52882 v2 The World Bank Study Croatia: Living Standards Assessment Volume 2: Promoting Social Inclusion and Regional Equity Background Papers Report No. 37992 November 2006 Report No. 37992 Croatia: Living Standards Assessment Volume 2: Background Papers November 2006 Poverty Reduction and Economic Management Unit Europe and Central Asia Region Document of the World Bank CURRENCY AND EQUIVALENT UNITS (as of May 30, 2006) Currency Unit = HRK 1 US$ = 5.67 HRK ABBREVIATIONS ALMM Active Labor Market Measures ALMP Active Labor Market Program ASSC Areas of Special State Concern BEEPS Business Environment and Enterprise Performance Survey CBS Central Bureau of Statistics CSW Centers of Social Welfare CARDS Community Assistance for Reconstruction, Development, and Stabilization CBN Cost of Basic Needs CES Croatian Employment Services HRK Croatian kuna DDP Development Data Platform ECA Europe and Central Asia EBRD European Bank for Reconstruction and Development EC European Commission FINA Financial Agency GDP Gross Domestic Product GNP Gross National Product GVA Gross Value Added NBS Household Budget Survey IT Information Technology IBRD International Bank for Reconstruction and Development IMF International Monetary Fund LFS Labor Force Survey NMS New Member States NUTS Nomenclature des Unites Territoriales Statistiques NGO Non-governmental Organization OECD Organization for Economic Co-operation and Development PEP Pre-Accession Economic Program PPP Purchasing Power Parity PPS Purchasing Power Standards ROP Regional Operations Program SME Small and Medium Enterprise UNESCO United Nations Educational, Scientific, and Cultural Organization USD US dollar Vice President : Shigeo Katsu Country Director : Anand K. Seth Sector Director : Cheryl W. Gray Sector Manager : Asad Alam Task Leader : Salman Zaidi ii TABLE OF CONTENTS BACKGROUND PAPER #1 POVERTY ESTIMATION: METHODS AND MEASUREMENT ISSUES..................................................... 1 INTRODUCTION ............................................................................................................................................................................ 1 DATA SOURCE ............................................................................................................................................................................. 1 CONSTRUCTION OF CONSUMPTION AGGREGATE ........................................................................................................................ 2 EQUIVALENCE SCALE .................................................................................................................................................................. 9 ESTIMATION OF THE POVERTY LINE......................................................................................................................................... 11 FINAL REMARKS ....................................................................................................................................................................... 15 BACKGROUND PAPER #2 POVERTY COMPARISON FOR CROATIA 2002-04 Danijel Nesti.................................................. 21 REGIONAL INEQUALITIES.......................................................................................................................................................... 28 INTERNATIONAL COMPARISONS ................................................................................................................................................ 31 CONCLUSION ............................................................................................................................................................................. 33 BACKGROUND PAPER #3 A POVERTY PROFILE FOR CROATIA 2004 ......................................................................... 45 THE CHARACTERISTICS OF POVERTY ....................................................................................................................................... 46 WHO AND WHERE ARE THE POOR? .......................................................................................................................................... 50 A Digression: International Poverty Comparisons............................................................................................................... 53 Location and Poverty .............................................................................................................................................................. 54 THE DETERMINANTS OF POVERTY ............................................................................................................................................ 61 The Method.............................................................................................................................................................................. 61 Main Findings.......................................................................................................................................................................... 62 SUMMARY AND CONCLUSIONS .................................................................................................................................................. 70 APPENDICES ......................................................................................................................................................................... 72 APPENDIX A: The Estimated Consumption Function ....................................................................................................... 72 APPENDIX B: STATISTICAL APPENDIX .................................................................................................................................. 74 BACKGROUND PAPER #4 REGIONAL DEVELOPMENT AND SOCIAL INDICATORS IN CROATIA ............................................... 91 INRODUCTION ............................................................................................................................................................................ 91 REGIONAL TRENDS IN THE NMS10 COUNTRIES AND EU15 ...................................................................................................... 91 DEMOGRAPHIC AND ECONOMIC CHARACTERISTICS OF THE CROATIAN REGIONS .................................................................... 97 SOURCES OF GROWTH OF THE CROATIAN COUNTIES ­ REGIONAL ECONOMY STRUCTURE ACCORDING TO SECTORS, ENTERPRISE SIZE AND OWNERSHIP.............................................................................................................................. 104 SECONDARY DISTRIBUTION OF INCOME .................................................................................................................................. 120 BACKGROUND PAPER #5 REGIONAL DISPARITIES IN LABOR MARKET PERFORMANCE IN CROATIA--ROLE OF INDIVIDUAL AND REGIONAL STRUCTURAL CHARACTERISTICS......................................................................................... 151 LABOR MARKET PERFORMANCE IN CROATIA ........................................................................................................................ 151 BACKGROUND PAPER #6 ASSESSING THE FLEXIBILITY OF THE CROATIAN LABOR MARKET ............................................ 185 INTRODUCTION ........................................................................................................................................................................ 185 OVERVIEW OF THE MAJOR TRENDS IN THE LABOR MARKET ................................................................................................... 185 RECENT LABOR MARKET INDICATORS IN COMPARATIVE PERSPECTIVE ............................................................................... 189 LABOR MARKET INSTITUTIONS AND UNEMPLOYMENT .......................................................................................................... 191 EMPLOYMENT PROTECTION LEGISLATION............................................................................................................................. 192 PASSIVE LABOR MARKET POLICIES ......................................................................................................................................... 195 ACTIVE LABOR MARKET POLICIES ........................................................................................................................................ 198 TAXES AND LABOR COSTS....................................................................................................................................................... 201 FLEXIBILITY AND LABOR MARKET FLOWS ............................................................................................................................ 203 LABOR MARKET ADJUSTMENT IN INTERNATIONAL PERSPECTIVE ......................................................................................... 205 LABOR MARKET ADJUSTMENT IN CROATIA ........................................................................................................................... 208 CONCLUSIONS ......................................................................................................................................................................... 218 References .............................................................................................................................................................................. 219 List of Tables Table 1.1: Estimated Deterioration Rate of Consumer Durables ...................................................................................................... 8 Table 1.2: Composition of Household Expenditures ........................................................................................................................ 9 Table 1.3: Nutritional Requirements and Equivalence Scale .......................................................................................................... 12 Table 1.4: Composition and Cost of Minimum Foods Basket by Product Groups ......................................................................... 13 Table 1.5: Poverty Lines using Various Equivalence Scales .......................................................................................................... 14 Table 2.1: Poverty Incidence, 2002-2004 ....................................................................................................................................... 22 Table 2.2: Changes in Poverty ........................................................................................................................................................ 23 Table 2.3: Decomposition of Changes in the Headcount Poverty Rates, in percent ....................................................................... 23 Table 2.4: Growth Effect on Poverty Rates .................................................................................................................................... 24 Table 2.5: Income Poverty Incidence, 2002-2004 .......................................................................................................................... 24 iii Table 2.6: Overlap between Income and Consumption Poverty, 2004 ........................................................................................... 26 Table 2.7: Poverty Rates Based on Implicit Social Welfare and Trade Union Poverty Lines, 2004............................................... 26 Table 2.8: Consumption and Income Inequality Statistics .............................................................................................................. 27 Table 2.9: Definition of Analytical Regions ................................................................................................................................... 28 Table 2.10: Poverty Incidence by Regions, 2002-2004 .................................................................................................................. 29 Table 2.11: Poverty Risk by Regions, 2002-2004........................................................................................................................... 29 Table 2.12: Poverty Risk by Regions and Urbanization Level, 2002-2004 .................................................................................... 30 Table 2.13: Consumption and Income by Regions, 2002-2004 ...................................................................................................... 30 Table 2.14: Average Consumption and Food Share........................................................................................................................ 32 Table 2.15: International Comparison of Poverty and Inequality ................................................................................................... 33 Table 3.1: Characteristics of Poverty .............................................................................................................................................. 49 Table 3.2: Estimates of Absolute Poverty for Croatia 2004............................................................................................................ 51 Table 3.3: Poverty Bands................................................................................................................................................................ 52 Table 3.4: Relative Poverty Risk by Region ................................................................................................................................... 65 Table 3.5­ Relative Poverty Risk by Educational Attainment of the Head of the Household ........................................................ 66 Table 3.6­ Relative Poverty Risk by Age of the Head of the Household ....................................................................................... 67 Table 3.7: Relative Poverty Risk by Employment Status of the Head of the Household................................................................ 68 Table 3.8: Relative Poverty Risk by Household Size ..................................................................................................................... 69 Table 4.1: Average Annual GDP p.c. PPS (1995-2002), Unemployment (1999-2003), Employment (1999-2003) and Population (1995-2002) Growth Rates................................................................................................................................................... 94 Table 4.2: Ratio between NUTS II Region with Highest/Lowest GDP p.c. PPS, 1995-2002......................................................... 95 Table 4.3: Real GDP Growth Rates of the EU25 According to the NACE Classification Activities, 2000-2003 .......................... 96 Table 5.4.4: Demographic Structure of Croatia, by County (NUTS III)......................................................................................... 97 Table 4.5: Demographic Structure of Croatia by Analytical Region .............................................................................................. 98 Table 4.6: Education by County, 2001, in % .................................................................................................................................. 99 Table 4.7: Education by region, 2001, in %.................................................................................................................................... 99 Table 4.8: Gross Domestic Product per capita, by County, Croatia =100..................................................................................... 100 Table 4.9: Gross Domestic Product per capita by regions, Croatia = 100..................................................................................... 100 Table 4.10: Employment Structure by Economic Activity and Unemployment Rate, by County, 2003, in % ............................. 101 Table 4.11: Employment Structure by Economic Activities, by Analytical Regions, 2003, in % ................................................ 101 Table 4.12: Estimation of Labor Productivity by Economic Activities, by Analytical Regions, 2003, Croatia = 100.................. 102 Table 4.13: Infrastructure Development Indicators ...................................................................................................................... 103 Table 4.14: Overview of Correlation Coefficients between County GDP p. c. and Various Variables ........................................ 104 Table 4.15: Regional Distribution of GVA, Current Prices in the Period 2001-2003, in %.......................................................... 106 Table 4.16: Gross Value Added Average Nominal Growth Rate, Current Prices, in the Period 2001-2003, in % ....................... 107 Table 4.17: Contribution to Total Gross Value Added Growth, Current Prices, in the Period 2001-2003, in % of Total Nominal GVA Increase..................................................................................................................................................................... 108 Table 4.18: Average Proportions in the GVA of Various Unit Groups According to Size in the Period 2001-2003.................... 110 Table 4.19: Average Nominal GVA Growth Rates in the Period 2001 - 2003, in % .................................................................... 112 Table 4.20: Estimated GVA Proportion According to the Ownership Structure in the Period 2001-2003, in % of Total GVA for the County .......................................................................................................................................................................... 114 Table 4.21: Average Annual Nominal GVA Growth According to the Ownership Structure in the Period 2001-2003 ............... 115 Table 4.22: Average Annual Growth Rates: GDP p.c. (2001-03), Population (2001-03) ............................................................. 117 Table 4.23: Underground Economy Size in Various Counties in Croatia (lower estimation boundary ­ Eurostat approach), 2002, in HRK thousands .............................................................................................................................................................. 118 Table 4.24: The Correlation Coefficients between Certain Variables and County Proportions of the Underground Economy in the GDP Figure ........................................................................................................................................................................ 119 Table 4.25: Gross Disposable Income of Households per capita, by county, Croatia =100.......................................................... 121 Table 4.26: Some Derivative Indicators on Disposable Income by Households, Primary Income and GDP by counties, ratio in % ........................................................................................................................................................................................... 122 Table 4.27: Some Derivative Indicators on Disposable Income by Households, Primary Income and GDP by analytical regions, ratio in %............................................................................................................................................................................ 123 Table 4.28: Social Transfers Relations to Primary Income, Disposable Income and GDP by counties, in % .............................. 124 Table 4.29: Social Transfers Relations to Primary Income, Disposable Income and GDP by Analytical Regions, in % ............. 125 Table 4.30: Pension Income Regional Distribution in Croatia...................................................................................................... 126 Table 4.31: Number of Social Welfare Beneficiaries, by counties, 2003 ..................................................................................... 127 Table 4.32: Number of Social Welfare Beneficiaries, by Analytical Regions, 2003 .................................................................... 128 Table 4.33: Number of Beneficiaries of Unemployment Benefits by Counties ............................................................................ 129 Table 4.34: Number of Beneficiaries of Unemployment Benefits by Analytical Regions............................................................ 129 Table 4.35: Child Allowance, by counties, 2001-2003 ................................................................................................................. 130 Table 4.36: Child Allowance, by Analytical Regions, 2001-2003................................................................................................ 130 Table 4.37: The Correlation Coefficients between Social Transfer Variables and the Level of Economic Development ............ 131 iv Table 5.1: Labor Force Participation Rate, Employment Rate, and ILO Unemployment Rate in Croatia (2002-2004) ............... 153 Table 5.2: Employment Rate, Unemployment Rate and Monthly Earnings, ................................................................................ 155 Table 5.3: Employment Rate, Unemployment Rate and Monthly Earnings ................................................................................. 155 Table 5.4: Employment Rate, Unemployment Rate and Monthly Earnings by Gender in Croatia (2002-04) .............................. 156 Table 5.5: Summary Statistics on Labor Market Indicators by region in Croatia (2002-04) ........................................................ 158 Table 5.6: Summary Statistics on Labor Market Indicators by County in Croatia (2002-04)....................................................... 159 Table 5.7: Inequality of Earning at the County Level in Croatia (2002-04).................................................................................. 159 Table 5.8: Determination of Employment and Earning at the National Level in Croatia (2002-04)............................................. 164 Table 5.9: Determination of Employment and Earning at the Regional Levels ............................................................................ 165 Table 5.10: Ranking of Impacts of Years of School on Earning and on Employment Across Regions........................................ 166 Table 5.11: Determination of Employment and Earning at the National Level in Croatia (2002-04)........................................... 167 Table 5.12: Determination of Employment and Earning at the Regional and National Levels Using Level of Schooling as Indicator of Human Capital................................................................................................................................................ 169 Table 5.13: Decomposition of the Effects of Individual and Other Characteristics on Regional Earning Differentials ............... 170 Table 5.14: Simulations of the Effects on Employment and Earning of Nation-wide Education Policy by Region ..................... 172 Table 5.15: Simulations of the Effects on Employment and Earning of regional Specific Labor Market Policies by Region...... 174 Table 6.1: Decomposition of the Change in Employment (In Percentage Points) ....................................................................... 187 Table 6.2: Structure of Employment in Central and Eastern Europe and EMU (in %)................................................................. 188 Table 6.3: Shares of Gross Manufacturing Output by Industry .................................................................................................... 188 Table 6.4: Spending on Passive Labor Market Policies in Selected EU Countries and Croatia.................................................... 196 Table 6.5: Comparative Indicators of Unemployment Allowance (Late 1990's) .......................................................................... 196 Table 6.6: Spending on Active Labor Market Policies ................................................................................................................. 198 Table 6.7: Structure of Spending on Labor Market Policies ......................................................................................................... 199 Table 6.8: Structure of Spending on Active Labor Market Policies.............................................................................................. 200 Table 6.9: Average Tax Wedge (as percent of average production worker's total labor cost, 2003)........................................................ 201 Table 6.10: Labor Costs................................................................................................................................................................ 203 Table 6.11: Types of Labor Market Flexibility............................................................................................................................. 203 Table 6.12: EPL Index in Transition Countries ............................................................................................................................ 206 Table 6.13: Job Flows in Transition Countries (Faggio-Konings dataset) ................................................................................... 206 Table 6.14: Job Flows in Transition Countries (Rutkowski Dataset)............................................................................................ 207 Table 6.15: Correlation Coefficients of Job Flows with GDP and Employment Changes (1994-2004) ....................................... 209 Table 6.16: Decomposition of Excess Job Reallocation Arising from Shifts within and between Different Regions, Ownership Types, Size Classes and Economic Activities .................................................................................................................... 211 Table 6.17: Distribution of Tenures in Transition Countries ........................................................................................................ 213 Table 6.18: Hiring Rates, Separation Rates and Worker Turnover in Transition Countries ......................................................... 214 Table 6.19: Hirings and Separation Rates, According to Sectors of Ownership (2002) ............................................................... 215 Table 6.20: Hirings and Separation Rates, According to Age Groups .......................................................................................... 215 Table 6.21: Employed Persons According to their Formal Status................................................................................................. 217 List of Figures Figure 1.1: The Change in Value of a Durable Good Due to Deterioration...................................................................................... 4 Figure 1.2: A Glance at the Household Consumption per Equivalent Adult Expenditure .............................................................. 11 Figure 2.1: Growth Incidence Curve, 2002-2004............................................................................................................................ 25 Figure 3.1: Expenditure Patterns of the Poor and the Nonpoor....................................................................................................... 47 Figure 3.2: Nutritional Assessment of the Diet of the Poor ............................................................................................................ 48 Figure 3.3: Sources of Income: A Comparison between Poor and Nonpoor Households............................................................... 48 Figure 3.4: Headcount Poverty Ratio as a Function of the Poverty Line ........................................................................................ 53 Figure 3.5: Cross-country Poverty Comparisons ............................................................................................................................ 53 Figure 3.6: Poverty Incidence in Croatia by Region ....................................................................................................................... 54 Figure 3.7: Distribution of Poverty by Region................................................................................................................................ 55 Figure 3.8: Poverty Incidence by Age of the Household Head ....................................................................................................... 55 Figure 3.9: Poverty Incidence over the Life Cycle ......................................................................................................................... 56 Figure 3.10: Incidence of Poverty and Pension Receipts among the Elderly.................................................................................. 57 Figure 3.11: Poverty Share by Age of the Household Head ........................................................................................................... 57 Figure 3.12: Poverty Incidence by Household Size and Urban/Rural Areas.................................................................................. 58 Figure 3.13: Distribution of the Poor by Household Size and Urban/Rural Areas.......................................................................... 58 Figure 3.14: Incidence of Poverty by Educational Attainment of the Household Head.................................................................. 59 Figure 3.15: Poverty Shares by Educational Attainment of the Household Head........................................................................... 59 Figure 3.16: Poverty Incidence by Employment Status of the Head of the Household................................................................... 60 Figure 3.17 ­ Poverty Shares By Employment Status of the Head of Household........................................................................... 60 Figure 3.18: Standardized Simulated Relative Poverty Risks by Region........................................................................................ 63 Figure 3.19: Standardized Simulated Relative Poverty Risks by Education of Household Head ................................................... 64 v Figure 4.1: Average Annual Real GDP Growth Rates of EU25 NUTS II Regions, ....................................................................... 93 Figure 4.2: The Relationship between the Proportion of Tertiary Sector...................................................................................... 109 Figure 4.3: Relationship between he Average Nominal Growth rate of the Total GVA and the proportion of Market Small and Middle Entrepreneurs in Total Gross Value Added ........................................................................................................... 113 Figure 4.4: The Relationship between the Gross Value Added Growth Rate and Private Sector Share ....................................... 115 Figure 4.5: Underground Economy Size According to County, 2002 (Eurostat Approach) ......................................................... 117 Figure 4.6: The Relationship between Social Transfers and Economic Development of Croatian Counties................................ 125 Figure 4.7: The Relationship between Pensions and Economic Development of Croatian Counties............................................ 127 Figure 4.8: The Relationship between Social Transfers, without Pensions and Economic Development of Croatian Counties... 127 Figure 5.1: Total Employment Rate of EU Countries (2002-2004) .............................................................................................. 152 Figure 5.2: Distribution of Nominal Monthly Earning of the Individuals Who are Employed (2002-2004) ................................ 154 Figure 5.3: Distribution of Human Capital by Region (2002-2004) ............................................................................................. 157 Figure 5.4: The Distribution of Monthly Earning in the County of Zagreb and the County of Lika-Senj (2002-04) ................... 160 Figure 5.5: Distribution of Earning by Region in Croatia (2002-2004) ........................................................................................ 160 Figure 5.6: Distribution of Earning by Age Group in Croatia (2002-2004).................................................................................. 161 Figure 5.7: Distribution of Earning by Education Group in Croatia (2002-2004) ........................................................................ 161 Figure 5.8: Distribution of Earning by Gender in Croatia (2002-2004)........................................................................................ 162 Figure 6.1: Dynamics of GDP, Average Net wage and Employment (1990=100) ....................................................................... 186 Figure 6.2: Administrative (registered) and Survey Unemployment Rates, 2004......................................................................... 189 Figure 6.3: Overall and Long-term Unemployment, 2004............................................................................................................ 190 Figure 6.4: EPL Indices - International Comparison .................................................................................................................... 193 Figure 6.5: EPL Configuration - Subindices for Temporary and Permanent Employment........................................................... 193 Figure 6.6: EPL Sub-index for Permanent Contracts and Temporary Employment (as a share of total, 2003) ........................... 194 Figure 6.7: Tradeoff between EPL and Passive Labor Market Policies........................................................................................ 197 Figure 6.8: EPL Index and Job Flows (Faggio-Konings dataset).................................................................................................. 207 Figure 6.9: GDP Change, Employment Change and Job Flows in Croatia................................................................................... 208 Figure 6.10: EPL Index and Worker Turnover (Cazes-Nesporova dataset + own calculations) .................................................. 214 vi MAIN TOPICS COVERED IN THE BACKGROUND PAPERS: A SHORT ROAD-MAP Six background papers have been prepared as part of the overall Croatia Living Standards Assessment work program, and are included in this volume. 1: Poverty Estimation for Croatia: Methods and Measurement Issues (Danijel Nesti and Giovanni Vecchi): The paper discusses the methodological steps involved in estimating poverty rates for Croatia on the basis of the 2004 Household Budget Survey. Part B of the paper deals with the construction of the consumption aggregate, the summary measure used to proxy living standards. Special attention is devoted to the estimation of the consumption flow from durable goods. Section C describes the choice of the equivalence scale used to adjust the consumption aggregate for differences in household composition. Finally, section D details the procedures followed in developing an absolute poverty line for Croatia, following Ravallion (1994). The poverty line derived using the methodology described in this paper is then held constant in real terms (i.e. adjusted only for inflation) and used in #2 below to derive poverty estimates for Croatia during the period 2002 ­ 2004. 2: Poverty Comparisons For Croatia: 2002-2004 (Danijel Nesti): Using data from the Household Budget Survey (HBS) series, this paper presents poverty comparisons for Croatia for the period 2002 ­ 2004. It focuses on three types of comparisons (i) over time, (ii) across regions, and (iii) across countries (i.e. international comparisons). Poverty estimates across regions are based on pooled data from the three HBS rounds to improve the precision of derived estimates. Part B provides poverty estimates for 2002-2004 (based on consumption as well as income) and assesses changes over time. Part C examines regional variation in living standards (both at the county-level as well 5 main analytic regions), while Part D provides poverty and inequality comparisons between Croatia and other counties of Central and Eastern Europe. 3: A Poverty Profile for Croatia (Giovanni Vecchi): Using data from the 2004 HBS, this paper develops a profile of the poor in Croatia. The focus is on two questions: (i) who and where are the poor?, and (ii) what are the micro-determinants of poverty? Section B provides a description of the main characteristics of the poor in terms of a number of both monetary and non-monetary indicators. Section C constructs the poverty profile for Croatia: in addition to estimating the scale of poverty (how many the poor are and how poor they are) it identifies the population groups most at risk of poverty. Section D analyzes the main determinants of consumption (and hence poverty) using a regression model of log per adult-equivalent consumption that includes a number of poverty correlates. The parameter estimates derived are then used to simulate the poverty rates that would be observed if households were given certain socio-demographic characteristics. Comparing simulated and actual poverty rates then helps assess the relative importance of different determinants of poverty in Croatia. 4: Regional Development and Social Indicators in Croatia (Zeljko Lovrincevi and Davor Mikuli): This paper provides a comprehensive profile of social and economic characteristics of Croatia's regions at NUTS III level (and also the five main analytic regions used in other background papers). Part B briefly reviews the experiences of EC new member states with regional development, based on an analysis of GDP dynamics, employment, unemployment, and population at the NUTS-II level. Part C reviews the demographic and economic characteristics of Croatia's region in relation to these main dimensions. Part D analyzes the relationship between economic growth of certain counties on the one hand, and regional economic structure according to sectors, enterprise size, and ownership on the other. Part E focuses on analyzing the distribution of GDP per capita, gross disposable income, primary and secondary incomes, thus highlighting the influence of redistribution policies in Croatia in equalizing incomes across counties and regions. The paper includes two appendices: the first presents regional GDP by counties for 2001-2003, while the second provides preliminary data on gross disposable income of the household sector in Croatia. 5: Regional Disparities in Labor Market Performance in Croatia (Xubei Luo): Using data from the Labor Force Survey (LFS) series, this paper reviews the labor market performance in Croatia for the period 2002 ­ 2004, at both the national as well as regional level. Section B describes the labor market performance in Croatia, and reviews disparities across individual groups and regions. Part C analyzes the main determinants of employment and earnings at the national and regional levels, focusing in particular on both individual as well as regional structural characteristics. Section D simulates the effects of nationwide educational policy and regional specific labor market policy on labor market developments in Croatia. The paper includes two appendices: the first presents more details pertaining to the various definitions used in the paper, while the second summarizes the main features of the 2002, 2003, and 2004 labor force survey data used in the analysis; this appendix also outlines the main rationale and validity of the data pooling exercise to improve the precision of the derived estimates. 6: Assessing the Flexibility of the Croatian Labor Market (Vedran Sosic): Using detailed firm-level data from the Financial Agency on job creation and destruction, this paper explores the extent of job and worker flows in the Croatian labor market over the period 1994 ­ 2004, and in particular analyzes trends by region, ownership type, firm size, economic activities. In addition, the first part of the paper provides a relatively comprehensive review of the main labor market features of Croatia, including an overview of employment protection legislation, active and passive labor market polices currently in place, as well as taxes and labor costs on labor. The main features of labor market adjustment are discussed, both with regard to their evolution over time as well as in light of similar experiences in other transition countries in the region. vii BACKGROUND PAPER #1 POVERTY ESTIMATION: METHODS AND MEASUREMENT ISSUES Danijel Nesti1 (Ekonomski Institut, Zagreb) Giovanni Vecchi (Universitŕ di Roma "Tor Vergata") Abstract The paper discusses the methodological steps involved in estimating poverty rates for Croatia on the basis of the 2004 Household Budget Survey. The first part of the paper deals with the construction of the consumption aggregate, the summary measure used to proxy living standards. Special attention is devoted to the estimation of the consumption flow from durable goods. The second part of the paper reviews the choices made for the construction of the absolute poverty line. INTRODUCTION The paper describes the main methodological issues underlying the estimation of poverty rates for Croatia on the basis of the 2004 Household Budget Survey. It focuses on (i) the construction of the consumption aggregate as a summary measure of living standards, and (ii) the estimation of the (absolute) poverty line. Both elements are at the core of the poverty measurement exercises carried out in two companion papers in this volume.2 This paper is organized as follows. Section B deals with the construction of the consumption aggregate. The key reference used throughout the section is Deaton and Zaidi (2002). Section C describes the choice of the equivalence scale used to adjust the consumption aggregate for differences in household composition. Section D details the procedure used to estimate the poverty line. The two single most important references in this section are Ravallion (1994) and Chen and Ravallion (1996). At various stages we have borrowed from the outstanding background work by Luttmer (2000). DATA SOURCE The main source of data for poverty analysis is the Household Budget Survey (HBS), conducted by the Croatian Bureau of Statistics (CBS) on an annual basis. The 2004 round of HBS is used to estimate the poverty line, as presented in this report. In 2004 the sample frame used for the selection of dwellings occupied by private households was based on the Census 2001 data. The sample was selected in two stages. At first, 26 samples were selected, each containing 13 1 The authors would like to thank Nicola Amendola and Salman Zaidi for many helpful comments. Useful comments were provided by the participants of a workshop held in Zagreb in Dec 2005. The usual disclaimer applies. Correspondence: dnestic@eizg.hr. 2 See "A Poverty Profile for Croatia 2004" (Chapter 2), and "Poverty Comparisons for Croatia 2002-2004" (Chapter 3). groups of neighbouring enumeration areas called segments, for each of 26 two-week periods, out of a set of 717 segments previously selected for the 2004 Labour Force Survey. In the second stage, out of 338 selected segments, 12 dwellings, occupied by private households, were selected, which were not previously selected into the 2004 Labour Force Survey. Thus, 4,056 dwellings occupied by private households were selected. For each selected occupied dwelling all private households were interviewed. 2,847 private households were successfully interviewed, corresponding to 8,222 individuals. Estimated population consists of 1,441,200 private households and 4,227,000 individuals.3 The HBS 2004 questionnaire consists of four forms: ˇ The Household Questionnaire refers to housing conditions, ownership of durable goods and expenditures on the number of non-food items. The reference period varies from one month to one year. ˇ The Household Members' Questionnaire gathers information on demographic characteristics, educational attainment and labor market status of household members, as well as their income (by source), and other information related to financial transactions (property sales, lending, savings). ˇ Diary ­ A form that households are expected to fill out daily (during a two-week period) as items are purchased. ˇ Questionnaire on the Consumption of Food, Drinks, Tobacco Products and Other Consumer Goods is a supplementary questionnaire used by households who fail to fill in the Diary. In our work, consumption is the main measure of material well-being and the basis for poverty estimation. Our preference of consumption over income is based on standard arguments, namely that (i) income is more prone to underreporting than expenditure; (ii) expenditure provides a better account of welfare in the presence of home-produced goods and other non-marketed transactions; (iii) expenditures are preferred to incomes when the size of grey economy is not negligible, and (iv) when income variability is high because of seasonal effects. All household expenditures were expressed in April 2004 prices by using Consumer Price Indices prepared by the CBS. CONSTRUCTION OF CONSUMPTION AGGREGATE The concept of consumption used in this study is similar to the concept of total household expenditures on final consumption applied in the Croatian national account system. However, in order to construct an appropriate measure of well-being, a number of adjustments are needed. In this section we describe how the main components of our baseline household consumption aggregate are defined. Food Expenditures Food consumption is calculated from the Diary and Supplementary Questionnaire. Our food aggregate includes both soft and alcoholic drinks. There is information on about 100 food and beverage items. Total food consumption includes (a) actual spending on food; (b) estimated value of home-produced food, and (c) estimated value of food gifts received. Values of food given in private transfers are not included in consumption. In estimation of the value for home produced foods and gifts, information on quantities is provided by households themselves, while CBS (the interviewers and the controllers during their field work) impute price information and calculate the value of food consumed. Original survey data on food are collected for a two-week consumption period. Yearly figures are calculated by multiplying the two-week values by 26. However, using this method, unusually large acquisitions of certain 3 Further details are available online at http://www.dzs.hr/Eng/2005/13-2-1_1e2005.htm. 2 items, either due to the seasonal nature of consumption or the building of stocks, could flaw comparisons among households. To neutralize this effect, at least in the most extreme cases, we have checked for large expenditures on individual items. Per capita consumption of every item in excess of 10 times the median of per capita consumption has been censored to the threshold of 10 times the average value. Although this is a somewhat arbitrarily procedure, similar "top coding" procedures were adopted by Gottschalk and Smeeding (1997) applied to income survey data. "Top coding" is also used in World Bank (2005). Housing Consumption related to dwelling consists of two parts: (i) rental value of the main residence, and (ii) expenditures for utilities. The rental value of the main residence is calculated as: ˇ Self-reported rental value for both owners and tenants with subsidized housing ˇ Actual rent paid by tenants ˇ Predicted rental value (hedonic regression) for missing observations. The main information on rents is the self-reported rental value of occupied dwellings. About 86 percent of households live in owner-occupied dwellings, an additional 11 percent of households are tenants with subsidized housing costs. Households belonging to both categories are asked to self-report the rental value of the occupied dwelling. Only 2.5 percent of households have reported market-based rent. In Croatia, the market for rentals is rather shallow, concentrated in large cities, and, thus, can not guarantee a reliable estimation of the imputed rent. Evaluation of the expenses incurred in buying/building a rental unit are also made difficult by very high past inflation rates, and the practice of continuous re-building of the unit with the help of family members and friends. For all of these reasons, we argue that self-reported rental value provides the best basis for estimating the rental value of missing observations.4 Rental value for missing observations (142 cases out of 2,847) was predicted after estimating a hedonic regression. The log of self-reported rental value was regressed on a number of covariates (about 20 indicators, including type of dwelling, year of construction, amenities, region etc.). Regression output is shown in Appendix I. About 10 percent of households report that they own a second house. The rental value of a second dwelling was excluded from the consumption aggregate, for two main reasons. First, the questionnaire does not report on the rental value of second dwellings. Second, the characteristics of these dwellings are poorly described in the HBS. Expenditures for utilities (both for main and second dwellings) were included in the consumption aggregate. More precisely, we included expenditures for electricity, gas, heating oil, water, central heating, waste removal, and other minor related expenses. Materials and services for household repairs and improvements were considered as an investment and not included in consumption. Taxes on houses were also excluded. In the presence of implausibly large values of expenditures on utilities we applied the rule described above (food expenditures), and replaced outlying values with the censored value of 10 times the average expenditure. 4 This is so even in spite of the fact that estimated rental value of the main residence is now approaching a high fraction of overall consumption, around 20 percent, as showed in Table 2.1 (see below) 3 Consumer Durables Durable goods require special treatment when constructing the consumption aggregate. The problem arises from the fact that it is not the purchase of durable goods that contributes to welfare, but their use. Thus, instead of using the current expenditures on durable goods service flows streaming from the goods' use need to be estimated. In order to estimate this component, different options are available in theory, although in practice, the choice usually depends on the data at hand. In the ideal case, the estimation of consumption flows from durable goods requires both the historical price of the good (that is, the price at which the good was purchased T years before) and the price at which the good could be sold at the time of the interview. Unfortunately, the Croatian HBS does not collect such information. What the 2004 questionnaire does report is (i) the vintage of durable goods (ownership relating to up to three durable goods is recorded in the questionnaire) ­ this information is reported by all households for all the owned durable goods ­ and (ii) the expenditure on (new) goods during the survey year. Information on expenditure is only available for a subset of households, that is, households that purchased durable goods during the last 12 months. The information gathered in the 2004 HBS therefore severely limits our ability to estimate the consumption flows of durable goods. We have therefore devised the following ad hoc procedure, which is best illustrated with the aid of Figure 1.1. On the y-axis is the value of the k-th durable good. On the x- axis is time, denoted by t. The t index runs from 0 (the age of a new good) to Tmax, which represents the total lifetime of the k-th good (i.e. the maximum age up to which the good delivers a strictly positive consumption flow). If the household owns a durable good older than Tmax, we decided, somewhat arbitrarily, to impute that good's age at exactly Tmax. How to calculate the consumption flow associated with (the ownership of) good k? In order to accomplish this task, we need to identify the dynamics of the good's value from the time it was manufactured (t=0) to the last year of its economically active life (t=Tmax). In Figure 1.1 we have denoted by Ak the price of one unit of a new good, and by Bk its value in the last year before becoming obsolete. The identification of the Ak-to-Bk path delivers the deterioration rate , which is key for imputing the consumption flow of durables owned by the households. Figure 1.1: The Change in Value of a Durable Good Due to Deterioration Ak Bk 0 Tmax The estimation of the Ak-to-Bk curve in Figure 1.1 requires the knowledge of four elements: (a) Ak ­ the price of one unit of the k-th good at the beginning of the year 2004 (b) Bk ­ the price of the good (at the end of) the last year of its economic life (c) Tmax ­ the total lifetime of the good (this parameter is good-specific, but we omit the k suffix to keep notation to a minimum) (d) The shape of the curve connecting Ak to Bk. 4 Starting from (a), two candidate methods for estimating Ak stand out. Method 1 consists in estimating Ak on the basis of the sample data: we use the expenditures reported by (the subset of) households who purchased one unit of the good during the last 12 months. In contrast, method 2 relies on extra-sample data, namely, the prices collected by the CBS for over 400 items (these are the prices that enter the CPI). Method 1 has the advantage that `households know better than CBS', i.e. reported expenditures are likely to control for the quality of owned goods better than others (including analysts). Method 1 has three major disadvantages. First, it can be argued that estimated prices are affected by self-selection bias; method 1 averages prices over households who reported exactly 1 purchase in the current year: to the extent to which the frequency of purchases is positively correlated with income, method 1 is likely to deliver prices for higher-than-average quality goods. Second, method 1 can be affected by small sample size problems: the number of households over which prices are averaged can amount to twenty or below, which implies a high proneness to misleading inference and fragility of the estimates. Third, method 1 can imply poor and/or partial coverage. This is owing to the fact that the HBS does not collect expenditures for all the durable items for which information on ownership and age is available. Out of 24 goods whose ownership is known, only 6 show expenditure-data exclusively related to them, while for others, information on expenditures is related to larger groups of goods disabling thereby a conclusion on the price (value) of a single item. The problems discussed in the above paragraph are absent in method 2. The main disadvantage associated with method 2 is, however, the potential lack of specificity, that is, the goods priced by the CBS could be different from those consumed (and reported) by the households surveyed. This seems to be a minor risk and we therefore adopt method 2. As far as the choice of Bk is concerned we have two "natural" options. One is Bk = 0. Choosing Bk = 0 makes sense logically, is tremendously intuitive, but is not convenient, as will be shown shortly. A more convenient choice is Bk = Ak, with alpha discretionary small (reasonable values being presumably in the range of 2.5 to 5 percent). This choice is also sensible, and can be justified by arguing that it accounts for the presence of transaction costs ( being the fraction of the transaction costs associated with the sale of the good at price Bk). The main disadvantage of this option is that the choice of is arbitrary. Preliminary runs show that the choice of affects the estimate of the consumption flow only by a small fraction. All in all, we decided to use =0.05 (i.e. Bk takes the value of 5 percent of Ak). Thirdly, the estimation of Tmax ­ the maximum age of the k-th good. Again, two options are discussed: Option 1 is outlined in Deaton and Zaidi (2002: 34) and summarized by the following formula: (1) Tmax = 2 × T , where T-bar is defined as the average vintage of the k-th good, and is computed through a weighted sample mean of reported ages of the k-th good over owner households. Option 2 consists of using some outlier-resistant statistics computed over the set of the most long-lived goods in the sample: one possibility, for instance, is the 95th percentile (denoted by T.95) of the sample distribution of the ages of the k-th good: (2) Tmax = T.95 Any other percentile will do, of course. All in all, we prefer (2) to (1). Again, the choice is an empirical matter and should ideally have no impact on the final result. 5 Finally, the choice of the shape of the A-to-B curve in Figure 1.1. From the outset, we ruled out the possibility of a linear deterioration rate. This is an unnecessarily unrealistic assumption for most durable goods. Typically, the value of the good decreases steeply immediately after its purchase, and undergoes a slower process of de-valuation in subsequent years. This path is qualitatively described by the solid line in Figure 1.1. It is easy to see that Ak takes a short time to halve, but a considerably longer time to reach its final value. Translating the curve in Figure 1.1 into mathematical formulae is useful, reasonably simple, and most of all necessary in order to make the procedure operational. We start from the following equation that links the price of a new good bought at the beginning of period t (we denote this price by pt0 ) with the unobserved price at time t (calendar year) of the same good purchased T years before (we denote this by ptT ): ptT = p t0 (1 - ) T (3) 1.1 From (3) the deterioration rate can be written as: 1 pT T (4) = 1 - t0 p t Equation (4) implies a path for the deterioration rate that is sometimes referred to as radioactive decay. The problem with equations (1) and (2) is that ptT is not observed in the case of Croatian HBS. However, by assuming (i) knowledge of the total economic lifetime (Tmax) of the good, and (ii) the fact that in the last year's value is equal to a small fraction of its initial value: (5) ptTmax = pt0 We are then able to simplify (4) dramatically. To see how, substitute (5) into (4) to get: 1 1 pt0 T max (6) = 1 - 0 = 1- Tmax p . t A main feature of equation (6) is that it does not depend on the purchase value of the good, which implies that the depreciation rate can be calculated by using data for all owner households and not only for those which purchased one good in the last 12 months. Hence, no selectivity bias is threatening our results. Once (6) has been calculated (separately for each durable good), the consumption flow accruing to the h- th household can be obtained as follows. Suppose that household h owns good k, which has an age of Th. The consumption flow can be calculated as: (7) C kh = p tTh ( k + r ) , ,k where r, the real interest rate, reflects a household's foregone earnings if it had invested money, for example, in bonds, a bank account or other low-risk investments. The fact that a household decided to own a durable good instead of putting money in alternative investment reveals that ownership gives at 6 least equal return as the alternative investment. In our case, we decided to estimate real investment rate as nominal yield on HRK denominated 5-year bond, as of in April 2004, minus inflation rate about that time, which yields the rate of 4 percent.5 Maintenance and repair expenditures on durable goods are not included in consumption flows.6 pT Although t in (5) is not known directly from the survey data (subscripts k and h can be neglected for a while), it can be easily calculated from (1), once we have estimated from (4) and calculated r as explained above. r is not good specific, but could vary with goods, depending on good specific Tmax ( is chosen to be the same for all goods). One more observation on p t0 (price of a new good) and ptT (price (value) of a good of age T) in line with the section explaining Ak.7 For Ak (or p t0 ) we decided to use the prices collected by the CBS for the purpose of CPI calculation. It is clearly not household specific, and this is a reasonable choice if we do not have information on the actual price or quality of the good possessed. Starting from the CBS price, and applying (4) we can calculate ptT . In some specific cases, when a household reports possession of exactly one unit of the good and reports non-zero expenditure on that consumer good in the last 12 months, it could be supposed that reported expenditure represents the current value of that good ( ptT or pt0 , depending on the age of the good bought). In that case, we calculate the consumption flow from (5) using expenditure data supplied by households as price of the good. Table 1.1 shows estimation results regarding the depreciation rate , for durable goods for which ownership information is available in 2004 HBS. A few durable goods are merged together in order to estimate the consumption flow (for example, cars and caravans, as well as motorcycles and scooters) because of assumed similarity of their characteristics and prices. Although ownership information is available in HBS, estimation of consumption flows from yachts and boats is not included in the consumption aggregate due to lack of reliable price information and a small number of observations (only 2 households report expenditures on new boats, and only 20 report possession of boat). Ownership information on music boxes, DVD players, mobile phones, video camera and vacuum cleaners is available in the 2004 HBS, but not earlier. In order to make comparable estimates of consumption flows from durables in the entire 2002-2004 period, we chose to exclude estimates for the above five durable items8. 5 In 2004, the 5-year HRK denominated government bond was security of the longest maturity in the market. 6 One reason for that is double counting: repair expenditures are already being imputed as part of the consumption flow. A second reason is that repair expenditures often increase the value of the good, and can be therefore seen more similar to investment than consumption. A further reason is the next-to-lumpy nature of these expenditures which make them pretty noisy indicator. 0 7 By now it should be clear that we are using p t and Ak interchangeably, to denote the value of a new unit of good k. 8 When we estimated service flow for durables in 2002 and 2003, we use exactly the same deterioration rate as in 2004, and current CBS prices. 7 Table 1.1: Estimated Deterioration Rate of Consumer Durables Purchase price (Ak) Deterioration No. of obs. Lifetime (kunas) rate Good purchase of a new od in last 12 months) median mean Tmax = T.95 k k k HBS CBS (%) Cars&Caravans 60 99,500 73,839 20 13.9 Motrocycles&Scooters 8 18,072 16,381 24 11.7 TV set - - 2,114 19 14.6 Radio set - - 636 29 9.8 HiFi set - - 1,713 17 16.2 Video recorders - - 1,025 16 17.1 Personal computers 90 5,988 4,261 10 25.9 Washers 87 2,277 2,356 22 12.7 Washer&Dryers** 8 4,300 - 22* 12.7 Dish washer 28 2,970 3,094 22* 12.7 Freezer 62 2,096 2,688 24 11.7 Refrigerator 109 2,393 2,590 24 11.7 Microwave - - 892 15 18.1 Notes: * Washer-dryer machines and dishwashers are relatively new goods for Croatian consumers. This implies that estimation of lifetime of these goods is based on the actual age of goods currently possessed (or more precisely age of 95th percentile) underestimates the true lifetime. Instead, estimation of in these cases is based on the estimated lifetime of similar goods; for both washer-dryer and dish washers that of washing machines. **There is no CBS price data for washer-dryers, but median price from HBS seems reasonable in spite of the fact that the result is based on the small number of observations. Therefore, median price is used in the estimation of deterioration rate and consumption flow. Other Non-food Expenditures The list of non-food items other than durable goods and housing included in the consumption aggregate includes: ˇ Clothing and Footwear ˇ Non-durable house furnishings (without expenditures for the purchase of durable goods) ˇ Non-durable transport equipment and services (mostly fuel and grease (but excluding expenditures t due to car use for business purposes), bus, taxi and ferry costs) ˇ Communication (expenditures on postal, phone and fax services, but excluding purchase of phone sets) ˇ Non-durable goods for recreation and culture ˇ Education ˇ Hotels and restaurants ˇ Miscellaneous goods (personal services, articles of personal care, clocks, watches, sunglasses, umbrellas, personal insurance against accidentals, voluntary health and pension insurance, car insurance, and house insurance). Maintenance and repair costs of durable goods for which we did not impute consumption flow were included in consumption. The idea is that expenditures for repair represent a proxy of the user cost (cost of repairs captures the depreciation of the good) and should be included in consumption (such as, for example, expenditures for furniture repairs). Expenditures Not Consistently Related to Well-being Certain types of expenditures are not good indicators of well-being in the sense that higher expenditures do not correspond to higher level of well-being. The natural candidates for such types are health and funeral 8 expenditures. Larger health expenditures do not necessarily point to higher living standards. On the contrary, they can point to illness and thus, a serious loss of well-being. Health expenditures were therefore excluded from the consumption aggregate. This decision is also supported by the rather low elasticity (around 0.2) of equivalent expenditure on health with respect to equivalent total consumption. Kindergarten expenditures are commonly means-tested in Croatia, meaning they are subsidized and depend on household income level. In that way, lower kindergarten expenditures do not point to a lower well-being. Expenses on kindergarten were therefore excluded from our consumption aggregate. Expenses for family celebrations (weddings, etc.) were excluded because infrequent. Expenditures related to social protection services were also excluded from the consumption aggregate as they are discriminatory (that is, targeted at special population subgroups) and means-tested. Overview of the Consumption Aggregate Table 1.2 shows the main components of the consumption aggregate, together with some expenditure categories that are excluded. Total household expenditure averaged at 77,597 HRK per year will be used in this study as a baseline aggregate for poverty analysis. Expenditures on food and beverages absorb 29 percent of overall consumption. The share for housing is also 29 percent. The imputed consumption flows from durables with ownership information account for around 5 percent of the total consumption. Purchases of other non-durable, non-food items accounts for the remaining 37 percent of consumption. Expenditures on the purchase of durable goods are excluded from the consumption aggregate, as well as health expenditures and others that are poor indicators of well-being. The excluded expenditures correspond to 11 percent of e household consumption. Table 1.2: Composition of Household Expenditures Household consumption Percentage of total (kuna/year) consumption (%) Food & beverages 22,515 29.1 Housing expenditures 22,522 29.1 o/w rents 15,361 19.9 o/w utilities 7,161 9.3 Imputed consumption flow from durables 3,839 5.0 Other non-food expenditures 28,484 36.8 Total household expenditure 77,330 100.0 Durables included in imputed flow 4,157 5.4 Durables without ownership information 2,159 2.8 Health expenditures 1,642 2.1 Old-age care, kindergarten and funeral exp. 428 0.6 Total excluded 8,387 10.8 EQUIVALENCE SCALE In order to compare well-being among households of different size and composition we adjusted the consumption aggregate by dividing total household expenditure by the equivalent size of the household. The simplest way of calculating the equivalent household size is (8) Equivalent Size = Adults + Kids 9 The main limitation of the above equation is that it does not account for either economies of scale or differences in needs due to differences in age and/or other demographic characteristics. In response to the above limitation a better measure of the equivalent size was constructed in the early 1970s by the Organization for Economic Development and Cooperation (OECD): (9) Equivalent Size = 0.3 + 0.7 × Adults + 0.5 × Kids , where Adults stands for the number of individuals aged 14 and older) and Kids stands for the number of children aged 13 and below. In what follows we refer to equation (9) as the OECD-I equivalence scale. Recently, the OECD has introduced a modified equivalence scale, characterized by larger economies of scale: (10) Equivalent Size = 0.5 + 0.5 × Adults + 0.3 × Kids Equation (10) is widely used by Eurostat in calculation of comparable income, poverty and social exclusion indicators for EU countries ("Laeken" indicators). We refer to the scale in equation (10) as OECD-II equivalence scale. Another option is represented by a simple class of equivalence scales which can be described by the following formula: (11) Equivalent Size = ( Adults + Kids ) , where is a parameter between 0 and 1 to be chosen or estimated. We use this equivalence scale with set to 0.75. Equation (11) was used by the World Bank in a number of studies on poverty in transition countries (see World Bank, 2000a). We call this the WB ECA equivalence scale. Note that if is set to 0.5, the formula describes the equivalence scale used by Luxembourg Income Study (LIS). To optimize our choice, we have considered a number of criteria that could drive our choice. Economic theory offers limited guidance. The literature is deluged with studies on the topic, but there are no generally accepted methods to choose among different equivalence scales. The fact that there is not a "theoretically correct" equivalence scale implies that economic theory cannot be invoked in support of any specific equivalence scale. Do national and/or international institutions suggest/require European countries to adopt a specific equivalence scale? In particular, does the Eurostat recommend any specific equivalence scale? To the best of our knowledge, the answer to both questions is negative. Hence, compliance with international/national requirements does not drive our choice for Croatia. In the absence of strong arguments based on statistical and economical theory we were left with the arbitrary choice among common/best practices. One option is to review the choices made by reports for other countries and see whether any clear tendency can be identified. The World Bank (2005) used per capita expenditures, which is in line with the Deaton and Zaidi (2002: 47 and 51) recommendation, but does not help to solve our problem. Another option is trusting the analysis of de Vos and Zaidi's (1997) in the study on equivalence scale sensitivity of poverty statistics for the member states of the European Community. "Since the final answer to the question how equivalence scales should be determined cannot be given, the results of this paper are useful in providing poverty statistics for three equivalence scales, of 10 which the two extremes ­ the OECD scale and the subjective scales ­ can probably be considered as upper and lower bounds on the values the equivalence scale can plausibly take. As a pragmatic choice, the modified OECD scale appears to be a reasonable compromise." All things considered, we gave preference to the modified-OECD equivalence scale. In doing so, we complied with the current Eurostat and CBS practice, and have chosen a tool that can be easily explained to the public. Figure 1.2 illustrates the main features of the distribution of household consumption expenditures per equivalent adult. Figure 1.2: A Glance at the Household Consumption per Equivalent Adult Expenditure 0.00003 Kernel density estimate Normal density 0.00002 Density 0.00002 0.00001 0.00001 0.00000 50000 100000 Equivalent CONSUMPTION (EU scale) Note: The two vertical lines are the median and the mean of equivalent consumption (calculated over the whole sample). The richest 1 percent (30 observations out of 2,847) was dropped from the sample before estimating the kernel density. ESTIMATION OF THE POVERTY LINE In this study, estimation of the poverty line follows Ravallion's (1994) recommendations. The main idea is to define the absolute poverty line as the level of total consumption at which households spend just enough on food to afford the cost of a minimum required energy intake plus an allowance to meet basic non-food needs. In this section we describe the choice of the minimum food basket, the procedure used to estimate its cost, and the estimation of the non-food allowance necessary to calculate the total poverty line. The Choice of the Minimum Food Basket The first step in constructing the minimum food basket is to define the food energy requirements for individuals of different age and sex. Since there is no official nutritional standard for Croatia, we relied on the World Health Organization (1985) and FAO (2004) recommendations. The FAO nutritional standard can be tailored to the Croatian context with the aid of the specifically designed software (CDROM Calculation Population Energy Requirements and Food Needs). The software takes into account the up- to-date representation of the Croatian demographic pyramid and appropriate values of the body mass index (BMI) by age and gender. By specifying Croatia-specific parameters regarding the birth and activity rates, the software delivers estimates of daily energy requirements by age groups. 11 A norm of 2,700 kcal per day per equivalent adult was adopted (see Table 1.3).9 According to FAO (2004), 2700 kcal/day is the minimum energy requirement for a reference person with the following characteristics: male, aged 18-30, with body weight in the range of 65-70 kilograms, a basal metabolic rate (BMR, that is the energy required for sustaining the basic functions of the body) equal to approximately 25.3, and enjoying a "lightly active lifestyle" (in technical terms, with a "physical activity level" (PAL) equal to 1.6). Nutritional requirement for energy intake of a household depends on age and sex of the household members. Total household energy requirement is calculated by multiplying the energy requirement per equivalent adult by equivalent size of a household. By construction, this equivalent size complies with international nutritional standards. The nutritional equivalence scale used to obtain the nutritional adult equivalent size of each household is shown in Table 1.3. Table 1.3: Nutritional Requirements and Equivalence Scale Daily energy requirement (Kcal) Equivalence scale Age (years) Female Male Female Male -1 573 631 0.212 0.234 1-4 1076 1168 0.399 0.433 5-17 2046 2395 0.758 0.887 18-29 2250 2809 0.833 1.040 30-59 2154 2700 0.798 1.000 60+ 1943 2265 0.720 0.839 Source: Authors' calculations based on FAO (2004). The next step is to calculate the food consumption (in kilos, liters, or units) per equivalent adult for each household in the sample by using 100 food items from the HBS. For each food item, we calculated the average food consumption per equivalent adult for households in the lowest quintile of the total equivalent consumption distribution in order to identify the composition of the minimum food basket.10 The choice of the lowest quintile is somewhat arbitrarily, of course, but it fits with the idea that the minimum food basket reflects actual consumption pattern of those just around poverty line, or more specific, of those who can just afford the minimum required calorie intake. Resulting quantities were transformed to calorie values by using conversion tables of Croatian Institute for Public Health (Zavod za zastitu zdravlja SR Hrvatske, 1990). The average calorie intake per equivalent adult for the lowest quintile turned out to be 2,859 kcal per day, which is slightly higher than the norm chosen. We scaled down quantities of all food items consumed by multiplying these quantities by a factor of 0.944 (= 2,700 kcal/2,859 kcal), so as to reach the food basket that yields exactly 2,700 kcal per day per equivalent adult. The vector identified is the minimum food basket. Its composition by major product groups is shown in Table 1.4. The complete content of the basket can be seen in Appendix II. 9 In per capita terms, the norm is equal to 2,250 kcal. 10 Here, total equivalent consumption is calculated using the EU equivalence scale. 12 Table 1.4: Composition and Cost of Minimum Foods Basket by Product Groups Product group Quantity KCal Percent of energy Monthly cost ay/ equivalent adult) requirement (kuna/eq. adult) Bread and cereals 430 1131 41.9% 98.08 Meat and meat preparations 161 383 14.2% 148.56 Fish 16 13 0.5% 11.50 Milk, milk products and eggs 327 230 8.5% 76.25 Oils and fats 47 411 15.2% 17.95 Fruits 93 42 1.6% 20.68 Vegetables 299 155 5.8% 63.71 Sugar, jam, chocolate and confectionery 44 169 6.3% 17.45 Other food items 30 36 1.3% 16.84 Beverages, coffee, and tea 141 129 4.8% 58.32 Total 2700 529.35 The Cost of the Minimum Food Basket (The Food Poverty Line) For the calculation of the cost of the minimum food basket, we need to find appropriate prices by items. One possibility is to use average prices collected by the price department of the CBS, and used in CPI calculations. The disadvantage of this approach lies in not knowing how closely goods whose prices are collected correspond to the goods actually consumed by households. The HBS itself is a rich source of data and, fortunately, provides enough information for our purposes. The Diary and the Supplementary Questionnaire contain information on the monetary value and quantity of food items consumed. Their quotient yields a unit value per item per household. This unit value can be used as a price substitute. It has the advantage that it corresponds exactly to the good consumed. For the purpose of estimating the cost of the food basket we calculate the median unit value for households in the lowest quintile of equivalent consumption. This measure is then used as the price.11 The resulting cost of the minimum food basket is 529 kunas per month (Table 1.4). Estimation of the Poverty Line Following the "cost-of-basic-need" approach, as described in Ravallion (1994), the (total) poverty line is estimated through a two-step procedure.12 Step 1 identifies the households whose food expenditures are approximately equal to the cost of the minimum food bundle. Step 2 estimates the poverty line by averaging total household expenditures on the subset of households identified in step 1. Precisely, the poverty line z is defined as follows: [ (12) z = E x x food z food , ] where x denotes total household expenditure (HRK/adult/year), xfood is the food household expenditure (HRK/adult/year), zfood is the food poverty line (HRK/adult/year), and the symbol is used as a short-cut for "approximately equal to". 11 Unit values of seemingly identical goods shows substantial variation depending on the position of households in the distribution of total equivalent consumption. Poorer households are usually faced with lower unit values, what probably reflects a bit poorer quality of food consumed. 12 This method was also implemented in the World Bank (2000) poverty study for the year 1998. See the Appendix IV. 13 The main features of the cost-of-basic-need method can be summarized as follows: ˇ The welfare indicator is the total expenditure on consumption (per adult equivalent); ˇ The poverty line includes an allowance for non-food expenditures. ˇ The key identifying condition x food z food is based on the sub-sample of households whose food expenditure equal the cost of the minimum food bundle. In order to estimate equation (12) parametrically we specify the following linear regression model: x food (13) ln = 0 + 1 ln x + 2 (ln x )2 , z food where betas are the unknown parameters. If the dependent variable takes a value higher than one, then household expenditure on food exceeds the cost of the minimum food basket. After estimating equation (13) by Weighted Least Squares13, the (total) poverty line is obtained by setting xfood = zfood in the predicted regression function ­ see also Luttmer (2000). Note that by setting the food consumption expenditure equal to the food poverty line in equation (13): (14) ln (1) = 0 = 0 + 1 ln z + 2 (ln z ) ^ ^ ^ 2 By solving the equation above and exponentiating the predicted values (so as to obtain predictions for x from a model involving the log of x), the median of x rather then the mean (see Wooldridge, 2002: pages 202-ff.), is derived, which is a desirable feature for our purposes (it adds robustness to the estimates) 14: - ą 2 - 4 ^ ^ ^ ^ [ (15) z = E x x food = z food ] = exp 1 1 ^ 2 2 0 2 Equation (15) estimates the poverty line. Table 1.5 shows the results of poverty line estimation using different equivalence scales. In the baseline case, corresponding to the use of OECD-II equivalence scale, the poverty rate is estimated at 11.1 percent of population. Table 1.5: Poverty Lines using Various Equivalence Scales Poverty line Equivalence scale: Poverty rate Single adult Couple w/o Single parent Couple w/ 2 (%) kids kids Baseline OECD-II equivalence scale 11.1 22,145 33,217 28,788 46,504 Other OECD-I equivalence scale 12.8 19,213 32,661 28,819 51,874 Per capita scale 14.5 15,122 30,243 30,243 60,487 WB ECA scale 12.4 19,780 33,266 33,266 55,946 13 OLS estimation provides similar results. 14 Estimating a median value instead of a mean value has also implications when comparing poverty lines across different methods. 14 FINAL REMARKS In this paper we have undertaken two tasks: (i) discussing the choices underlying the construction the consumption aggregate, and (ii) estimating the absolute poverty line for Croatia on the basis of the 2004 HBS. With regard to task (i), the above discussed details need not be repeated. With regard to task (ii), the estimation of the absolute poverty line, we have omitted the results of a number of exercises that were carried out before identifying the "best" method. More precisely, we estimated poverty lines using the Orshanki's (1965) method, as well as other methods described in Luttmer's (2000) paper.15 After evaluating the theoretical underpinnings of each method as well as the empirical results, we opted for the cost-of-basic need method described in section D. Inevitably, this choice is arbitrary. Each of the methods examined is defensible, at least to some extent, on the basis of `technical' arguments. The two single most important reasons behind our choice were (i) the fact that poverty estimates obtained by this method can be consistently compared with previous poverty estimates ­ the World Bank (2001) report; and (ii) the cost-of-basic need approach is well understood and popular both in literature and among practitioners. 15 Details are available from the author upon request. 15 References Deaton A. and S. Zaidi. 2002. "Guidelines for Constructing Consumption Aggregates for Welfare Analysis." Living Standards Measurement Study Working Paper Nr. 135. World Bank, Washington D.C. Chen, S. and M. Ravallion. 1996. "Data in transition: Assessing Rural Living Standards in Southern China." China Economic Review, 7(1): 23-56. FAO. 2004. Human Energy Requirements; Report of a Joint FAO/WHO/UNU Expert Consultation. FAO Food and Nutrition Technical Report Series No.1. Food and Agriculture Organization, Rome. Gottschalk, P. and T. Smeeding. 1997. "Cross-National Comparisons of Earnings and Income Inequality.", Journal of Economic Literature, 35(June): 633-687. Luttmer E. 2000. "Methodlogy." Background Paper No. 2 in World Bank (2000). Ravallion, M. 1994. Poverty Comparisons, Fundamentals of Pure and Applied Economics 56. Chur, Switzerland: Harwood Academic Press. Ravallion, M. 1998. "Poverty Line in Theory and Practice." LSMS Working Paper No. 133. World Bank, Washington D.C. de Vos K., and M. A. Zaidi . 1997. "Equivalence Scale Sensitivity of Poverty Statistics for the Member States of the European Community." Review of Income and Wealth, 43(3): 319-333. Wooldridge, J.M. 2002, Introduction to Econometrics: A Modern Approach. 2nd edition Thomson South- Western. World Bank, 2000, Croatia: Economic Vulnerability and Welfare Study, Volume II: Technical Papers. Document of the World Bank, Poverty Reduction and Economic Management Unit, Europe and Central Asia Region, World Bank, Washington D.C. World Bank, 2000a, "Making Transition Work for Everyone: Poverty and Inequality in Europe and Central Asia." World Bank, Washington D.C. World Bank. 2005. Growth, Poverty an Inequality: Eastern Europe and the Former Soviet Union. World Bank, Washington D.C. World Health Organization. 1985. Energy and protein requirements: Report of a joint FAO/WHO/UNU expert consultation. WHO Technical Report Series No. 724. Geneva. Zavod za zastitu zdravlja SR Hrvatske, 1990, Tablice o sastavu namirnica i pia (by Kai- Rak A. and K. Antoni). Zagreb: Zavod za zastitu zdravlja SR Hrvatske. 16 APPENDIX I: HEDONIC REGRESSION RESULTS A1.1: Regression of the Log of Rental Value on Housing Characteristics Coeff. S.E . P-value Constant 7.449 0.171 0.000 Dwelling type Apartment (omitted) House -0.110 0.065 0.090 Year of build Before 1918 (omitted) 1918-1945 -0.006 0.065 0.928 1946-1960 0.042 0.058 0.472 1961-1970 0.063 0.055 0.252 1971-1980 0.075 0.056 0.177 1981-1990 0.065 0.057 0.251 1991-2000 0.082 0.060 0.168 2000- 0.075 0.122 0.539 Building type Detached house w/ one appt (omitted.) House w/ one apt 0.045 0.043 0.296 House w/ two apts 0.053 0.029 0.071 Building w/ 3+ apts 0.105 0.072 0.145 No. of rooms Studio (omitted) 1-room 0.124 0.116 0.288 2-rooms 0.292 0.119 0.014 3-rooms 0.395 0.125 0.002 4-rooms 0.410 0.133 0.002 5-rooms 0.638 0.143 0.000 6+ rooms 0.649 0.172 0.000 Location Central Croatia urban (omitted) Central Croatia rural 0.224 0.042 0.000 Eastern Croatia urban -0.037 0.037 0.323 Eastern Croatia rural 0.229 0.043 0.000 Zagreb Region urban 0.562 0.034 0.000 Zagreb Region rural 0.310 0.052 0.000 Adriatic North urban 0.328 0.048 0.000 Adriatic North rural 0.277 0.050 0.000 Adriatic South urban 0.513 0.040 0.000 Adriatic South rural 0.130 0.049 0.008 Amenities Hot water 0.050 0.031 0.111 Collective central heating 0.153 0.039 0.000 Own central heating 0.134 0.028 0.000 Balcony 0.121 0.021 0.000 Garden 0.019 0.027 0.474 Kitchen (in-house) 0.268 0.057 0.000 Bathroom (in-house) 0.094 0.052 0.073 Complains Rotten -0.177 0.035 0.000 Humidity -0.083 0.041 0.044 Lack of lights 0.023 0.057 0.682 Installation Running water 0.268 0.055 0.000 Sewage 0.103 0.032 0.001 Gas 0.106 0.026 0.000 Telephone 0.044 0.042 0.296 Usable space Space (m2) 0.010 0.002 0.000 Space squared (*1000) -0.024 0.006 0.000 No. of obs. 2,846 Adj. R2 0.636 17 APPENDIX II: The Minimum Food Basket Table A2.1: The Minimum Food Basket Quantity Percent of Energy Monthly cost (g/day/equivalent energy Product (kcal) (kuna/eq.adult) adult) requirement Rice 15 51 1.9% 3.12 Flour 82 277 10.3% 11.34 Corn and other cereals 1 5 0.2% 0.33 White bread 80 187 6.9% 16.21 Bread rolls etc 5 14 0.5% 2.71 Other bread 199 438 16.2% 34.81 Long-lasting bread products 9 33 1.2% 7.76 Cakes 1 5 0.2% 1.57 Doughnuts 1 4 0.1% 0.92 Pizza, "burek" etc. 2 7 0.3% 1.93 Macaroni products 25 91 3.4% 7.50 Cereal products (flakes etc.) 0 2 0.1% 0.44 Dietary cereal products 0 1 0.0% 0.10 Beef w/o bones 8 16 0.6% 9.34 Beef with bones 9 16 0.6% 10.20 Baby beef w/o bones 1 1 0.0% 1.12 Baby beef with bones 1 1 0.1% 1.71 Pork w/o bones 15 41 1.5% 17.05 Pork with bones 23 51 1.9% 23.22 Lamb 2 3 0.1% 3.10 Poultry 64 95 3.5% 40.54 Other meat 0 0 0.0% 0.29 Sub-products 3 5 0.2% 2.01 Smoked, dried sausage 42 163 6.1% 47.55 Other meat products (canned) 1 5 0.2% 1.77 Fresh water fish 2 1 0.0% 1.61 Sea fish 11 7 0.3% 6.62 Other seafood 1 1 0.0% 1.27 Smoked, dried, salted fish 0 0 0.0% 0.21 Canned fish 2 3 0.1% 1.78 Milk (full fat) 149 95 3.5% 20.89 Milk (reduced fat) 106 51 1.9% 14.52 Powdered milk 0 0 0.0% 0.03 Other milk products ( fruit yoghurt, cream) 12 14 0.5% 6.30 Yoghurt 19 11 0.4% 6.05 Fresh farmer cheese 10 7 0.3% 6.24 Cheese 4 15 0.6% 5.45 Soft cheese 0 1 0.0% 0.65 Eggs 27 36 1.3% 16.13 Butter 0 3 0.1% 0.30 Margarine 6 47 1.7% 3.25 Oil 32 289 10.7% 9.75 Olive oil 1 10 0.4% 2.13 Other fat 7 62 2.3% 2.52 18 Quantity Percent of Energy Monthly cost (g/day/equivalent energy Product (kcal) (kuna/eq.adult) adult) requirement Citrus fruits 20 4 0.2% 4.78 Bananas 17 8 0.3% 3.61 Apples 34 12 0.5% 5.59 Pears 3 1 0.0% 0.64 Peach, apricots, etc 7 2 0.1% 2.11 Berries and grapes 2 1 0.0% 0.85 Watermelon 7 3 0.1% 0.80 Dried fruits 1 1 0.0% 0.43 Walnuts, cleaned 1 5 0.2% 0.97 Walnuts, in shells 0 1 0.0% 0.29 Eatable grains 0 3 0.1% 0.31 Fruits, canned 1 1 0.0% 0.29 Vegetables-leafy 23 3 0.1% 6.87 Vegetables-grassy 0 0 0.0% 0.36 Cabbage 31 6 0.2% 4.85 Tomatoes, green pepper, and cucumbers 49 7 0.2% 15.07 Onions, carrots, mushrooms, and olives 30 9 0.3% 6.15 Vegetables for soups 1 0 0.0% 2.26 Dried vegetables 11 30 1.1% 4.94 Vegetables, preserves and canned 20 6 0.2% 6.35 Dietary vegetable products 0 0 0.0% 0.02 Potatoes 133 93 3.4% 16.16 Potato products (chips) 1 2 0.1% 0.68 Sugar 34 134 5.0% 6.38 Marmalade, compote, and honey 5 13 0.5% 2.89 Chocolate 2 12 0.5% 3.99 Candies 1 6 0.2% 1.95 Chewing gum 0 0 0.0% 0.76 Ice cream 2 3 0.1% 1.48 Dressing (mayonnaise, ketchup, vinegar) 8 15 0.6% 3.08 Spices and salt 19 13 0.5% 9.80 Soup concentrate and other powders 2 5 0.2% 2.63 Baby food 1 2 0.1% 0.79 Deserts and other food 1 0 0.0% 0.54 Coffee beans 9 25 0.9% 13.16 Tea 1 1 0.0% 2.66 Cocoa 0 1 0.1% 0.62 Mineral water 0 0 0.0% 5.40 Carbonized drinks 21 10 0.4% 3.06 Fruit juices 27 14 0.5% 6.56 Syrup 15 37 1.4% 5.18 Liqueurs 1 3 0.1% 2.18 Wines 21 18 0.7% 7.15 Beer 45 19 0.7% 12.34 Total 2700 100% 529.35 19 Table A2.2: Food Items with the Highest Cost Share in the Minimum Food Basket Product Unit Quantity Price = Monthly cost (unit/day/equivalent median unit (HRK/eq.adult) adult) value (HRK/kg) Smoked, dried sausage g 42 37.68 47.55 Poultry g 64 20.94 40.54 Other bread g 199 5.76 34.81 Pork with bones g 23 33.60 23.22 Milk (full fat) ml 149 4.63 20.89 Pork w/o bones g 15 38.04 17.05 White bread g 80 4.00 16.21 Potatoes g 133 6.67 16.16 Eggs piece 27 1.00* 16.13 Tomatoes, green pepper, and cucumbers g 49 10.14 15.07 Milk (reduced fat) ml 106 4.52 14.52 Coffee beans g 9 49.15 13.16 Beer ml 45 8.98 12.34 Flour g 82 4.54 11.34 Note: *Price per piece. 20 BACKGROUND PAPER #2 POVERTY COMPARISON FOR CROATIA 2002-04 Danijel Nesti INTRODUCTION This Chapter presents poverty comparisons for Croatia in the period between 2002 and 2004. It concentrates on three types of comparisons (i) over time, (ii) across regions and (iii) international comparisons. In the comparisons over time we focus on the 2002-2004 period, since any inter-temporal comparability of poverty estimates as regards an earlier period is severely limited. Poverty estimates are based on a detailed analysis of the Household Budget Survey (HBS). However, changes of its sample frame, design and coverage restrict the use of a longer series. The HBS sample frame was updated in 2002 to use the 2001 Census as the basis. The HBS questionnaires became stable in 2002-2004 after several revisions conducted in previous years. We were also tempted to use a previous World Bank (2000) study on poverty in Croatia as a point of reference to the current poverty estimate. However, the previous study was based on the 1998 HBS, which only covered those parts of Croatia that were not directly affected by war. It therefore failed to cover refugees and displaced persons, groups that are likely to be facing a high poverty risk. Moreover, recent surveys provide a slightly different data set, making it necessary to modify the poverty estimation methodology used in the current study when compared to that in the World Bank (2000). Hence, reliable poverty comparisons have been possible for the 2002-2004 period only, and no direct reference to the 1998 poverty figures should be made. Poverty estimates at the regional level are based on the pooled data sets from three HBS rounds between 2002 and 2004. Pooling was made in order to increase the sample size and enable deriving of representative statistics at a sub-national level. Nevertheless, county-level estimates are still subject to large standard errors, impeding a reliable statistical inference. This paper concentrates instead on five analytical regions arranged as the groups of counties that allow the recognition of regional differences in living standards. The Chapter is organized as follows. Section B provides poverty figures for 2002-2004 and assesses changes over that time. Besides consumption-based poverty estimates, it also presents the poverty figures based on income and various inequality indicators. Section C deals with regional variations in the living standard. Section D provides poverty and inequality comparisons between Croatia and other countries of Central and Eastern Europe. Section E concludes. The main body of the paper is supplemented by two appendices with a number of additional tables that should facilitate further poverty comparisons and future work on this topic. 2002-2004 POVERTY TRENDS Poverty Line and Poverty Incidence In this study, poverty is defined as a lack of material consumption; it is estimated on the basis of individual-level data from the HBS. In order to allow consumption to cover material well-being to as broad an extent as possible, we define a consumption aggregate that includes both household expenditures on non-durable goods and the imputed flow of consumption for durable goods. Rental value of owner-occupied dwelling is also imputed. Actual spending on the purchase of durable goods or houses 21 is excluded so as to avoid double-counting. For a better comparison of consumption between households of different sizes and composition we apply a modified OECD equivalence scale. Consumption-based absolute poverty line is estimated after applying the main features of the cost-of-basic- need method (Ravallion 1994; 1998). Poverty line is set at the level of consumption that allows one to cover minimum non-food expenses without the need to cut down food consumption below the level that satisfies nutritional requirements. In other words, poverty line resources should provide for a satisfactory diet and minimum consumption of other goods and services at the same time. An extensive procedure of step-by-step construction of poverty line is conducted for 2004. Poverty line for other years, such as 2002 and 2003 in this study, is determined as the 2004 line adjusted for inflation. We use the headline consumer price index to adjust for inflation. More details on the construction of the consumption aggregate and the estimation of the poverty line can be found in the methodology paper in Chapter 1 of this report. Table 2.1 reports the poverty incidence in Croatia in the period 2002-2004. In 2004, poor in Croatia represented 11.1 percent of the population. In other words, some 11.1 percent of individuals were estimated to be living with the equivalent consumption below the poverty line. If we take into account a standard error of estimation as a measure of imprecision of our estimate, it should be accepted that this rate could be a bit lower or higher. A margin of two standard errors around the central rate gives a 95 percent confidence level in the statistical conclusion of the true value. Therefore, we can be quite confident that the true poverty rate in Croatia was between 9.3 percent and 12.9 percent in 2004. Poverty line itself is estimated at the level of consumption of 22,145 kuna per equivalent adult per year, or 1,845 kuna per month. This amount should suffice for a single adult to escape absolute material poverty. For households composed of more than one person, poverty line is determined by multiplying the poverty line for equivalent adult by the equivalent size of the household. In determining the equivalent size we apply the modified OECD scale. In case of a household consisting of two adults and two children, for example, equivalent household size is 2.1 (1 for adult + 0.5 for other adult + 2*0.3 for children under 14). Poverty line for such a household is estimated at 46,504 kuna per year. The poverty gap was 2.6 percent in 2004. This poverty indicator shows that resources representing 2.6 percent of the poverty line per equivalent adult should be sufficient to lift all the poor out of poverty if perfectly distributed to every poor household in the amount filling the gap between actual consumption and the poverty line. The severity of poverty also shows how far consumption of the poor is from the poverty line, but attaches even higher importance to the poor that are deeper into poverty. The average poverty deficit of 23.8 percent shows how much smaller consumption of the poor is on average than the poverty line. Table 2.1: Poverty Incidence, 2002-2004 2002 2003 2004 Poverty line (eq. adult), kn/year 21,390 21,732 22,145 Poverty line (couple & 2 children), kn/year 44,918 45,637 46,504 Poverty headcount rate (s.e.) 11.2% (0.8) 12.4% (1.0) 11.1% (0.9) Poverty gap (s.e.) 2.6% (0.3) 3.1% (0.4) 2.6% (0.3) Poverty severity (s.e.) 1.0% (0.2) 1.2% (0.2) 1.0% (0.1) Average poverty deficit (s.e.) 23.5% (1.2) 25.1% (1.3) 23.8% (1.2) Notes: Linearized standard errors based on sample specification are reported in parentheses. Poverty estimates are based on equivalent consumption using the modified OECD scale (1; 0.7; 0.3). The baseline poverty line is estimated for 2004, while poverty lines for 2002 and 2003 are adjusted from the 2004 line by using the consumer price index. Poverty lines are expressed in April prices of each given year. Consumption-based poverty incidence in Croatia exhibits minor changes over the 2002-2004 period. Poverty headcount rates, as well as other poverty measures, were almost the same both in 2002 and 22 2004, with mildly higher values in 2003. We should take into account the statistical nature of estimated measures to see how significant the changes observed in that period are. Table 2.2 reports changes in the poverty measures and associated standard errors. Calculated P-values suggest that we cannot reject the hypothesis of no-change in any observed case. In other words, we are unable to discern statistically significant changes in poverty incidence over the period between 2002 and 2004. Table 2.2: Changes in Poverty 2003 2004 Change P-value on Change P-value on over previous test of over previous test of year change=0 year change=0 Poverty headcount rate 1.1% 0.40 -1.3% 0.35 (1.3) (1.4) Poverty gap 0.5% 0.30 -0.5% 0.31 (0.5) (0.5) Poverty severity 0.2% 0.44 -0.2% 0.28 (0.2) (0.2) Average poverty deficit 1.6% 0.37 -1.3% 0.47 (1.8) (1.8) Notes: Linearized standard errors based on sample specification are reported in parentheses. Consumption-based poverty estimates are compared. Findings on a stagnant poverty rate over the 2002-2004 period are challenging for positive GDP growth observed at the same time.16 A major reason for the discrepancy between vigorous economic growth and sluggish changes in poverty should be found in the household consumption records used in poverty calculations. The average household consumption, as calculated from the HBS data, has been practically stagnant in real terms between 2002 and 2004, with a small decrease in 2003 and a small increase in 2004.17 Table 2.3 shows the results of a decomposition of changes in the poverty rate on growth and inequality component, illustrating clearly that the expected positive impact of economic growth on poverty was absent. Negative consumption growth recorded in 2003 caused an increase by about 0.5 percentage points in the headcount poverty rate. Consumption growth in 2004 helped to decrease poverty rate by about 0.2 percentage points. Similarly, changes in inequality have worked for a higher poverty rate in 2003 and lower poverty rate in 2004. Table 2.3: Decomposition of Changes in the Headcount Poverty Rates, in percent 2003 2004 Total 1.1 -1.3 Due to growth 0.5 -0.2 Due to inequality 0.6 -1.1 Growth is a powerful driver out of poverty, especially if it is equally or pro-poor distributed. In a simple illustration given in Table 2.4 we could see the impact of stable medium and long-term growth on poverty rates. Simulation starts from the 2004 consumption distribution, poverty rate and poverty line. With the 16 The average GDP growth over 2002-2004 was 4.9 percent per year (CBS First Release 12.1.1/4 from March 30, 2006). 17 In many countries consumption growth is lower when calculated from household surveys than if taken from the national account statistics (see, for example, Paci et al. 2004 for evidence in case of Poland). However, the discrepancy in Croatia is rather large and puzzling. 23 growth rate of 4 to 5 percent per year in the next 5 years (this seems like an affordable scenario for Croatia) the poverty rate could be halved if only this growth is equally distributed across the whole population. If similar growth is recorded in the next 10 years, the poverty rate could drop below 3 percent. Slow consumption growth, on the other hand, severely reduces the chances of poverty reduction even in the long term. Table 2.4: Growth Effect on Poverty Rates Average annual growth Poverty rate after 5 year Poverty rate after 10 years 0 11.1 11.1 1 9.4 8.6 2 8.6 6.3 3 7.4 4.2 4 6.3 3.0 5 5.4 2.1 6 4.3 1.4 Note: Average annual growth rate refers to the household consumption growth starting from 2004. Projected poverty rates assume that the distribution of equivalent consumption is not changed, i.e. each household experiences the same consumption growth. Income Poverty Poverty estimates presented so far were based on consumption because consumption is widely accepted as a more robust welfare measure than income. Detailed household records on consumption were used as the basis for estimating poverty line. However, once we have set the poverty line in money-metric terms, it is possible to look at the income side and analyze poverty as a lack of income. Table 2.5 presents the results for income poverty. Poverty rate was reduced from 13.0 percent in 2002 to 10.4 percent in 2004, and the same trend is observed in other poverty measures. Poverty reduction over the 2002-2004 period is statistically significant at usual confidence levels for all measures except for the average poverty deficit (results not shown here). 18 This gives a contrasting picture to consumption-based poverty changes and deserves closer inspection. Table 2.5: Income Poverty Incidence, 2002-2004 2002 2003 2004 Poverty line (eq. adult), kn/year 21,390 21,732 22,145 Poverty headcount rate (s.e.) 13.0% (0.9) 12.4% (1.0) 10.4% (0.9) Poverty gap (s.e.) 3.2% (0.3) 3.0% (0.3) 2.4% (0.3) Poverty severity (s.e.) 1.2% (0.2) 1.2% (0.2) 0.9% (0.1) Average poverty deficit (s.e.) 18.8% (1.6) 16.1% (1.3) 15.5% (1.6) Note: Linearized standard errors based on sample specification are reported in parentheses. Poverty estimates are based on equivalent income using the modified OECD scale (1; 0.7; 0.3). An examination of the changes in per equivalent adult consumption and incomes across the population over the 2002-2004 period gives an insight into differing patterns of poverty changes. Figure 2.1 depicts growth 18 A relatively bright picture of changes in income poverty is enforced by the household perception about their living standard. In the HBS, there is a following question: "With respect to your disposable income, evaluate how your household lives". In 2002, some 13.4 percent of the population said they lived with great difficulties, and two years later such an unpleasant perception was expressed by just 10 percent of population. The proportion of those living with difficulties also decreased over the 2002-2004 period, while the proportion of those perceiving that they live well and very well increased (see Appendix A3 for raw figures). 24 incidence curves for income and consumption and shows how different the changes in per equivalent terms actually were, and at different parts of the population. Per equivalent consumption was stagnant on average and across the population, except for positive growth observed at a very low-end of the distribution and negative growth at a very high-end of the distribution. On the other hand, income growth was around 6 percent over two years, with all income-groups recording a positive real income growth. Above-average growth is found at the bottom and the top of the income distribution. The difference between the evolution of income and consumption is somewhat puzzling. It is possible that the population perceived a recent increase in income as transitory. Since households smooth their consumption to equalize their long-term incomes, a sudden income jump is more likely to be saved or invested than consumed, especially by better-off households. In that way, income growth is only partially reflected in consumption as the interest or property income. Another explanation could be that the current design of the HBS is valid for a snapshot estimate of poverty and consumption profile only, but can not provide a reliable picture of changes over a short period of time because there is no panel part of the sample.19 Figure 2.1: Growth Incidence Curve, 2002-2004 14.0% 12.0% 10.0% 8.0% Growth (in %) 6.0% 4.0% 2.0% 0.0% -2.0% -4.0% 0 10 20 30 40 50 60 70 80 90 100 Percentiles Consumption Income Notes: Changes in per equivalent adult consumption and incomes in constant prices are shown. Modified OECD scale is used. Table 2.6 matches the group of consumption-poor with the group of income-poor to see to what extent these two groups overlap. Let us keep in mind that the poverty line applied for both estimations is the same, still, the consumption poverty rate was 11.2 percent and the income poverty rate 10.4 percent, both estimated for 2004. Around 60 percent of the consumption-poor are found to be income-poor too. In other words, 6.4 percent of the population has both equivalent income and equivalent consumption below the poverty line. Some 4.7 percent of the population has equivalent consumption below the poverty line, but their incomes are above poverty line. Another 4 percent of the population has equivalent consumption higher than the poverty line but its equivalent incomes are below the poverty line. Finally, around 85 percent of the population can not be regarded as poor either by their equivalent consumption or by their equivalent incomes. 19 It is also possible that the CBS staff (interviewers, controllers and data processing staff) invest a lot more effort in collecting the income side of household resource flows in the HBS since income records are more important for regular statistical reporting of monetary poverty indicators (Laeken indicators). 25 Table 2.6: Overlap between Income and Consumption Poverty, 2004 Income-poor Income-non-poor Total Consumption-poor 6.4% 4.7% 11.1% (0.5) (0.4) Consumption-non-poor 4.0% 84.9% 88.9% (0.4) (0.7) Total 10.4% 89.6% 100% Note: Standard errors are reported in parentheses. Implicit Poverty Lines Poverty estimates could be prepared by applying the implicit poverty lines that are in use either by the social welfare system as a benefit threshold, or by trade unions for the cost-of-living calculations. For example, social assistance of 400 kuna per person per month may be tested as a possible poverty line. In our estimate, based on per capita household consumption, it can be seen that this standard is by no means extremely low; it results in the headcount poverty rate of 0.15 percent (Table 2.7). In other words, there is a minor proportion of the population whose consumption is smaller than the social assistance. Two large trade union confederations, the Association of the Independent Trade Unions of Croatia (SSSH) and the Independent Croatian Trade Unions (NHS), constructed their own minimum consumer baskets. A basket consists of goods and services that are considered a necessity, with minimum quantities of these per month defined. The cost of the basket for a 4-member household was 5,307.69 kuna per month in the SSSH basket for April 2004, and 5,800.43 kuna per month in the NHS basket for May 2004. After applying these implicit poverty lines to our consumption records equalized by the per capita scale, the resulting poverty rates were 17.3 percent and 22.4 percent for the SSSH and the NHS basket, respectively (Table 2.7). Table 2.7: Poverty Rates Based on Implicit Social Welfare and Trade Union Poverty Lines, 2004 Poverty line Consumption poverty (kuna/year/per capita) rate Social assistance (400 kuna/month/person) 4,800 0.15% Trade union basket (SSSH- April 2004) 15,923 17.3% Trade union basket (NHS -May 2004) 17,401 22.4% Note: Poverty rate estimation is based on the per capita household consumption. Consumption and Income Inequality Inequality in the distribution of material resources is an important factor determining the poverty incidence. Table 2.7 summarizes the main features of the income and consumption inequality in Croatia. In the distribution of equivalent consumption, the bottom decile (the poorest 10 percent of the population) has a share of 3.8 percent in 2004 whereas in the equivalent income distribution it has a share of 3.6 percent. The top decile (the richest 10 percent of population) in the consumption distribution used around 20.4 percent of resources, while it used about 21.8 percent of the total income in 2004. Thus, the poorest population in the income distribution commanded fewer resources and the richest population in the income distribution commanded more resources than in the consumption distribution. This suggests higher inequality in the income distribution. Decile share d9/d1 in the consumption distribution shows that the poorest person in the 26 top decile consumes around 3.18 times more than the richest person in the bottom decile in 2004. Decile share d9/d1 in the income distribution takes value of 3.55, suggesting a larger difference between the rich and the poor than is the case with the consumption distribution. More sophisticated inequality measures that take into account the whole distribution, the Gini coefficient, the Theil entropy index and the mean log deviation, all point similarly to a higher inequality in the income distribution compared to the consumption distribution. For example, the Gini coefficient for the income distribution of 0.275 in 2004 is higher than the Gini of 0.253 for the consumption distribution. It could be noted that our income aggregate for poverty and inequality estimations is inclusive of the income in-kind and imputed rents for owner-occupied dwellings, causing slightly lower values of inequality measures than would be the case for the distribution of monetary income. As for changes over the period 2002-2004, both the consumption and the income distribution seem quite stable. Decile shares and decile ratio d9/d1 for 2002 and 2004 are quite closely tied in case of both distributions (Table 2.8). The Gini coefficient for the distribution of equivalent consumption is a bit lower in 2004 than in 2002. But we have to take into account statistical uncertainty about their true values via calculated standard errors. For example, there is a 95 percent probability that the true value of the Gini coefficient lies within 2 standard errors around our point estimate (0.253 in 2004). In that way, we could be quite confident that the true Gini coefficient lies between 0.247 and 0.259 in 2004. Similarly, we cannot be confident that any change occurred as compared to the Gini coefficient of 0.258 in 2002. Table 2.8: Consumption and Income Inequality Statistics 2002 2003 2004 Consumption Cons. share of the bottom decile 0.038 (0.001) 0.036 (0.001) 0.038 (0.001) Consumption share of the top decile 0.210 (0.003) 0.203 (0.002) 0.204 (0.002) Decile ratio d9/d1 3.181 (0.038) 3.348 (0.056) 3.182 (0.048) Gini coefficient 0.258 (0.003) 0.259 (0.003) 0.253 (0.003) Theil entropy measure 0.123 (0.007) 0.113 (0.004) 0.107 (0.003) Mean log deviation 0.116 (0.004) 0.116 (0.003) 0.109 (0.002) Total Income Income share of the bottom decile 0.034 (0.000) 0.035 (0.000) 0.036 (0.000) Income share of the top decile 0.211 (0.002) 0.210 (0.002) 0.218 (0.002) Decile ratio d9/d1 3.592 (0.044) 3.488 (0.040) 3.553 (0.051) Gini coefficient 0.270 (0.002) 0.269 (0.002) 0.275 (0.003) Theil entropy measure 0.122 (0.003) 0.120 (0.003) 0.127 (0.003) Mean log deviation 0.126 (0.002) 0.124 (0.002) 0.128 (0.002) Memorandum items: Mean equivalent consumption 43,346 (305) 42,830 (258) 43,074 (245) Mean equivalent income 44,653 (251) 44,195 (259) 47,327 (278) Notes: Bootstraped standard errors are in parentheses. Consumption aggregate definition is explained in the methodology paper. Total income includes the in-kind income and self-reported rental values for owner-occupied dwellings. Mean equivalent consumption and income are expressed in April 2004 prices. Equivalent income and consumption are calculated by using the modified OECD scale. The Theil index and the mean log deviation for the consumption distribution are notably lower in 2004 than in 2002. The Theil entropy measure is by its construction rather sensitive to changes at the upper part of the distribution, and the mean log deviation to the lower part of the distribution. Figure 2.1 shows large changes in those parts of the distribution over the period 2002-2004, which is here reflected in the two aforementioned inequality measures. Changes in inequality measures are even less pronounced in the income distribution. 27 With the help of Figure 2.1 again we could loosely conclude that improved incomes at the lower part of the distribution are "counter-balanced" by notably higher incomes at the richer part of the distribution, while aggregate inequality remains roughly the same. REGIONAL INEQUALITIES Estimation of regional poverty in Croatia was a true challenge having in mind relatively small sample size in the HBS that can be employed for that purpose. We decided to pool the samples for 2002, 2003 and 2004 and to proceed with poverty estimates employing this enlarged sample. Sampling procedure currently used for the HBS allows us to pool data sets without running into any significant statistical problem. In spite of that, the pooled sample did not enable a low-standard error estimate of poverty incidence at the county level. We decide to group counties into analytical regions that allow deriving representative statistics at a sub-national level and recognize regional variations in living standards. We apply the 5-region disaggregation, which is actually borrowed from the previous World Bank (2000) Living Standard Assessment and provides regional statistics combined with urban/rural division for the location of residence.20 These five analytical regions are Central Croatia, Eastern Croatia, Zagreb Region, Adriatic North and Adriatic South (Table 2.9). Table 2.9: Definition of Analytical Regions Analytical Region County Krapina-Zagorje, Sisak-Moslavina, Karlovac, Varazdin, Koprivnica- Central Croatia Krizevci, Bjelovar-Bilogora, Medimurje Virovitica-Podravina, Pozega-Slavonia, Slav. Brod-Posavina, Eastern Croatia Osijek-Baranja, Vukovar-Sirmium Zagreb Region Zagreb County, Zagreb City Adriatic North Primorje-Gorski kotar, Lika Senj, Istria Adriatic South Zadar, Sibenik-Knin, Split-Dalmatia, Dubrovnik-Neretva Regional inequality in the poverty incidence is considerable. Table 3.10 reports the poverty headcount rate from below 4 percent in the Zagreb Region and Adriatic North to around 18 percent in Eastern Croatia and over 20 percent in Central Croatia. A poverty gap of 6.2 percent in Central Croatia suggests that poverty in this region is not only widespread but relatively deep as well, and the poor lack substantial resources. The poor in Central Croatia have an average consumption that is 30 percent below the poverty line. A similarly unfavorable situation is found in the Eastern region. The Northern Adriatic is a region with low and relatively shallow poverty. The average poverty deficit of about 13 percent means that the poor in the North Adriatic have average consumption that is 13 percent below the poverty line, a notably better position than in other regions, including Zagreb. The Southern Adriatic lies half-way between low- and high-poverty regions. 20 Appendix B3 gives certain county-level estimates that should be taken as illustrative of a variation in living standard conditions and not as precise point-estimates. 28 Table 2.10: Poverty Incidence by Regions, 2002-2004 Poverty Average poverty Poverty gap Poverty severity headcount rate deficit % s.e. % s.e. % s.e. % s.e. Analytical region Central Croatia 21.2% (1.4) 6.2% (0.6) 2.7% (0.3) 29.4% (1.1) Eastern Croatia 17.5% (1.3) 3.8% (0.4) 1.3% (0.2) 21.9% (1.2) Zagreb Region 3.8% (0.5) 0.8% (0.1) 0.2% (0.1) 19.7% (1.9) Adriatic North 3.7% (0.6) 0.5% (0.1) 0.1% (0.0) 13.1% (1.6) Adriatic South 9.1% (1.0) 1.8% (0.3) 0.6% (0.1) 19.5% (1.6) Croatia 11.6% (0.5) 2.8% (0.2) 1.1% (0.1) 24.2% (0.7) Note: Linearized standard errors based on sample specification are reported in parentheses. Relative poverty risk, defined as a ratio of the region specific poverty rate to the national poverty rate, reflects a substantial variation in the poverty incidence observed above. Combined with a relative proportion of the poor by region it gives an impression of the relative deprivation by region. Table 2.11 shows that the poverty rates in both Central and Eastern Croatia are more than 50 percent higher than the national average. Almost ž of the poor in Croatia live in these two regions, as opposed to less than 1/2 of the total population residing there. The Southern Adriatic experiences a slightly lower poverty risk than the average for Croatia as a whole. The Zagreb Region and Adriatic North are two parts of the country with a relatively low risk of poverty, with around 37 percent of the total population but only 12 percent of the poor in Croatia living there. Table 2.11: Poverty Risk by Regions, 2002-2004 Relative poverty risk Proportion of the poor Proportion of total pop. Risk s.e. Prop. s.e. Prop. s.e. Analytical region Central Croatia 1.831 (0.121) 0.425 (0.022) 0.232 (0.004) Eastern Croatia 1.517 (0.109) 0.300 (0.019) 0.198 (0.004) Zagreb Region 0.330 (0.042) 0.082 (0.010) 0.249 (0.005) Adriatic North 0.318 (0.054) 0.040 (0.007) 0.127 (0.003) 0.784 (0.089) 0.152 (0.016) 0.194 (0.004) Adriatic South Note: Relative poverty risk is a ratio of the region specific poverty rate to the national poverty rate. Linearized standard errors based on sample specification are reported in parentheses. A dimension of regional inequality that usually strongly affects poverty is the level of urbanization. Table 2.12 reports the results of poverty assessment by regions and urbanization level. The variation in the poverty incidence is striking. Rural areas generally face a substantially higher poverty risk ­ that risk is 3 times higher than in urban areas. Poverty rates in rural areas of Central and Eastern Croatia are double the national poverty rate and more than 50 percent of the poor in Croatia live in these two areas. On the other hand, urban Zagreb Region and urban Northern Adriatic are locations with the smallest poverty risk in Croatia, followed by urban parts of Southern Adriatic. Central and Eastern Croatia, even in urban areas, are faced with an above-average poverty risk. 29 Table 2.12: Poverty Risk by Regions and Urbanization Level, 2002-2004 Poverty Relative poverty Proportion of the Proportion of headcount rate risk poor total pop. % s.e. Risk s.e. Prop. s.e. Prop. s.e. Region and urbanization Central Croatia Urban 14.3% (1.5) 1.240 (0.132) 0.098 (0.013) 0.079 (0.006) Eastern Croatia Urban 12.1% (1.4) 1.048 (0.122) 0.098 (0.013) 0.093 (0.007) Zagreb Region Urban 2.6% (0.4) 0.221 (0.036) 0.043 (0.007) 0.195 (0.006) Adriatic North Urban 2.8% (0.7) 0.243 (0.059) 0.020 (0.005) 0.084 (0.005) Adriatic South Urban 4.5% (0.8) 0.386 (0.069) 0.041 (0.008) 0.106 (0.007) Central Croatia Rural 24.7% (1.8) 2.133 (0.158) 0.328 (0.024) 0.154 (0.007) Eastern Croatia Rural 22.4% (1.9) 1.936 (0.167) 0.202 (0.020) 0.105 (0.007) Zagreb Region Rural 8.4% (1.6) 0.725 (0.137) 0.039 (0.008) 0.054 (0.006) Adriatic North Rural 5.4% (1.3) 0.464 (0.109) 0.020 (0.005) 0.043 (0.005) Adriatic South Rural 14.6% (1.9) 1.262 (0.163) 0.111 (0.016) 0.088 (0.007) Location Urban 6.2% (0.4) 0.538 (0.036) 0.300 (0.022) 0.557 (0.014) Rural 18.3% (0.9) 1.580 (0.082) 0.700 (0.022) 0.443 (0.014) Note: Relative poverty risk is a ratio of the region specific poverty rate to the national poverty rate. Linearized standard errors based on sample specification are reported in parentheses. One suspected cause for a large variation in poverty rates by regions is their different level of economic development. The average consumption or average income level could approximate the development level. Table 2.13 shows a substantially lower variation in the average consumption per capita and the average income per capita by regions than was the case for poverty rates. For example, the average income level in the richest region (Zagreb) is some ź above the national average and some 50 percent above the poorest region (Eastern Croatia). Regions with a lower average income and consumption are faced with a higher poverty risk. In other words, differences in the average economic development across regions could only partially explain regional variation in the poverty incidence. Table 2.13: Consumption and Income by Regions, 2002-2004 Consumption per capita Income per capita Index Index s.e. s.e. (Croatia=100) (Croatia=100) Analytical region Central Croatia 83.7 (1.6) 87.0 (1.6) Eastern Croatia 85.0 (2.2) 80.3 (1.5) Zagreb Region 122.1 (1.8) 123.3 (2.0) Adriatic North 114.4 (2.0) 117.1 (2.3) Adriatic South 97.1 (1.5) 94.5 (1.7) Location Urban 112.7 (1.1) 112.8 (1.2) Rural 84.0 (1.2) 83.9 (1.1) Note: Linearized standard errors based on sample specification are reported in parentheses. 30 Variations in the regional poverty rates could be exaggerated due to differences in the regional cost-of-living, especially in food prices and housing rents. However, little can be done to account for the cost-of-living differences across regions in Croatia. The CBS price statistics employ a collection system which ensures that the prices of identical articles are picked up all across the country. Therefore, in the price comparison we could stick to the HBS data, and the only available detailed information on values and quantities are those for food. In fact, we could only calculate unit value prices. However, the unit values vary not only because of different prices, but also due to quality differences. Variations in the rental values across regions could be estimated only by hedonic regressions that are not a robust starting point for the estimation of regional cost-of-living differences. All in all, maneuvering space for a confident estimation of regional price differences is narrow. INTERNATIONAL COMPARISONS In this part of the paper we present international comparisons of poverty and inequality that are based on the World Bank (2005) study on poverty in Eastern Europe and appended by our poverty estimates for Croatia. In this case, calculations for Croatia are prepared following the methodology applied to other countries from the study. As the first step, we have had to construct a material well-being indicator as close as possible to the indicator from the World Bank (2005). Following the explanations from the study, we choose a consumption aggregate not inclusive any estimates of the flow of services of durables, expenditures on durable purchases or rents. Health expenditures were excluded from the consumption aggregate. However, in-kind consumption (food consumption from one's own agricultural activity, consumption in kind of products from other own business activities, value of gifts and transfers in-kind) is included. Consumption data are "trimmed" for extreme values, so they are bottom-coded at 1 percent of the per capita mean real consumption and top-coded at 10 times the median of household consumption. Our consumption aggregate for Croatia is not corrected for regional price differences as opposed to the consumption aggregate for other countries, but this discrepancy will not influence the results of comparisons significantly. All estimates are made on a per capita basis. Purchasing power parity (PPP) for Croatia is estimated by a procedure proposed by the World Bank (2005). PPP for Croatia taken from the OECD (2003) and valid for 2000 was transformed from the euro terms to the US dollar terms, than multiplied by the US inflation rate in the 1993-2000 period, and finally multiplied by the Croatian CPI inflation over the 2000-2004 period. In that way, we have constructed a PPP for Croatia comparable with the other countries from the World Bank study. The poverty rate of PPP$2.15 per capita per day is calculated as PPP multiplied by 2.15, and by 365 to convert to yearly values. An analogue procedure was done for the PPP$4.30 poverty rate. Table 2.14 shows per capita consumption in PPP $ terms, in local currency terms and the food share. Croatia has the highest average per capita consumption among the countries under consideration, some 20 percent higher than Latvia, and some 30 and 40 percent higher than Macedonia and Hungary, respectively. The consumption share of food is around 40 percent, which roughly corresponds to a relatively high per capita consumption, although one could expect an even lower share. A higher than expected food share could be explained by findings from other studies that food prices in Croatia are relatively high, as compared both to other consumption items and other counties in transition (see, for example, OECD 2004, and Eurostat 2005). 31 Table 2.14: Average Consumption and Food Share Year of Average consumption per capita per year Food share survey in PPP $ in local currency in % Croatia 2004 4,156 21,193 41.6 Latvia 2003 3,401 1,051 41.0 Macedonia 2003 3,171 72,735 54.2 Hungary 2002 2,890 417,447 38.7 Lithuania 2003 2,762 4,850 44.5 Estonia 2003 2,753 23,457 42.2 Belarus 2002 2,704 1,399 68.1 Poland 2002 2,611 6,199 39.8 Ukraine 2003 2,496 2,982 72.2 Bulgaria 2003 2,248 1,562 58.7 Russia 2002 2,179 25,467 55.8 Serbia 2002 1,993 85,313 60.8 Turkey 2002 1,816 14,364mio 38.8 Romania 2003 1,624 22,797,434 57.8 Albania 2002 1,388 96,518 61.7 Moldova 2003 1,046 4,573 66.4 Note: Food share is expressed as a percentage share of expenditures on food, beverages and tobacco in the consumption aggregate. Dollar values of average consumption are expressed at 2000 PPP. Source: World Bank (2005) and authors' estimates for Croatia. Table 2.15 reports poverty rates and inequality indices for a number of Eastern European countries. Croatia is the best performer regarding poverty rates. If an international poverty line of $PPP 4.30 per day per person is applied, the poverty rate for Croatia is around 4 percent, much lower than in other countries in the regions for which data are available. If an international poverty line of $PPP 2.15 is used, the proportion of the poor in Croatia falls quite low (to 0.4 percent). It seems that a relatively high level of average consumption is the main factor behind low poverty rates, although inequality could contribute too. The Gini coefficient for Croatia is at the low end in the region and consumption share of the lowest 20 percent is below average. A comparatively low poverty incidence in Croatia stands in contrast with its income level. GDP per capita in PPP terms in Croatia was at around 46 percent of the EU25 average in 2004, which is lower than that for Hungary, Estonia and Lithuania (Eurostat 2005a). There are several possible explanations for the observed inconsistency. First, the HBS in Croatia may be rather comprehensive compared to other surveys in the region, meaning that a larger portion of the total household consumption is covered and more information on consumption is collected. Second, "gray" (or unrecorded) economy may be larger in Croatia than in other countries, especially taking into account its rather large service sector. Third, the actual standard of living in Croatia may be higher than current GDP figures imply. 32 Table 2.15: International Comparison of Poverty and Inequality Year Poverty rate (%) Inequality of survey $PPP $PPP Gini coefficient (per Share of the 4.30/day 2.15/day capita) lowest 20% Croatia 2004 4 0 0.264 9 Hungary 2002 12 0 0.250 10 Latvia 2003 17 3 0.350 7 Belarus 2002 21 2 0.292 9 Ukraine 2003 22 1 0.268 10 Macedonia 2003 24 4 0.373 6 Lithuania 2003 24 4 0.325 8 Estonia 2003 26 5 0.330 7 Poland 2002 27 3 0.320 8 Bulgaria 2003 33 4 0.277 9 Bosnia & 2004 35 4 0.295 9 Herz. Russia 2002 41 9 0.338 7 Serbia 2002 42 6 0.292 9 Romania 2003 58 12 0.288 9 Turkey 2002 58 20 0.393 6 Albania 2002 71 24 0.319 8 Moldova 2003 85 43 0.328 8 Source: World Bank (2005) and authors' estimates for Croatia. CONCLUSION Poverty estimates presented in this Chapter should facilitate a better understanding of recent poverty trends and regional dimension of poverty in Croatia. Somewhat puzzling findings on stagnant consumption-based poverty and a striking variation in poverty incidence by region may remain open to discussion. Certain trends are surely hidden since they were tried to be observed over a relatively short period of time. But if careful observation could continue over a longer period, underlying trends will surely be observed. This study may be seminal in using the proposed methodology for the construction of the consumption aggregate and the poverty line. The poverty line that is defined is, in our view, meaningful, consistent and tailored to satisfy the national purposes. It could be applied for many other poverty calculations and over a longer period of time. It is the author's strong belief that continued efforts in preparing regular poverty estimates that broadly follow the method proposed in this study will result in a number of important insights into the poverty incidence in Croatia. 33 References Eurostat. 2005. "Food, Beverage, Tobacco, Clothing and Footwear: Comparative Price Levels for Five Countries in the Western Balkan Region for 2003", Statistics in Focus ­ Theme Economy and Finance, No. 30/2005. Eurostat. 2005a. "GDP per capita in Purchasing Power Standards", News Release No. 164/2005, 20 December 2005. Eurostat, Luxembourg. OECD. 2003. "Summary Results of ECP 2000: A Note by the OECD Secretariat." Working Paper No 6, Statistical Commission and Economic Commission for Europe, Conference of European Statisticians, Joint Consultation on the ECP. Geneva, 31 March - 2 April 2003. OECD, 2004, Purchasing Power Parity and Real Expenditures: 2002 Benchmark Year. OECD, Paris. Paci, Pierella, Marcin J. Sashin, and Jos Verbeek. 2004. "Economic Growth, Income Distribution, and Poverty in Poland during Transition." World Bank Policy Research Working Paper No. 3467. World Bank, Washington D.C. Ravallion, Martin. 1994. Poverty Comparisons, Fundamentals of Pure and Applied Economics 56. Chur, Switzerland: Harwood Academic Press. Ravallion, Martin. 1998. "Poverty Line in Theory and Practice." LSMS Working Paper No. 133. World Bank, Washington D.C. World Bank, 2000, Croatia: Economic Vulnerability and Welfare Study, Volume II: Technical Papers. Document of the World Bank, Poverty Reduction and Economic Management Unit, Europe and Central Asia Region, World Bank, Washington D.C. World Bank, 2000a, "Making Transition Work for Everyone: Poverty and Inequality in Europe and Central Asia." World Bank, Washington D.C. World Bank. 2005. Growth, Poverty an Inequality: Eastern Europe and the Former Soviet Union. World Bank, Washington D.C. 34 APPENDIX A2 ­ Household Perception of the Living Standard Table A2.1: Households' Opinion about Their Living Standard With its disposable monthly 2002 2003 2004 income, the household lives: In % of In % of In % of s.e. s.e. s.e. total pop. total pop. total pop. With great difficulties 13.4% (0.8) 11.4% (0.7) 10.0% (0.7) With difficulties 25.6% (1.0) 26.7% (1.0) 22.7% (0.9) With some difficulties 29.5% (1.0) 31.8% (1.0) 28.6% (1.1) Fairly well 21.8% (0.9) 18.7% (0.8) 17.5% (0.9) Well 8.0% (0.6) 9.7% (0.7) 19.1% (1.0) Very well 1.7% (0.3) 1.7% (0.3) 2.1% (0.3) Notes: Linearized standard errors based on sample specification are reported in parentheses. Table A2.2: Overlap between Consumption Poverty and Subjective Poverty, 2004 Subjective-poor Subjective-non-poor Total Consumption-poor 4.0% 7.1% 11.1% (0.4) (0.5) Consumption-non-poor 6.0% 82.9% 88.9% (0.5) (0.8) Total 10.0% 90.0% 100.0% Note: Standard errors are reported in parentheses. Subjective-poor are those who report to live with great difficulties on thier disposable income. Table A2.3: Overlap between Income Poverty and Subjective Poverty, 2004 Subjective-poor Subjective-non-poor Total Income-poor 4.4% 6.0% 10.4% (0.4) (0.5) Income-non-poor 5.7% 84.0% 89.6% (0.5) (0.8) Total 10.0% 90.0% 100.0% Note: Standard errors are reported in parentheses. Subjective-poor are those who report to live with great difficulties on their disposable income. 35 APPENDIX B2 ­Poverty Indicators by County Table B2.1: Poverty Incidence by County, 2002-2004 Poverty Poverty gap Poverty severity Average poverty headcount rate P(1) P(2) deficit P(0) % s.e. % s.e. % s.e. % s.e. Zagreb County 6.6 (1.3) 1.5 (0.4) 0.5 (0.2) 23.2 (2.8) Krapina-Zagorje 19.2 (2.8) 4.6 (0.7) 1.6 (0.3) 23.9 (2.0) Sisak-Moslavina 28.3 (3.6) 8.8 (1.5) 3.8 (0.7) 31.1 (2.0) Karlovac 33.8 (5.9) 11.3 (2.5) 5.3 (1.3) 33.4 (3.1) Varazdin 15.6 (2.4) 3.9 (0.8) 1.6 (0.4) 24.7 (2.9) Koprivnica-Krizevci 20.8 (4.3) 6.1 (1.6) 2.5 (0.7) 29.3 (2.9) Bjelovar-Bilogora 21.7 (4.3) 7.0 (1.8) 3.3 (1.2) 32.0 (2.8) Primorje-Gorski kotar 3.4 (0.8) 0.5 (0.1) 0.1 (0.1) 13.6 (2.7) Lika-Senj 2.5 (1.1) 0.2 (0.1) 0.0 (0.0) 8.7 (3.1) Virovitica-Podravina 19.8 (2.2) 4.0 (0.6) 1.3 (0.3) 20.1 (2.4) Pozega-Slavonia 10.2 (3.0) 1.2 (0.5) 0.3 (0.1) 12.0 (2.9) Slav Brod-Posavina 16.4 (3.3) 3.9 (1.0) 1.3 (0.4) 23.5 (3.0) Zadar 8.2 (1.6) 1.1 (0.3) 0.3 (0.1) 13.7 (2.6) Osijek-Baranja 19.9 (2.3) 4.9 (0.8) 1.9 (0.4) 24.6 (2.0) Sibenik-Knin 13.6 (3.4) 2.8 (0.8) 0.9 (0.3) 20.4 (2.8) Vukovar-Sirmium 16.3 (2.2) 3.0 (0.4) 0.9 (0.2) 18.3 (1.7) Split-Dalmatia 8.9 (1.5) 1.9 (0.4) 0.7 (0.2) 21.7 (2.5) Istria 4.4 (1.2) 0.6 (0.2) 0.1 (0.1) 13.3 (2.1) Dubrovnik-Neretva 6.2 (2.0) 1.0 (0.4) 0.2 (0.1) 16.5 (2.8) Medimurje 8.0 (1.9) 1.8 (0.6) 0.6 (0.2) 22.9 (2.5) Zagreb City 2.7 (0.4) 0.4 (0.1) 0.1 (0.0) 16.3 (2.5) Note: Linearized standard errors based on sample specification are reported in parentheses. Poverty calculations are based on the baseline equivalent consumption using the modified OECD scale (1; 0.7; 0.3). 36 Table B2.2: Poverty Risk by County, 2002-2004 Relative poverty risk Proportion of the poor Proportion of total pop. Risk s.e. Prop. s.e. Prop. s.e. Zagreb County 0.570 (0.112) 0.041 (0.008) 0.072 (0.003) Krapina-Zagorje 1.657 (0.240) 0.052 (0.007) 0.031 (0.001) Sisak-Moslavina 2.447 (0.314) 0.103 (0.013) 0.042 (0.002) Karlovac 2.923 (0.507) 0.086 (0.015) 0.029 (0.002) Varazdin 1.352 (0.205) 0.058 (0.009) 0.043 (0.001) Koprivnica-Krizevci 1.799 (0.372) 0.050 (0.011) 0.028 (0.001) Bjelovar-Bilogora 1.877 (0.368) 0.057 (0.012) 0.030 (0.001) Primorje-Gorski kotar 0.293 (0.070) 0.020 (0.005) 0.067 (0.002) Lika-Senj 0.213 (0.095) 0.003 (0.001) 0.013 (0.001) Virovitica-Podravina 1.711 (0.187) 0.036 (0.005) 0.021 (0.002) Pozega-Slavonia 0.885 (0.256) 0.015 (0.004) 0.017 (0.002) Slav Brod-Posavina 1.420 (0.283) 0.055 (0.011) 0.039 (0.001) Zadar 0.710 (0.142) 0.026 (0.005) 0.037 (0.002) Osijek-Baranja 1.719 (0.200) 0.132 (0.015) 0.077 (0.003) Sibenik-Knin 1.176 (0.291) 0.031 (0.008) 0.027 (0.002) Vukovar-Sirmium 1.407 (0.192) 0.062 (0.008) 0.044 (0.002) Split-Dalmatia 0.772 (0.132) 0.080 (0.013) 0.104 (0.003) Istria 0.382 (0.103) 0.018 (0.005) 0.047 (0.002) Dubrovnik-Neretva 0.540 (0.170) 0.014 (0.004) 0.026 (0.001) Medimurje 0.691 (0.169) 0.020 (0.005) 0.028 (0.001) Zagreb City 0.233 (0.037) 0.041 (0.007) 0.177 (0.004) Note: Linearized standard errors based on sample specification are reported in parentheses. Poverty calculations are based on the baseline equivalent consumption using the modified OECD scale (1; 0.7; 0.3). Table B2.3: Groups of Countries by Poverty Risk, 2002-2004 Relative poverty risk: County: Low risk (up to 0.5) Zagreb City, Primorje-Gorski kotar, Istria, Lika-Senj Moderate risk (0.5-1.5) Dubrovnik-Neretva, Zagreb-County, Medimurje, Zadar, Split- Dalmatia, Pozega-Slavonia, Sibenik-Knin, Varazdin, Vukovar- Sirmium, Slav.Brod-Posavina, High risk (1.5-2) Krapina-Zagorje, Virovitica-Podravina, Osijek-Baranja, Koprivnica-Krizevci, Bjelovar-Bilogora Very high risk (above 2) Karlovac, Sisak-Moslavina Source: Author's classification build upon consumption-based poverty estimates by counties. 37 Table B2.4: Consumption and Income by Counties, 2002-2004 Consumption per capita Income per capita Index Index s.e. s.e. (Croatia=100) (Croatia=100) Zagreb County 100.5 (2.7) 99.4 (3.1) Krapina-Zagorje 81.2 (2.5) 87.9 (3.4) Sisak-Moslavina 79.4 (4.1) 80.1 (2.9) Karlovac 76.1 (6.0) 79.6 (4.9) Varazdin 84.5 (2.9) 93.6 (3.5) Koprivnica-Krizevci 82.5 (4.8) 87.8 (6.3) Bjelovar-Bilogora 84.4 (4.7) 86.6 (5.3) Primorje-Gorski kotar 122.1 (2.7) 121.3 (3.2) Lika-Senj 115.1 (4.4) 107.3 (4.9) Virovitica-Podravina 77.9 (4.7) 67.8 (2.6) Pozega-Slavonia 108.7 (19.4) 83.8 (4.3) Slav Brod-Posavina 83.1 (3.0) 86.2 (3.3) Zadar 93.7 (3.0) 86.8 (3.4) Osijek-Baranja 81.3 (2.5) 77.5 (2.6) Sibenik-Knin 93.4 (4.0) 90.7 (4.8) Vukovar-Sirmium 86.9 (2.8) 84.5 (3.4) Split-Dalmatia 97.9 (2.2) 95.8 (2.2) Istria 103.3 (3.4) 113.8 (4.2) Dubrovnik-Neretva 102.6 (3.7) 104.1 (5.3) Medimurje 99.8 (4.2) 93.7 (3.6) Zagreb City 130.9 (2.3) 133.0 (2.5) Note: Linearized standard errors based on sample specification are reported in parentheses. 38 Table B2.5: Ranking of Counties by Living Condition Indicators Consumption per capita Income per capita GDP per capita, Poverty rate 2002-2004 2002-2004 2003 2002-2004 (descending) (descending) (descending) (ascending) Zagreb City Zagreb City Zagreb City Lika-Senj Primorje-Gorski Primorje-Gorski Istria Zagreb City Lika-Senj Istria Primorje-Gorski Primorje-Gorski Pozega-Slavonia Lika-Senj Lika-Senj Istria Istria Dubrovnik-Neretva Koprivnica-Krizevci Dubrovnik-Neretva Dubrovnik-Neretva Zagreb County Varazdin Zagreb County Zagreb County Split-Dalmatia Dubrovnik-Neretva Medimurje Medimurje Medimurje Medimurje Zadar Split-Dalmatia Varazdin Zadar Split-Dalmatia Zadar Sibenik-Knin Karlovac Pozega-Slavonia Sibenik-Knin Krapina-Zagorje Sisak-Moslavina Sibenik-Knin Vukovar-Sirmium Koprivnica-Krizevci Virovitica-Podravina Varazdin Varazdin Zadar Osijek-Baranja Vukovar-Sirmium Bjelovar-Bilogora Bjelovar-Bilogora Split-Dalmatia SlBrod-Posavina SlBrod-Posavina SlBrod-Posavina Bjelovar-Bilogora Krapina-Zagorje Koprivnica-Krizevci Vukovar-Sirmium Zagreb County Virovitica-Podravina Osijek-Baranja Pozega-Slavonia Krapina-Zagorje Osijek-Baranja Krapina-Zagorje Sisak-Moslavina Pozega-Slavonia Koprivnica-Krizevci Sisak-Moslavina Karlovac Sibenik-Knin Bjelovar-Bilogora Virovitica-Podravina Osijek-Baranja SlBrod-Posavina Sisak-Moslavina Karlovac Virovitica-Podravina Vukovar-Sirmium Karlovac Sources: Author's calculations based on the 2002-2004 HBS for consumption, income and poverty rate; the CBS for regional GDP per capita. 39 APPENDIX C2: CONSTRUCTION OF REGIONAL PRICE INDICES Data sources for the construction of regional price indices in Croatia are scarce. The price collection system applied in the construction of the consumer price index is oriented toward measuring changes over time, not across the regions. Therefore, we have to rely on the HBS data. Housing price differences seem to enter the largest noise into the comparison of regional living standards. At this stage, we are trying to account for this problem. The HBS collects information on the number of housing characteristics. At the same time, respondents are asked to assess the rental value of the place that they actually live in, or to report their actual rent if it is agreed on market principles. These two pieces of information were used the basis for estimating regional differences in housing prices. In the first step, the log of rental value is regressed on the number of housing characteristics including 10 dummies for urban/rural regions (see Table C3.1). The estimated coefficient could be treated as percentage differences with respect to omitted characteristics. These coefficients are regional price indices of a sort. However, in the second step we should define the national average price in order to calculate regional price indices. The national price is set at the level to satisfy the condition (for each year) that the national average of the total rental value (across population of individuals) is the same before and after applying regional price indices. In these calculations we have implicitly applied the Pasche price index, meaning that the deflator is different for each household depending on the share of its housing costs in total consumption. Resulting regional price indices for housing are shown in Table C3.2. Urban Zagreb region and urban parts of Northern Adriatic are places with the highest rents in Croatia and rural parts of Central and Eastern Croatia are areas with the lowest rents. Since prices of all other goods and services are taken to be the same all across the country, the overall regional price indices are easily calculated. Once again, the average equivalent consumption is the same with or without adjustment for regional price differences, for each year. Table C3.3 reports poverty regional poverty estimates using consumption aggregate adjusted for regional price differences. Table C2.1: Regression of Log of Rental Value on Housing Characteristics Coeff. (s.e.) P-value Constant 7.498 (0.101) 0.000 Dwelling type Apartment (omitted) House -0.112 (0.040) 0.005 Year of build Before 1918 (omitted) 1918-1945 -0.002 (0.043) 0.965 1946-1960 0.016 (0.040) 0.685 1961-1970 0.051 (0.038) 0.172 1971-1980 0.060 (0.038) 0.113 1981-1990 0.060 (0.039) 0.119 1991-2000 0.113 (0.041) 0.006 2000- 0.205 (0.076) 0.007 Building type Detached house w/ one appt (omitted.) House w/ one apt 0.057 (0.029) 0.051 House w/ two apts 0.060 (0.020) 0.002 Building w/ 3+ apts 0.069 (0.044) 0.114 No. of rooms Studio (omitted) 1-room 0.140 (0.062) 0.025 2-rooms 0.310 (0.063) 0.000 3-rooms 0.424 (0.067) 0.000 4-rooms 0.457 (0.073) 0.000 5-rooms 0.574 (0.083) 0.000 6+ rooms 0.531 (0.114) 0.000 Location Central Croatia urban (omitted) Central Croatia rural -0.239 (0.031) 0.000 Eastern Croatia urban 0.009 (0.028) 0.738 Eastern Croatia rural -0.190 (0.031) 0.000 Zagreb Region urban 0.589 (0.026) 0.000 Zagreb Region rural 0.264 (0.037) 0.000 Adriatic North urban 0.410 (0.031) 0.000 Adriatic North rural 0.237 (0.036) 0.000 Adriatic South urban 0.538 (0.029) 0.000 Adriatic South rural 0.090 (0.038) 0.020 Amenities Hot water 0.054 (0.026) 0.039 Collective central heating 0.121 (0.029) 0.000 Own central heating 0.145 (0.018) 0.000 Balcony 0.092 (0.014) 0.000 Garden -0.036 (0.017) 0.038 Kitchen (in-house) 0.237 (0.035) 0.000 Bathroom (in-house) 0.110 (0.035) 0.002 Complains Rotten -0.132 (0.024) 0.000 Humidity -0.049 (0.028) 0.077 Lack of lights -0.051 (0.040) 0.203 Installation Running water 0.219 (0.046) 0.000 Sewage 0.125 (0.021) 0.000 Gas 0.114 (0.016) 0.000 Telephone 0.107 (0.028) 0.000 Usable space Space (m2) 0.008 (0.001) 0.000 Space squared (*1000) -0.016 (0.003) 0.000 No. of obs. 5,827 Adj. R2 0.630 Note: Robust standard errors are reported in parentheses. Based on pooled sample of the HBS data for 2003 and 2004. 42 Table C2.2: Regional Price Indices Regional price index for housing Overall regional price index (Croatia=1.0) (Croatia=1.0) Central Croatia urban 0.825 0.964 Central Croatia rural 0.628 0.924 Eastern Croatia urban 0.832 0.965 Eastern Croatia rural 0.669 0.936 Zagreb Region urban 1.310 1.064 Zagreb Region rural 1.043 1.008 Adriatic North urban 1.163 1.032 Adriatic North rural 1.020 1.004 Adriatic South urban 1.268 1.046 Adriatic South rural 0.899 0.982 Note: Based on regression estimates of the rental values of dwellings. Overall regional price index takes into account only the regional differences in housing prices whereas other prices are taken to remain the same all across the country. Price indices are of the Paasche-type. Table C2.3: Poverty Incidence and Consumption Adjusted for Regional Price Differences by Counties, 2002-2004 Baseline consumption Consumption adjusted for regional price differences (unadjusted) Poverty headcoun Poverty headcoun Average poverty Consumption per capita rate rate deficit (Croatia=100) % s.e. % s.e. % s.e. % s.e. Zagreb County 6.6% (1.3) 7.9% (1.4) 22.5% (2.6) 98.0 (2.5) Krapina-Zagorje 19.2% (2.8) 16.8% (2.5) 21.7% (1.8) 88.0 (2.7) Sisak-Moslavina 28.3% (3.6) 26.9% (3.7) 29.9% (2.0) 83.4 (4.1) Karlovac 33.8% (5.9) 31.9% (5.8) 33.2% (3.2) 80.8 (6.2) Varazdin 15.6% (2.4) 12.7% (2.2) 25.8% (3.1) 90.4 (2.9) Koprivnica-Krizevci 20.8% (4.3) 19.5% (4.1) 27.8% (3.0) 88.2 (4.9) Bjelovar-Bilogora 21.7% (4.3) 20.8% (3.9) 31.1% (3.3) 88.4 (4.9) Primorje-Gorski kotar 3.4% (0.8) 4.2% (0.9) 15.0% (2.4) 119.0 (2.6) Lika-Senj 2.5% (1.1) 2.5% (1.1) 11.7% (3.0) 114.2 (4.2) Virovitica-Podravina 19.8% (2.2) 18.5% (2.1) 18.3% (2.6) 81.9 (4.7) Pozega-Slavonia 10.2% (3.0) 7.2% (2.8) 11.7% (2.4) 113.4 (19.5) Slav Brod-Posavina 16.4% (3.3) 14.6% (3.1) 23.8% (2.7) 88.4 (3.2) Zadar 8.2% (1.6) 9.7% (1.8) 14.5% (2.4) 93.4 (2.9) Osijek-Baranja 19.9% (2.3) 17.7% (2.3) 25.6% (2.2) 85.1 (2.6) Sibenik-Knin 13.6% (3.4) 14.1% (3.3) 20.7% (2.7) 93.8 (3.9) Vukovar-Sirmium 16.3% (2.2) 14.0% (1.9) 17.2% (1.8) 91.3 (2.8) Split-Dalmatia 8.9% (1.5) 10.3% (1.6) 21.2% (2.4) 95.1 (2.0) Istria 4.4% (1.2) 6.1% (1.4) 13.9% (1.8) 101.2 (3.3) Dubrovnik-Neretva 6.2% (2.0) 7.5% (2.1) 18.7% (2.8) 100.9 (3.6) Medimurje 8.0% (1.9) 6.9% (2.0) 21.2% (2.6) 108.6 (4.7) Zagreb City 2.7% (0.4) 4.7% (0.6) 16.4% (1.8) 123.3 (2.1) Note: Linearized standard errors based on sample specification are reported in parentheses. Poverty calculations are based on the baseline equivalent consumption using the modified OECD scale. 43 Table C2.4: Poverty Incidence and Consumption Adjusted for Regional Price Differences by Regions, 2002-2004 Baseline consumption Consumption adjusted for regional price differences (unadjusted) Poverty Poverty Average Consumption per capita headcount rate headcount rate poverty deficit (Croatia=100) % s.e. % s.e. % s.e. % s.e. Central Croatia 21.2% (1.4) 19.4% (1.4) 28.6% (1.2) 89.3 (1.6) Eastern Croatia 17.5% (1.3) 15.4% (1.2) 22.1% (1.3) 89.3 (2.2) Zagreb Region 3.8% (0.5) 5.6% (0.6) 18.9% (1.6) 116.0 (1.7) Adriatic North 3.7% (0.6) 4.8% (0.7) 14.3% (1.4) 112.0 (2.0) Adriatic South 9.1% (1.0) 10.3% (1.1) 19.6% (1.5) 95.4 (1.4) Urban 6.2% (0.4) 7.2% (0.4) 19.9% (1.0) 109.7 (1.0) Rural 18.3% (0.9) 17.1% (0.9) 25.3% (0.9) 87.8 (1.2) Note: Linearized standard errors based on sample specification are reported in parentheses. Poverty calculations are based on the baseline equivalent consumption using the modified OECD scale. 44 BACKGROUND PAPER #3 A POVERTY PROFILE FOR CROATIA 2004 Giovanni Vecchi* Department of Economics University of Rome "Tor Vergata" Abstract This paper presents the essential facts on poverty in Croatia according to the Household Budget Survey dataset for the year 2004. The focus is on two questions: (i) Who and where are the poor?, and (ii) What are the micro-determinants of poverty? The paper presents an eclectic mix of descriptive material including standard contingency tables and graphs, but also micro-simulations based on multivariate regression analysis. We find that the region of residence and the educational attainment of the household head are the salient determinants of poverty in Croatia. We also find that social safety nets are inadequate to protect the retired, the unemployed and other inactive persons from poverty. Finally, we find that low labor market activity rates and old age poverty should be priority issues in the poverty reduction strategy of the country. INTRODUCTION The estimates provided in this paper update the poverty profile contained in World Bank Report no. 22079, Croatia: Economic Vulnerability and Welfare Study, based on the 1998 HBS data. Poverty comparisons between 2004 and 1998, however, are flawed by a number of inconsistencies in the sampling design underlying the two surveys. The single most important difference is in the sample coverage, and is caused by the fact that the 1998 HBS did not collect information from areas of Croatia occupied during the war. These areas are thought to be significantly poorer than average ­ see Luttmer (2000) ­ which would distort comparisons of both poverty levels and patterns between the two surveys. Four assumptions underlie the results discussed in the paper. First, our welfare measure is total household (not family) expenditure (not income, even if available in the HBS datasets). Following the guidelines provided in Deaton and Zaidi (2002), we defined total household expenditure as including (i) self-reported rental values of dwellings, (ii) (estimated) user costs of consumer durables, and (iii) the value of food items produced at home or received as gifts. Second, household expenditures were adjusted using the so-called OECD-II equivalence scale. The use of the equivalence scale allowed us to account for different needs among households with different size and age structures. Third, no spatial price adjustment was applied to household expenditures, in consequence of the lack of a suitably disaggregated price index. Finally, the absolute poverty line used to mark the threshold of expenditure below which households are classified as poor was estimated using the method described in Ravallion (1994). The line used throughout this paper is * I would like to thank Nicola Amendola, Daniel Nesti and Salman Zaidi for their helpful (and countless) comments. The usual disclaimer applies. Correspondence: giovanni.vecchi@uniroma2.it. 45 equal to 22,145 HRK per adult-equivalent per year. A complete description of these methodological issues is contained in a companion paper in the same volume of this report. The paper is organized as follows. Section 2 contains a description of the characteristics of poverty. We compare poor versus non-poor families in terms of a number of both monetary and non-monetary indicators. Section 3 constructs the poverty profile for Croatia; in addition to estimating the scale of poverty (how many the poor are and how poor they are) it identifies the population groups most at risk of poverty. The main limitation of the analysis carried out in sections 2 and 3 is that they are both based on simple correlations. Simple correlations are often informative statistics, but they may as often be a poor guide in assessing causality. For instance, the fact that households in urban areas have a lower probability of being poor than households in rural areas may have little to do with the characteristics of the areas, rather than the differences in the individuals who live in those areas. For instance, individuals in urban areas may be more educated than their rural counterparts. In section 0 we analyze the determinants of consumption (and hence poverty) by running micro- simulations. We regress log per adult-equivalent consumption on a number of poverty correlates, and use the parameter estimates to simulate the poverty rates that would be observed if households were given certain socio-demographic characteristics. The comparison between simulated and actual poverty rates provides useful evidence for assessing the relative importance of poverty determinants. Section 0 summarizes the main findings and offers concluding remarks. THE CHARACTERISTICS OF POVERTY In this section we use the 2004 household survey data to compare poor versus non-poor families on a number of poverty dimensions. By analyzing indicators other than income and/or expenditure levels we will investigate the factors that make the life of the poor different from the nonpoor, thereby initiating the identification of poverty correlates. Disadvantageous consumption pattern. Figure 3.1 compares the budget shares, that is the expenditures on a group of goods as fraction of total expenditure, between poor and non-poor households. These basic data reveal that: ˇ The budget share absorbed by food consumption exceeds 50 percent among the poor. As expected, poor households spend proportionally more on food (53 percent) than the non-poor households (37 percent). ˇ Out of the 50-60 percent of total expenditure not devoted to food, housing is the most important item: it absorbs 20 percent of the budget among the poor, compared to 13 percent among the non poor. ˇ Other expenditures (which include health, education, and recreational items) and transport and communication are about equally important (8-10 percent among the poor, 15-17 percent among the non poor). ˇ Expenditures for clothing, liqueur and durable goods come last with roughly 5 percent. The share devoted to clothing shows the largest difference between poor and non-poor households (less than 3 percent among the former, more than 7 percent among the latter). 46 Figure 3.1: Expenditure Patterns of the Poor and the Nonpoor 53 nonpoor 50 poor Budget shares (%) 40 37 30 20 20 17 15 13 10 10 8 7 5 5 4 3 2 0 food other transport utilities clothing durables liqueur and and communication tobacco Source: Author's estimates on 2004 data from HBS. On average, the diet of the poor meets caloric norms ... According to our estimates based on the 2004 HBS, the average calorie intake in Croatia amounts to 3,333 kilocalories/equivalent adult/day. Among the poor, the calorie intake amounts to 2,715 kilocalories/adult/day (2,789 kcal in rural areas, and 2,519 in urban areas), which compares to 3,410 kcal/adult/day among non-poor households (3,596 kcal and 3,264 kcal in rural and urban area, respectively). ... but average values may mask pockets of severe deprivation. The fact that the average calorie intake among the poor roughly coincides with the minimum energy intake (2.7000 kcal/adult/day according to FAO 2004), suggests that extreme poverty may be present in the country.21 The diet of the poor is little varied, but ­ on average ­ nutritional standards are met. With regard to the adequacy of the diet, Figure 3.2 shows that: ˇ Carbohydrates dominate the diet of the Croatian population, even if the differences between the poor and non-poor households are not large. ˇ There is no sign of protein deficiency; the average daily intake is safely above the minimum daily requirement 62 grams/day. This holds true for both non-poor households (108 grams/day) and the poor (88 grams/day). 21 According to FAO (2004), 2700 kcal/day is the minimum energy requirement after assuming a reference person with the following characteristics: male, aged 18-30, weighing between 65 to 70 kilograms, with a basal metabolic rate (BMR, that is the energy required for sustaining the basic functions of the body) equal to approx. 25.3, and enjoying a "lightly active lifestyle" (that is with "physical activity level" (PAL) set equal to 1.6). 47 Figure 3.2: Nutritional Assessment of the Diet of the Poor 400 nonpoor poor grams/equivalent adult/day 300 200 100 0 carbohydrates fats proteins Source: Author's estimates on 2004 data from HBS. Figure 3.3 shows how the sources of income of the poor differ from those of the non poor. Three main points stand out. First, the poor get the largest share of their income from transfers (pensions plus state transfers account for 57 percent of total income). In contrast, transfers represent 31 percent of total income among the nonpoor. Second, the share of income from self-employment is 5 percent among the poor and 10 percent among the non-poor. Third, the structure of income in Figure 3.3 suggests that, overall, the poor get a small fraction of their resources from productive activities: wages plus self-employment income plus in-kind income contribute to less than one-forth of total income. This is consistent with both the low labor force participation rates among the poor, and their relatively high age, as shown in Table 3.1 below. Figure 3.3: Sources of Income: A Comparison between Poor and Nonpoor Households 40 38 nonpoor 35 poor 30 Income shares (%) 22 20 18 19 19 12 10 10 9 5 6 4 3 2 0 wages pensions imputed self-empl. State in-kind other rent income transfers income sources Source: Author's estimates on 2004 data from HBS. In Table 3.1: we compare a selection of demographic and socio-economic indicators between poor and non-poor households. 48 Table 3.1: Characteristics of Poverty Characteristics of household Urban Rural Overall or household head Poor Not poor Poor Not poor Poor Not poor All Gender (% individuals in households with head with given characteristic) Female household head 40 27 31 15 33 22 23 (6) (1) (3) (1) (3) (1) (1) Age (average) Household Head 66 55 66 56 66 55 57 (1.7) (0.5) (0.8) (0.5) (0.8) (0.4) (0.4) Education (% individuals in households with head with given characteristic) Unfinished primary 38 5 40 17 39 11 14 (6) (1) (4) (1) (3) (1) (1) Primary 26 12 41 25 37 18 20 (5) (1) (4) (2) (3) (1) (1) Vocational Secondary 27 28 16 36 19 32 30 (5) (2) (3) (2) (3) (1) (1) General Secondary 9 28 3 13 5 22 20 (3) (1) (1) (1) (1) (1) (1) Post Secondary 0 27 0 8 0 19 17 (0) (2) (0) (1) (0) (1) (1) Employment status (% individuals in households with head with given characteristic) Employee 20 48 11 38 14 44 41 (6) (2) (3) (2) (3) (1) (1) Self-employed 4 12 30 28 23 19 19 (2) (1) (4) (2) (3) (1) (1) Unemployed 12 2 5 3 7 2 3 (3) (0) (2) (1) (2) (0) (0) Retired 58 35 42 29 47 32 34 (6) (2) (4) (2) (3) (1) (1) Other inactive 6 3 11 2 9 3 4 (2) (0) (2) (1) (2) (0) (0) Housing and access to basic services (individuals, %) Less than 10 m2 per person 19 4 8 2 11 3 4 (6) (1) (2) (1) (3) (0) (1) No electricity 1.3 0.1 2.2 0.2 1.9 0.1 0.3 (0.9) (0.1) (0.5) (0.1) (0.7) (0.1) (0.1) No water supply 12 1 29 10 24 5 7 (3) (0) (4) (2) (3) (1) (1) No connection to sewage 16 5 56 33 45 17 20 (4) (1) (4) (3) (4) (2) (2) No telephone 27 5 31 7 30 6 8 (5) (1) (3) (1) (3) (0) (1) No toilet in dwelling 23 3 39 8 35 5 8 (6) (1) (4) (1) (1) (1) (1) Durable Goods (individuals, %) No car 85 27 79 22 81 25 31 (5) (1) (3) (1) (3) (1) (1) No TV set 10 1 15 2 13 1 3 (4) (0) (2) (0) (1) (0) (0) No personal computer 95 53 97 71 96 61 65 (3) (2) (2) (2) (1) (1) (1) No refrigerator 13 4 12 5 13 4 5 (4) (1) (2) (1) (2) (0) (0) No washing machine 25 7 46 10 40 8 12 (4) (1) (4) (1) (3) (1) (1) 49 Characteristics of household Urban Rural Overall or household head Poor Not poor Poor Not poor Poor Not poor All Indebtedness (% individuals) No savings (last 12 months) 6 25 3 21 4 10 13 (5) (2) (1) (2) (2) (0) (0) Payed off loan (last 12 15 46 13 35 13 41 38 months) (5) (2) (3) (2) (2) (1) (1) Subjective Income Poverty (individuals, %) Disposable income makes 39 10 35 10 36 10 13 life difficult/very difficult (3) (1) (2) (0) (2) (0) (0) Health (% individuals) Bad/Very Bad Health (self- 36 14 44 15 42 14 18 reported) (5) (1) (2) (1) (2) (1) (1) Note: standard errors (corrected for sample design) in parentheses. Source: Author's estimates on 2004 data from HBS. A number of points are worth underlining in Table 3.1: ˇ Gender. Poor households are female-headed more frequently than the non-poor ones (33 percent versus 22 percent, respectively). ˇ Age. The average age of the head of the household among the poor is 66 years, compared to 55 years among the non poor. ˇ Education. Around 75 percent among the poor live in households headed by individuals who attained at most the primary level of education, compared to 30 percent among the non-poor. Only 5 percent of the poor live in households, whose head has completed general secondary school. ˇ Employment status. About one half of the poor live in households headed by retired individuals. Labor force participation seems to offer relative protection against poverty, as shown by the fact that 37 percent of the poor are classified as occupied. ˇ Housing and access to basic services. Table 3.1 points to a surprisingly high percentage of individuals, not only among the poor, with limited or no access to basic services. About 8 percent of the population live in dwelling without toilet, and 8 percent without a telephone line. There is also sign of severe deprivation, as suggested by the fact that one fourth of the poor live in dwellings without water supply. Overall, the distance separating the poor from the non poor seems remarkably large when measured in terms of access to basic services. ˇ Savings and access to credit. The evidence in Table 3.1suggests that (i) the income of the poor does not allow positive savings (only 4 percent of the poor report positive savings during the recall period); (ii) only 13 percent of the poor report access to borrowing (from either the banking system or intermediaries other than relatives) during the last 12 months. The combination of low capacity of saving with limited access to borrowing suggests that the poor are also vulnerable to shocks and hence to income fluctuations. WHO AND WHERE ARE THE POOR? This section provides a description of poverty rates across Croatia's regions and population groups. Three key questions are addressed: (i) Who are the poor?, (ii) Where do they live?, and (iii) What do they do? The scale of poverty in Croatia can be captured by the estimates shown in Table 3.2. Schematically, the main results are as follows. 50 ˇ In 2004 almost half a million people ­ representing about 11 percent of the Croatian population ­ lived in poverty. Individuals in poverty consumed less than the poverty line of 22,145 HRK per equivalent adult per annum. ˇ Rural populations are three times more likely to be poor than urban populations. The evidence points to the existence of a considerable gap between urban and rural areas, both in terms of the incidence of poverty (17 percent in rural areas versus 5.7 percent in urban areas) and its depth (4.2 percent versus 1.2 percent). ˇ Poverty is shallow: on average, the poor have an expenditure shortfall of ca. 24 percent of the poverty line. The "depth" of poverty, as measured by the poverty gap index, amounts to 2.6 percent. It is the average distance (expressed as a percentage of the poverty line) separating the population from the poverty line, with the non-poor assigned a distance of zero. ˇ The cost of eradicating poverty under the (highly unrealistic) assumption of perfect targeting, is approximately 1.5 billion HRK. This represents about 0.7 percent of the Croatian GDP in 2004 prices. ˇ Shallow poverty is associated with relatively severe poverty. According to the estimates in Table 3.2, the "severity" of poverty (measured by the squared poverty gap) is about 1 percent. If all inequality among the poor was removed (for instance by a mean-preserving redistribution) the squared poverty gap would decrease from 1 percent (actual) to 0.6 percent. This finding suggests that there are some groups in the population who are likely to experience extreme poverty, an issue that calls for further investigation. Table 3.2: Estimates of Absolute Poverty for Croatia 2004 CROATIA RURAL URBAN National absolute poverty line = 22,145 HRK/year/equiv. adult Headcount Ratio (%) 11.1 17.0 5.7 95% confidence interval [9.4, 12.8] [13.9, 20.2] [4.1, 7.4] Poverty Gap (%) 2.6 4.2 1.2 Poverty Gap Squared (%) 1.0 1.6 0.4 Number of poor persons 468,170 340,355 127,715 Relative Poverty Risk 1.0 1.5 0.5 International poverty line = 4.30 USD/day/at PPP Headcount Ratio (%) 4.0 5.6 2.5 Poverty Gap (%) 0.9 1.3 0.6 Poverty Gap Squared (%) 0.3 0.5 0.2 Relative Poverty Risk 1 1.4 0.6 Background statistics Population share 100.0 47.2 52.8 Average expenditure 43,229 36,634 49,035 Average expenditure of the poor 16,864 16,641 17,453 Average poverty gap 5,281 5,504 4,692 Gini Index 25.3 24.2 24.1 Note: See Table in the Appendix for the estimated standard errors. Source: Author's estimates on 2004 data from HBS. 51 Table 3.3 shows the results after slicing the distribution of per equivalent adult expenditure into intervals, either centered around the poverty line (z) or defined as a proportion of it, and counts how many individuals fall within each interval. The purpose of this exercise is to come up with a rough estimate of households' vulnerability to poverty, loosely defined. In column one the categories have been named in completely arbitrary manner with the only purpose of facilitating communication. Table 3.3: Poverty Bands Poverty Bands Population Share (%) Cum. (%) CROATIA 4,227,000 100.0 100.0 Non-poor PEA > 2z 1,682,511 39.8 100.0 Transient non-poor 1.25z < PEA < 2z 1,629,140 38.6 60.2 Transient vulnerable z < PEA < 1.25z 447,178 10.6 21.7 Transient poor 0.75z < PEA < z 283,146 6.7 11.1 Chronically poor 0.5z < PEA < 0.75z 143,225 3.4 4.4 Extremely poor PEA < 0.5z 41,800 1.0 1.0 Notes: PEA is per equivalent adult expenditure, z is the absolute poverty line, equal to 22,145 HRK/equivalent adult/ year. By reading Table 3.3 bottom up we obtain an account of how rapidly the count of the poor changes in response to changes in the poverty line. For instance, the last row shows that about 40 thousand individuals (1 percent of the population) live with a desperately low level of expenditures (less than half the poverty line). If the threshold of poverty is raised up to three-quarters of the poverty line, then about 4.4 percent of individuals are classified as poor. In the presence of such a low level of expenditure poverty is likely to persist over time, which is why this group is labeled "chronic poor" in Table 3.3. According to Table 3.3 a considerable number of poor (280 thousand individuals) are concentrated just below the poverty line: these households (`transient poor') representing almost 7 percent of the entire population, need relatively little extra purchasing power in order to be lifted from poverty. On the other hand, an even more numerous group of people (447 thousand individuals) is located just above the poverty line ­ they were not poor at the time of the survey, but were close enough to the poverty line to be classified as `transient non-poor', to emphasize that they face a relatively high risk of becoming poor in the future. Figure 3.4 builds on the analysis carried out in Table 3.3, and shows how the headcount poverty ratio varies as a function of the poverty line. The curve shows that the incidence of poverty increases at a relatively slow pace for poverty lines up to approximately 10,000 HRK/equiv. adult/year. Thereafter, the shape of the curve becomes clearly non-linear, implying that the incidence of growth increases more than proportionally with the poverty line. This suggests that even relatively small economic shocks may be responsible for large changes in poverty rates. 52 Figure 3.4: Headcount Poverty Ratio as a Function of the Poverty Line 100 80 Poverty Incidence (%) 60 40 20 0 5000 15000 25000 35000 45000 55000 65000 75000 85000 95000 Poverty Line (HRK/equivalent adult/year) Source: Author's estimates on 2004 data from HBS. A Digression: International Poverty Comparisons Poverty incidence in Croatia is, perhaps, surprisingly lower than in other countries of the region. Figure 3.5 compares Croatia with other countries for which poverty estimates are available and consistently defined. Croatia stands out as an outlying observation, far below its predicted value (the quadratic fit represented by a dotted line). There are several possible reasons that aid in the explanation of this apparent anomaly. First, the quality of the household survey in Croatia (that is, its comprehensiveness) may be higher than in other countries. Second, the extent of underestimation of the country's gray economy may be greater than in other countries. Third, the overall actual standard-of-living in Croatia may be somewhat higher (that is, in comparison to other countries) than previously thought. Figure 3.5: Cross-country Poverty Comparisons 60 ROMANIA Headcount Poverty Ratio 40 BULGARIA (%) POLAND ESTONIA LITHUANIA 20 LATVIA HUNGARY CROATIA 0 30 40 50 60 GDP per capita (PPS, EU-25=100) Source: Author's calculations based on the HBS (2004) and World Bank (2005). 53 Location and Poverty Rural households are three times more likely to be poor than their urban counterparts. Rural poverty is also more deep and severe than poverty in urban areas. On average, the consumption of the rural poor is 25 percent below the poverty line, compared to 20 percent for the urban poor. The squared poverty gap is four times higher in rural areas (1.6 percent) than in its urban counterparts (0.4 percent). There are large regional differences in the extent of poverty. As shown in Figure 3.6, the incidence of poverty ranges from about 3 percent the Zagreb region to 19 percent in the Eastern region. Differences are even larger between urban and rural regions (Table in the Appendix). Even after accounting for the configuration of the Croatian territory, it is striking to observe- 1:6-difference in poverty rates between the poorest and richest regions. The identification of the factors that underlie the regional variation of poverty rates deserves a high priority in analyzing poverty in Croatia. Section 4 explores the extent to which differences in educational attainment levels (as a proxy for the stock of human capital) can explain the regional disparities in poverty rates. Other factors such as physical capital endowments, presence of infrastructure, political and institutional framework should also be investigated, though it is beyond the scope of the present paper. Figure 3.6: Poverty Incidence in Croatia by Region National Average Zagreb Adriatic North Adriatic South Central Eastern 0 5 10 11.1 15 20 Headcount Poverty Ratio (%) Source: Author's estimates on 2004 data from HBS. More than 70 percent of all poor individuals are concentrated in the Central and Eastern regions. They account for 43 percent of the population. Figure 3.7 shows the distribution of the poor by region. 54 Figure 3.7: Distribution of Poverty by Region 15% 5% 38% 7% Regions: Central 34% Eastern Zagreb Adriatic North Adriatic South Source: Author's estimates on 2004 data from HBS. Female-headed households have a higher risk of poverty than their male counterparts. Overall, one-third of the population lives in households headed by a female. The incidence of poverty among female-headed households is 16 percent, compared to 9.6 percent among male-headed households. The risk of poverty increases with age. Figure 3.8 shows that the incidence of poverty is highest among households headed by the elderly. They face a poverty risk twice the average (Table 3.6). Individuals belonging to households headed by a female aged 65 or above have the highest incidence of poverty, equal to 26 percent (12 percent and 30 percent in urban and rural households, respectively ­ see Table in the Appendix). Figure 3.8: Poverty Incidence by Age of the Household Head 26.4 male 25 female Headcount Poverty Ratio (%) 20 18.8 15 National Average 10.9 10 9.4 8.6 6.2 5.1 5 2.1 0 16-30 31-49 50-64 65+ Age of Household Head Source: Author's estimates on 2004 data from HBS. 55 At the individual level, Figure 3.9 shows that the incidence of poverty (left axis) is remarkably flat over the life cycle, but surges when it comes to the elderly. The pattern is by and large unaltered by the consideration of the poverty gap index (right axis). With regard to the depth of poverty, however, a peak is observed among the youngest children (aged 0-4), who have the second highest value for the poverty gap index. This suggests that households with babies stand out as a group deserving special attention: their risk of poverty is similar to households with older kids, but their hardship is significantly higher. Figure 3.9: Poverty Incidence over the Life Cycle 25 6 headcount ratio poverty gap 5 Headcount Poverty Ratio (%) 20 Poverty Gap Index (%) 4 15 3 10 2 5 0-4 5-9 10-15 16-24 25-34 35-44 45-54 55-64 65+ Age Source: Author's estimates on 2004 data from HBS. Protection offered by pensions is not sufficient in helping the elderly to overcome the risk of poverty. Focusing on the 65+ group of individuals, we find that the risk of poverty of the elderly with pension is almost twice the average risk. This category represents 28 percent of the poor. Not having a pension (12 percent of the poor belong to this category) raises the risk, dramatically, to 3.5 times the average risk. Within the group of households headed by the 65+ individuals, those without a pension have a poverty risk more than five times the national average (Figure 3.10). Since one-fourth of the population belongs to households headed by 65+, they account for almost 50 percent of the poor (Figure 3.11). The protection offered by the household to its 65+ members without a pension is significant but far from being able to fill the gap left by the social security system. A comparison between the relative poverty risk of an elderly individual heading a household with the risk for an elderly person not heading a household, may be used as a proxy, admittedly crude, of the extent to which households offer protection against poverty in the absence of a pension. We find that being elderly and not head of the household decreases the relative poverty risk by 40 percent compared to elderly heads of households. 56 Figure 3.10: Incidence of Poverty and Pension Receipts among the Elderly 65 62 60 Headcount Poverty Ratio (%) 50 40 30 19 20 National Average 9 10 7 3 0 16-30 31-49 50-64 65+ pens. 65+ no pens. Age of Household Head Source: Author's estimates on 2004 data from HBS. Figure 3.11: Poverty Share by Age of the Household Head 100 23% 80 Share of the poor (%) 27% 60 42% 40 Age of Household Head 16-30 31-49 20 50-64 65+ pensioners 7% 65+ w/o pension 0 Source: Author's estimates on 2004 data from HBS. According to the 2004 HBS data, average household size in Croatia is 2.7 persons (2.6 in urban areas, 2.9 in rural areas). The link between poverty and household size is best characterized by the following three issues: ˇ Figure 3.12 shows an U-shaped pattern between poverty risk and household size. Small households (1 and 2 person) and large households face above-average poverty risk. ˇ According to Table B3.12, there are considerable differences across rural and urban areas. One-person rural households face the highest risk of poverty (about 3.5 higher than average), followed by rural two-person households (2 times the average). Urban households with 6+ persons face a higher poverty risk than all other household size-groups in urban areas, but the size of standard errors requires caution. 57 ˇ Figure 3.12 shows that the distribution of the poor by household size is pretty uniform, with the exception of two-person households, a phenomenon explained by the fact that a two- person household is the modal size, accounting for 30 percent of the population. Figure 3.12: Poverty Incidence by Household Size and Urban/Rural Areas 40 39 rural urban Headcount Poverty Ratio (%) 30 23 20 16 15 12 11 11 11 10 10 6 3 1 0 1 person 2 persons 3 persons 4 persons 5 persons 6+ persons Household Size Source: Author's estimates on 2004 data from HBS. Figure 3.13: Distribution of the Poor by Household Size and Urban/Rural Areas 30 18 rural urban Share of Poor (%) 20 14 12 11 11 10 7 10 5 3 3 4 1 0 1 person 2 persons 3 persons 4 persons 5 persons 6+ persons Household Size Source: Author's estimates on 2004 data from HBS. Education. Like in most other countries across the world, Croatia shows a strong negative correlation between poverty risk and the level of education of its head of household. Figure 3.14 shows the pattern of poverty risk by educational level of the household head. The covariation is clearly negative, but does not vary with the urban/rural location. Irrespective of the educational level, however, rural households face systematically greater poverty incidence rates than urban households. Secondary education stands out as a threshold above which the probability of being poor becomes lower than the national average. 58 Figure 3.14: Incidence of Poverty by Educational Attainment of the Household Head 32 32 31 croatia 30 rural urban 25 Headcount Povertt Ratio (%) 21 20 12 National Average 10 8 7 5 5 2 2 0 1 0 0 unfinished primary vocational general post primary secondary secondary secondary Educational Attainment of the Household Head Source: Author's estimates on 2004 data from HBS. More than three-quarters of the poor live in households headed by individuals with primary or lower level of education (Figure 3.15). The share of the poor decreases monotonically with the educational attainment of its head of household Figure 3.15: Poverty Shares by Educational Attainment of the Household Head 5% 19% 39% Educational Attainment Unfinished Primary 37% Primary Vocational Secondary General Secondary Post Secondary Source: Author's estimates on 2004 data from HBS. Poverty is tightly associated with the activity status of the main breadwinner. Households headed by a "retired", "unemployed", or "other inactive" person (i) show the highest rates of poverty incidence (the peak of 47 percent belongs the other inactive in the rural areas ­ see Figure 3.16), and (ii) represent 62 percent of the total poor (Figure 3.17). 59 Retirement doubles the risk of poverty in rural, but not in urban areas. The incidence of poverty among households headed by a retired person is below the average in urban areas (9 percent) but close to twice the average in rural households (Figure 3.16). This can be explained by the following three factors: (i) the proportion of the population living in households headed by 65+ individuals without pension is 2 percent in rural areas, compared to 0.3 percent in urban areas, and (ii) individuals in urban areas benefit from a higher degree of protection from other household members than their rural counterparts (about 87 percent of households headed by 65+ individuals without pension live in rural areas). (iii) Given the contributory pension system in Croatia, pensions in rural areas may be significantly lower than in urban areas. Figure 3.16: Poverty Incidence by Employment Status of the Head of the Household 47 other inactive 10 28 unemployed 26 23 retired 9 18 self-employed 2 6 employee rural 2 National Average urban 0 10 20 30 40 50 Headcount Poverty Ratio (%) Source: Author's estimates on 2004 data from HBS. Self-employment decreases dramatically the poverty risk in urban areas, while increases the risk in rural areas. Table B3.9 shows a wide gap in the headcount ratios between urban and rural areas (2 percent versus 18 percent, respectively). The unemployed is a relatively small group (3 percent of households are headed by an unemployed person), but they face a considerably higher risk of poverty as compared to the national average, both in rural and urban areas (28 and 26 percent, respectively). Finally, with regard to other inactive persons, Figure 3.17 shows that there are large disparities in the incidence of poverty between urban and rural areas: the group of other inactive accounts for 3.5 percent of the total population, and 9 percent of the poor. Figure 3.17 ­ Poverty Shares By Employment Status of the Head of Household 100 7% 9% 80 14% Share of the poor (%) 23% 60 46% 40 Employment status retired self-employed 20 employee other inactive unemployed 0 Source: Author's estimates on 2004 data from HBS. 60 THE DETERMINANTS OF POVERTY The analysis contained in the previous section relies on simple bivariate associations between poverty rates and a pool of factors potentially responsible for the observed pattern of poverty across population groups. The main limitation of the above analysis is that simple correlations can be spurious, that is, driven by factors omitted from bivariate comparisons. To illustrate this, suppose we are interested in the relationship between the incidence of poverty and the region of residence of households. We saw in the previous section that poverty incidence varies significantly across regions: the headcount ratio in the Central region (18 percent) is 6 times higher than in the Zagreb area (3 percent), poverty incidence in the Eastern region (19 percent) is more than twice as much as in the Adriatic Southern region (9 percent). The relationship between poverty and region, however, may not be direct (regions matter because of differences in hydro- oro-graphic conditions, lack of infrastructure, poor access to basic services, etc.), but may be caused by a third variable such as, say, education. To the extent that educational attainment is unevenly distributed across regions, the relationship between poverty risk and region can be dubbed spurious: poverty risk is related to region indirectly, via education. One way of identifying the nature of the relationship between poverty risk and regions is by purging the effect of education from the simple correlation between poverty and region. This can be achieved by using partial correlations instead of simple correlations. Partial correlation between two variables (x and y, say) is defined as the correlation observed after holding constant (that is, eliminating the effects of) a third variable (say z). Partial correlations may differ substantially from simple correlations, and comparisons are often informative about the relationship between the two variables. Resuming the above example, suppose we control for z: if there is little difference between the simple and partial correlations of x and y, it can be inferred that the relationship between x and y is "genuine", or at least not driven by z. If, on the other hand, partial correlation differs significantly from simple correlation, the simple correlation is spurious (at least to some extent). This section investigates whether the relationships between poverty rates and a selection of poverty covariates are genuine rather than spurious. For this purpose we have carried out a partial correlation analysis by means of micro- simulations. Section 4.1 describes the method used to run micro-simulations, and section 4.2 summarizes the main findings. The Method The method outlined in this section permits the estimation of relative poverty risk after controlling for one or more variables. We shall refer to this poverty risk as the simulated relative poverty risk. The idea can be illustrated as follows. We control the relationship between poverty and region for education by means of a three-step procedure: (i) We regress (per adult equivalent) consumption on a set of household characteristics and poverty covariates (ii) We generate predicted consumption levels that would be observed after assigning the same educational level to all individuals in the sample (iii) We calculate the relative poverty risk using the consumption levels simulated/predicted under step (ii). 61 In step (i) estimates of the partial effects for each covariate in the consumption function are obtained; in step (ii) we control for (eliminate) the variation of educational levels (both within and between regions), so as to simulate household consumption levels that would be observed in the absence of differences in educational levels. Finally, in step (iii) we estimate the counterfactual/simulated relative poverty risk on the basis of the predicted values from step (ii). Once we have simulated the relative poverty risk, we compare the simulated versus actual patterns of risk. If the comparison shows little difference, we conclude that education is not responsible for the regional variation in poverty. Hence, the correlation between poverty and region is not spurious. If after controlling for education the poverty risk pattern changes significantly, we conclude that the correlation between poverty and region is spurious (that is, driven by the uneven distribution of education across regions). In practice, simulated relative poverty risk is obtained by adapting Luttmer's (2000) procedure: 1) Specify a household consumption function model: yh = x h + , where yh is household expenditure (per equivalent adult), xh denotes a vector of household characteristics, is a vector of unknown parameters, and is an error term with zero mean and variance 2. The consumption function specified in this study is described in the Appendix. 2) Estimate the consumption function model by ordinary least squares. Regression results are shown in the Appendix. 3) Use regression estimates to calculate simulated equivalent expenditures. These are defined as follows: ^ y SIM ,h = y h + edu ( x edu - x edu ,h ) , where xedu,h denotes the actual educational level of the head of the h-th household, and xedu is the average educational level (the average being calculated over the whole sample). Luttmer (2000: 8) contains an excellent interpretation of the above equation. 4) Obtain the simulated relative poverty risks, i.e. the relative poverty risks that would be observed if all households in the population were given the average educational level. 5) Bootstrap standard errors associated with the simulated relative poverty risks (or estimate them through linearized variance estimators). The above procedure refers to the case in which education is controlled for. In the next section we extend our analysis by controlling for other factors, like: employment status of the head of household, the region of residence of the household, head of household's age, and the household size. Main Findings Simulation results, shown in Tables 3.4-3.7. Table 3.7 can be summarized as follows. Regional variation of poverty cannot be accounted for by differences in the distribution of education, labor market status and other demographic factors. Table 3.4 shows that the pattern of simulated poverty risks does not change significantly when compared to the pattern of actual poverty risks: this holds true for all controls (columns in Table 3.4) with the exception of the region itself (column 3). After controlling for regional effects, Table 3.4 shows that poverty risks converge towards unity. Table 62 3.4 leads to the same conclusion, even if seen from a different point of view. The stratification by urban/rural (Table B3.11 in the Appendix) does not alter the pattern found at the national level. Figure 3.18: Standardized Simulated Relative Poverty Risks by Region 100 actual Region Education deviation from national average (%) Employment status 50 Relative poverty risk Age Household size 0 -50 -100 zagreb adriatic adriatic central eastern north south Note: The figure shows the patterns of percentage deviation of the actual and simulated relative poverty risk from the national average. If the pattern simulated for factor j (say education) remains close to the actual pattern (thick solid line), we infer that factor j plays an insignificant role in the explanation of the correlation between poverty and region. If, on the other hand, the simulated pattern flattens towards the zero horizontal axis, we infer that factor j plays a significant role in explaining the relationship between poverty and educational attainment. In other words, the correlation between poverty and region is mediated by factor j. In the above figure, only the curve simulated for the factor region (dashed line) flattens significantly, which suggests that the relationship between poverty and region is not spurious. Source: Author's estimates on 2004 data from HBS. Education has a major effect on decreasing the risk of poverty, but does not affect the regional variation of relative poverty risks. After controlling for education (column 1, Table 3.4) the overall headcount ratio decreases from 11 percent (actual) to 9 percent (simulated), while the odds ratios of simulated poverty risks hardly change.22 Education is a powerful and independent micro-determinant of poverty. Table 3.5 shows that employment status, region, age and household size does not explain the relationship between poverty and education: the simulated patterns of the relative poverty risks in columns 2 to 5 are almost identical to the actual pattern in column 0. In contrast, by controlling for education (columns 1) we observe a clear-cut convergence towards a uniform distribution. Figure 3.19 illustrates. 22 This is in line with the regression estimates of the consumption function in the Appendix: the parameters beta associated with the education variables are "large" and statistically significant. 63 Figure 3.19: Standardized Simulated Relative Poverty Risks by Education of Household Head 200 actual education employment status deviation from national average (%) region age Relative poverty risk 100 household size 0 -100 unfinished primary vocational general post primary secondary secondary secondary Source: Author's estimates on 2004 data from HBS. Age affects poverty risk significantly. The evidence in Table 3.6 is mixed. By comparing actual versus simulated relative poverty risks for members of households headed by young individuals (that is aged 16 to 30 years) we are led to conclude that the link between age and poverty is direct. Table B3.12 in the Appendix qualifies this result, and suggests that this link is especially strong in rural areas. On the other hand, if we carry out the same comparison for households headed by elderly individuals, we conclude that the link between poverty and age is mediated by education. After controlling for the partial effect of education the risk of poverty for households headed by individuals aged 65 or above, decreases from 93 percent above the average (actual relative poverty risk) to 60 percent (simulated poverty risk). The link between unemployment and poverty is not accounted for by other socio-demographic factors. Households headed by unemployed face the second-highest risk of poverty (about two and a half times the average risk), which does not change significantly after controlling for factors in columns 1, 3 4 and 5 in Table 3.7. It does, however, change dramatically when controlled for the employment status itself; according to the estimate in column 2, poverty risk decreases to 30 percent below the average relative risk. There is a strong link between poverty and membership in a rural household headed by an inactive individual. The result shown in Table 3.7 (compare columns 0 and 2) points to direct association between poverty and inactive status. However, the estimates in Table B3.15 (Appendix) suggest that the entire effect is driven by rural households, whereas in urban areas controlling for employment status does not imply a large discrepancy between actual and simulated relative poverty risks. Household size and poverty risk are linked indirectly, their relationship mediated by a combination of other socio-demographic factors. This conclusion is suggested by the evidence in Table 3.8, which shows a stable pattern of relative poverty risk by household size, regardless of the variables controlled for. 64 Table 3.4: Relative Poverty Risk by Region (0) SIMULATED ACTUAL Relative Poverty Risk after controlling for the partial effect of: Relative (1) (2) (3) (4) (5) (6) Region Poverty Household All Risk Education Employment status Region Age size factors Central 1.65 1.71 1.65 1.29 1.68 1.65 1.10 (0.20) (0.22) (0.20) (0.20) (0.20) (0.20) (0.22) Eastern 1.69 1.61 1.67 1.29 1.71 1.67 1.07 (0.23) (0.24) (0.23) (0.18) (0.23) (0.23) (0.25) Zagreb 0.30 0.22 0.35 0.61 0.25 0.33 0.86 (0.07) (0.06) (0.07) (0.10) (0.06) (0.07) (0.17) Adriatic North 0.41 0.46 0.34 0.65 0.32 0.37 0.51 (0.14) (0.15) (0.11) (0.19) (0.12) (0.12) (0.20) Adriatic South 0.79 0.85 0.79 1.07 0.85 0.78 1.31 (0.16) (0.17) (0.15) (0.19) (0.16) (0.15) (0.23) Overall poverty rate (%) 11.1 9.1 11.6 10.0 11.0 10.8 6.6 (0.86) (0.8) (0.9) (0.80) (0.9) (0.83) (0.65) Notes: Standard errors are in parentheses, corrected for sample design effects. 65 Table 3.5­ Relative Poverty Risk by Educational Attainment of the Head of the Household (0) SIMULATED Relative Poverty Risk after controlling for the partial effect of: ACTUAL (1) (2) (3) (4) (5) (6) Educational Attainment of Relative Poverty Household Head Risk Employment Household All Education Region Age status size factors Unfinished Primary 2.85 1.80 2.83 2.77 2.53 2.98 1.17 (0.24) (0.25) (0.23) (0.26) (0.24) (0.25) (0.24) Primary 1.87 1.57 1.96 1.76 1.87 1.90 1.16 (0.20) (0.21) (0.19) (0.20) (0.19) (0.20) (0.21) Vocational secondary 0.62 0.90 0.56 0.69 0.73 0.56 1.02 (0.10) (0.13) (0.09) (0.10) (0.11) (0.09) (0.16) General Secondary 0.24 0.49 0.23 0.27 0.30 0.22 0.73 (0.07) (0.11) (0.06) (0.08) (0.08) (0.07) (0.16) Post Secondary 0.02 0.45 0.06 0.06 0.02 0.02 0.97 (0.01) (0.11) (0.04) (0.04) (0.01) (0.01) (0.19) Overall poverty rate (%) 11.1 9.1 11.6 10.0 11.0 10.8 6.6 (0.86) (0.8) (0.8) (0.80) (0.9) (0.83) (0.65) Notes: Standard errors are in parentheses, corrected for sample design effects. 66 Table 3.6­ Relative Poverty Risk by Age of the Head of the Household (0) SIMULATED Relative Poverty Risk after controlling for the partial effect of: ACTUAL (1) (2) (3) (4) (5) (6) Age of the Head of the Relative Poverty Household Risk Employment Household All Education Region Age status size factors 16 ­ 30 0.27 0.32 0.16 0.18 0.68 0.28 0.86 (0.17) (0.22) (0.13) (0.15) (0.25) (0.17) (0.38) 31 ­ 49 0.61 0.75 0.58 0.68 0.72 0.56 0.84 (0.09) (0.11) (0.08) (0.10) (0.10) (0.08) (0.13) 50 ­ 64 0.82 0.91 0.73 0.76 0.85 0.79 1.16 (0.11) (0.14) (0.11) (0.12) (0.11) (0.11) (0.17) 65 or above 1.93 1.60 2.13 1.93 1.67 2.04 1.05 (0.16) (0.16) (0.16) (0.17) (0.14) (0.16) (0.15) Overall poverty rate (%) 11.1 9.1 11.6 10.0 11.0 10.8 6.6 (0.86) (0.8) (0.8) (0.80) (0.9) (0.83) (0.65) Notes: Standard errors are in parentheses, corrected for sample design effects. 67 Table 3.7: Relative Poverty Risk by Employment Status of the Head of the Household (0) SIMULATED Relative Poverty Risk after controlling for the partial effect of: ACTUAL (1) (2) (3) (4) (5) (6) Employment Status of Relative Poverty Household Head Risk Employment Household All Education Region Age status size factors Employee 0.34 0.47 0.26 0.29 0.41 0.30 0.58 (0.07) (0.09) (0.06) (0.07) (0.08) (0.07) (0.11) Self-employed 1.20 1.31 1.42 1.28 1.27 1.28 1.44 (0.19) (0.22) (0.19) (0.20) (0.20) (0.19) (0.25) Unemployed 2.41 2.35 0.71 2.64 2.48 2.22 1.29 (0.46) (0.51) (0.28) (0.49) (0.46) (0.46) (0.44) Retired 1.38 1.19 1.59 1.39 1.25 1.40 1.14 (0.12) (0.12) (0.12) (0.12) (0.11) (0.12) (0.14) Other inactive 2.71 2.34 1.76 2.52 2.62 2.64 1.90 (0.41) (0.46) (0.35) (0.43) (0.41) (0.41) (0.52) Overall poverty rate (%) 11.1 9.1 11.6 10.0 11.0 10.8 6.6 (0.86) (0.8) (0.8) (0.80) (0.9) (0.83) (0.65) Notes: Standard errors are in parentheses, corrected for sample design effects. 68 Table 3.8: Relative Poverty Risk by Household Size (0) SIMULATED Relative Poverty Risk after controlling for the partial effect of: ACTUAL (1) (2) (3) (4) (5) (6) Relative Poverty Household size Risk Employment Household All Education Region Age status size factors 1 2.13 1.71 2.13 2.14 1.95 2.37 1.32 (0.18) (0.19) (0.19) (0.20) (0.18) (0.20) (0.20) 2 1.43 1.15 1.46 1.40 1.28 1.56 1.20 (0.14) (0.13) (0.14) (0.14) (0.13) (0.15) (0.17) 3 0.74 0.81 0.74 0.71 0.78 0.70 0.77 (0.13) (0.14) (0.13) (0.13) (0.14) (0.13) (0.18) 4 0.41 0.49 0.36 0.38 0.44 0.37 0.74 (0.08) (0.09) (0.07) (0.08) (0.08) (0.08) (0.14) 5 0.80 1.12 0.88 .90 1.00 0.70 1.01 (0.18) (0.25) (0.18) (0.20) (0.20) (0.17) (0.23) 6+ 1.37 1.52 1.36 1.42 1.43 1.23 1.35 (0.28) (0.32) (0.27) (0.31) (0.29) (0.27) (0.37) Overall poverty rate (%) 11.1 9.1 11.6 10.0 11.0 10.8 6.6 (0.86) (0.8) (0.9) (0.80) (0.9) (0.83) (0.65) Notes: Standard errors are in parentheses, corrected for sample design effects. 69 SUMMARY AND CONCLUSIONS The poverty analysis carried out in this paper leads to several conclusions. In this section we offer a summary of the main findings and wrap up the analysis with a few final, policy-oriented remarks. The poverty profile constructed in sections 2 and 3 identifies the following factors: ˇ Type of settlement. Poverty is deeper, more severe and widespread in rural areas than in urban areas. ˇ Region. Poverty is concentrated in the Central and Eastern regions. Regional variation of poverty incidence ranges from about 3 percent in the Zagreb region to 18-19 percent among households in the Eastern and Central regions. Even more pronounced is the variation among poverty gaps and the squared poverty gaps. ˇ Education. The risk of poverty decreases sharply with the level of educational attainment of the head of household. A head of household with primary or lower education is associated with a poverty risk which is two times the average risk, while attainment of secondary education reduces the risk to one-third of the average risk. ˇ Activity rates. Low activity rates are clearly mirrored in the structure of poverty rates. The single most important group is the pensioners: Apart from being associated with poverty risk twice the average, they account for 46 percent of the total poor. Households headed by unemployed and other inactive persons are also subject to an above-average poverty risk, but together they make up 16 percent of the poor. ˇ Old age and pension receipts. Poverty rates increase over the life cycle of the head of household. While cohorts below 64 years of age have below-average risk of poverty, households headed by 65+ persons have a poverty risk which is two times the national average. Within the 65+ group, those without pensions are at risk more than five times the national average. The largest fraction of the elderly classified as poor is concentrated in rural areas. As far as the micro-determinants of poverty are concerned, the analysis in section 4 has identified two main factors that stand out in terms of explanatory power for the household poverty status: ˇ Region. The regional disparities in poverty rates across regions persist after controlling for differences in education, labor market and other demographic factors. To the extent that regional differences are structural, poverty is more likely to be an enduring rather than a transitory phenomenon. ˇ Education. Croatia is no exception to the common finding that human capital improves standards of living. The poverty risk literally collapses when calculated over population groups with relatively high educational attainment levels. Simulation results confirm that the link between poverty and education is strong and genuine. The large disparities in poverty rates across regions are associated with similarly large disparities in both unemployment rates and average earnings. This leads us to the introduction of two issues: (i) increase investment in human capital, and (ii) design policies to enhance labor mobility in Croatia. The combination of a thin housing market and a social safety net inadequate in supporting families who provide in ­home care for the elderly (especially in rural areas) is likely to negatively affect the willingness of workers to move across regions. In other words, the benefits from migration may be more than offset by opportunity costs. 70 References Deaton, A. and S. Zaidi (2002), "Guidelines for Constructing Consumption Aggregates for Welfare Analysis", Living Standards Measurement Study Working Paper Nr. 135. World Bank: Washington D.C. FAO (2004), Human Energy Requirements; Report of a Joint FAO/WHO/UNU Expert Consultation. FAO Food and Nutrition Technical Report Series No.1. Food and Agriculture Organization: Rome. Luttmer E. 2000. "Methodology." Background Paper No. 2 in World Bank (2000). Ravallion, M. (1994), Poverty Comparisons. Harwood Academic Press: Chur, Switzerland. World Bank (2000), Croatia: Economic Vulnerability and Welfare Study, Volume II: Technical Papers. World Bank: Washington D.C. World Bank (2005), Growth, Poverty an Inequality: Eastern Europe and the Former Soviet Union. World Bank: Washington D.C. 71 APPENDICES APPENDIX A: The Estimated Consumption Function The model is the following: ln( EQCONSi ) = 0 + 1 DEMO i + 2 LOCi + 3 LFPi + 4SOCIOi + 5 EDUC i + i where: ln(EQCONSi) log of total per adult equivalent expenditure (HKN/year/adult equivalent) DEMOGRAPHIC VARIABLES (DEMO) MALEi = 1 if head of the household is male AGE age of the head of the household HSIZEi five variables controlling for household size: hs1 = 1 if household size equals 1 member hs2 = 1 if household size = 2 members hs3 = 1 if household size = 3 members hs4 = 1 if household size = 4 members hs5 = 1 if household size = 5 members hs6 = 1 if household size = 6 members DEPRATIOi two control variables: (i) DEPKID = (number of children aged 5-14 as a fraction of household size), and (ii) DEPELD = (number of elderly as a fraction of household size). LOCATION (LOC) REGIONi set of regional dummy variables: loc1 = 1 if Zagreb Urban loc2 = 1 if Central Urban loc3 = 1 if North Adriatic Urban loc4 = 1 if South Adriatic Urban loc5 = 1 if Central Urban loc6 = 1 if Zagreb Rural loc7 = 1 if Central Rural loc8 = 1 if North Adriatic Rural loc9 = 1 if South Adriatic Rural loc10 = 1 if Eastern Rural LABOR FORCE PARTICIPATION (LFP) NEARNERSi = (# of labor income earners in last 12 months)/(# of adults) LMSi five dummies controlling for the labor market status of the head of the household (in bold type is the reference group): employee = 1 if employee selfempl = 1 if selfemployed retired = 1 if retired unemployed = 1 if unemployed othinact = 1 if not in the labor force A set of dummies was defined analogously for the spouse and included in the regression. SOCIOLOGICAL CONTROLS (SOCIO) MARRIEDi = 1 if married EDUCATION (EDUC) EDUCi set of binary variables for the head of the household education (highest level of education attained): ownedu1 = 1 if unfinished primary wnedu2 = 1 if primary ownedu3 = 1 if vocational secondary ownedu4 = 1 if general secondary ownedu5 = 1 if post secondary A set of dummies was defined analogously for the spouse and included in the regression. j is the error term 72 Regression Results Dependent variable: log of consumption (HRK/year/adult equivalent) Coef. Std.Err. male 0.003 (0.023) age -0.003 (0.001)** Household size (reference group: household with 1 member) 2 members -0.024 (0.028) 3 members -0.063 (0.033) 4 members -0.088 (0.035)* 5 members -0.105 (0.039)** 6 or more members -0.148 (0.046)** Dependency ratios (number of kids)/ household size -0.152 (0.049)** (number of elderly)/ household size -0.146 (0.031)** Location (reference group: central urban) zagreb urban 0.319 (0.029)** north adriatic urban 0.191 (0.033)** south adriatic urban 0.226 (0.033)** eastern urban 0.087 (0.033)** zagreb rural 0.199 (0.042)** central rural -0.019 (0.033) north adriatic rural 0.277 (0.038)** south adriatic rural 0.163 (0.034)** eastern rural -0.034 (0.037) Labor force participation (no. of labor income earners)/(no. adults) 0.304 (0.034)** Head of the household (reference group: employee) self-employed 0.112 (0.024)** unemployed -0.253 (0.049)** retired 0.106 (0.026)** inactive -0.173 (0.052)** Spouse (reference group: employee) self-employed 0.116 (0.030)** unemployed -0.107 (0.037)** retired 0.049 (0.029) inactive -0.015 (0.026) Education (reference group: unfinished primary) Head of the household Primary 0.110 (0.030)** Vocational Secondary 0.282 (0.030)** General Secondary 0.364 (0.032)** Post Secondary 0.516 (0.034)** Spouse Primary 0.061 (0.033) Vocational Secondary 0.070 (0.033)* General Secondary 0.126 (0.033)** Post Secondary 0.218 (0.040)** Sociological variables married -0.082 (0.037)* constant 10.219 (0.072)** No. observations 2,847 R-squared 0.47761 Note: Standard errors in parentheses. * p < 0.05, ** p < 0.01. 73 APPENDIX B: STATISTICAL APPENDIX Table B3.1: Headcount Poverty Ratios by Region Urban Rural Overall Fraction Fraction Fraction Relative Relative Relative of total Headcount of total Headcount of total Headcount Region median median median population (%) population (%) population (%) consumption consumption consumption (%) (%) (%) Central 7.74 0.90 13.80 15.48 0.80 20.54 23.23 0.84 18.29 (1.24) (0.03) (3.09) (1.31) (0.02) (2.96) (0.63) (0.02) (2.16) Eastern 8.16 0.96 10.89 11.91 0.75 23.99 20.07 0.84 18.66 (1.18) (0.03) (2.91) (1.39) (0.02) (3.59) (0.74) (0.02) (2.56) Zagreb 18.95 1.33 2.56 5.59 0.92 5.83 24.54 1.25 3.31 (1.17) (0.03) (0.74) (1.23) (0.05) (1.94) (0.94) (0.02) (0.73) Adriatic North 8.15 1.16 3.18 4.60 1.06 6.99 12.74 1.11 4.56 (0.92) (0.04) (1.71) (0.98) (0.05) (2.86) (0.74) (0.03) (1.52) Adriatic South 9.76 1.11 3.30 9.66 0.93 14.15 19.41 1.04 8.70 (1.24) (0.03) (1.10) (1.41) (0.04) (3.02) (0.71) (0.02) (1.75) CROATIA 52.76 1.15 5.73 47.24 0.84 17.04 100.00 1.00 11.08 (2.59) (0.02) (0.82) (2.59) (0.01) (1.58) - - (0.86) Notes: Standard errors are in parentheses, corrected for sample design effects. Relative median consumption is the median equivalent consumption of the population subgroup divided by the national median equivalent consumption. Source: Data are from the 2004 Croatian Household Budget Survey. 74 Table B3.2: Headcount Poverty Ratios by Age of the Head of the Household Urban Rural Overall Fraction Fraction Fraction Relative Relative Relative of total Headcount of total Headcount of total Headcount Age median median median population (%) population (%) population (%) consumption consumption consumption (%) (%) (%) 16-30 2.63 1.58 0.00 1.15 0.95 9.92 3.77 1.39 3.01 (0.34) (0.08) (0.00) (0.22) (0.09) (5.68) (0.39) (0.09) (1.84) 31-49 21.35 1.26 3.03 17.03 0.98 11.41 33.38 1.13 6.75 (1.33) (0.02) (0.80) (1.18) (0.03) (1.84) (1.05) (0.02) (0.96) 50-64 16.77 1.20 5.41 15.86 0.94 13.00 32.63 1.08 9.10 (1.16) (0.02) (1.21) (1.09) (0.03) (2.30) (1.02) (0.02) (1.25) 65+ 12.01 0.94 12.24 13.20 0.69 29.78 25.21 0.80 21.43 (0.89) (0.02) (2.06) (0.94) (0.03) (2.74) (0.94) (0.01) (1.75) CROATIA 52.76 1.15 5.73 47.24 0.84 17.04 100.00 1 11.08 (2.59) (0.02) (0.82) (2.59) (0.01) (1.58) - - (086) Notes: Standard errors are in parentheses, corrected for sample design effects. Relative median consumption is the median equivalent consumption of the population subgroup divided by the national median equivalent consumption. Source: Data are from the 2004 Croatian Household Budget Survey. 75 Table B3.3: Headcount Poverty Ratios by Household Size Urban Rural Overall Fraction Fraction Fraction Relative Relative Relative Household of total Headcount of total Headcount of total Headcount median median median Size population (%) population (%) population (%) consumption consumption consumption (%) (%) (%) 1 4.98 1.04 11.19 4.08 0.69 38.59 9.07 0.87 23.53 (0.37) (0.03) (2.13) (0.35) (0.04) (3.38) (0.42) (0.02) (2.02) 2 11.45 1.08 10.69 8.84 0.83 22.56 20.29 0.96 15.86 (0.76) (0.03) (1.95) (0.64) (0.03) (2.56) (0.75) (0.02) (1.54) 3 11.35 1.25 3.26 7.68 0.91 15.42 19.03 1.10 8.17 (0.83) (0.04) (1.14) (0.69) (0.04) (3.05) (0.83) (0.02) (1.48) 4 15.58 1.25 0.59 11.03 1.01 10.12 26.61 1.16 4.55 (1.08) (0.02) (0.41) (0.86) (0.05) (1.94) (0.97) (0.02) (0.87) 5 6.07 1.15 6.08 7.30 0.86 11.27 13.37 0.98 8.91 (0.62) (0.05) (2.36) (0.70) (0.03) (3.00) (0.76) (0.03) (2.00) 6+ 3.32 0.92 12.33 8.32 0.82 16.35 11.64 0.84 15.20 (0.50) (0.04) (5.94) (0.87) (0.03) (3.72) (0.90) (0.02) (3.12) CROATIA 52.76 1.15 5.73 47.24 0.84 17.04 100.00 1 11.08 (2.59) (0.02) (0.82) (2.59) (0.01) (1.58) - - (086) Notes: Standard errors are in parentheses, corrected for sample design effects. Relative median consumption is the median equivalent consumption of the population subgroup divided by the national median equivalent consumption. Source: Data are from the 2004 Croatian Household Budget Survey. 76 Table B3.4: Headcount Poverty Ratios by Educational Attainment of the Head of the Household Urban Rural Overall Fraction Fraction Fraction Relative Relative Relative of total Headcount of total Headcount of total Headcount Educational Attainment median median median population (%) population (%) population (%) consumption consumption consumption (%) (%) (%) Unfinished Primary 3.65 0.72 31.23 10.04 0.67 31.74 13.69 0.69 31.60 (0.47) (0.03) (5.22) (0.86) (0.03) (3.25) (0.86) (0.02) (2.73) Primary 6.55 0.91 12.01 13.26 0.73 25.02 19.81 0.78 20.72 (0.65) (0.02) (2.71) (1.02) (0.03) (2.93) (1.00) (0.02) (2.17) Vocational Secondary 14.89 1.07 5.48 15.25 0.97 8.32 30.14 1.03 6.92 (1.12) (0.02) (1.39) (1.12) (0.03) (1.66) (1.07) (0.02) (1.08) General Secondary 14.27 1.25 1.87 5.50 1.11 4.79 19.77 1.22 2.28 (0.96) (0.02) (0.66) (0.62) (0.05) (2.14) (0.92) (0.02) (0.76) Post Secondary 13.40 1.51 0.11 3.19 1.50 0.50 16.59 1.51 0.18 (0.99) (0.03) (0.10) (0.42) (0.09) (0.50) (0.96) (0.03) (0.13) Overall 52.76 1.15 5.73 47.24 0.84 17.04 100 1 11.08 (2.59) (0.02) (0.82) (2.59) (0.01) (1.58) - - (086) 77 Table B3.5: Headcount Poverty Ratios by Employment Status of the Head of the Household Urban Rural Overall Fraction Fraction Fraction Relative Relative Relative Employment of total Headcount of total Headcount of total Headcount median median median Status population (%) population (%) population (%) consumption consumption consumption (%) (%) (%) Employee 24.50 1.31 2.44 16.00 1.05 5.76 40.51 1.21 3.75 (1.47) (0.02) (0.85) (1.16) (0.03) (1.50) (1.18) (0.02) (0.77) Self-employed 6.04 1.33 2.03 13.31 0.82 18.39 19.35 0.95 13.29 (0.61) (0.05) (1.18) (1.14) (0.02) (2.90) (1.09) (0.04) (2.06) Unemployed 1.46 0.73 25.56 1.54 0.75 27.84 3.00 0.75 26.73 (0.24) (0.05) (6.58) (0.28) (0.09) (7.67) (0.35) (0.05) (5.08) Retired 19.13 0.99 9.21 14.56 0.75 23.26 33.69 0.89 15.28 (1.20) (0.02) (1.43) (1.11) (0.02) (2.37) (1.07) (0.01) (1.31) Other inactive 1.63 1.17 10.40 1.83 0.51 47.39 3.46 0.75 29.96 (0.25) (0.09) (4.43) (0.27) (0.06) (6.40) (0.35) (0.06) (4.57) Overall 52.76 1.15 5.73 47.24 0.84 17.04 100 1 11.08 (2.59) (0.02) (0.82) (2.59) (0.01) (1.58) - - (0.86) 78 Table B3.6: Poverty Gap and Poverty Gap Squared Indices by Region Region Urban Rural Overall Poverty Poverty Gap Poverty Poverty Gap Poverty Poverty Gap Mean cons. of Mean cons. of Mean cons. of Gap Squared Gap Squared Gap Squared the poor the poor the poor (%) (%) (%) (%) (%) (%) Central 38,984 3.5 1.4 32,702 6.1 2.6 34,936 5.2 2.2 (1,857) (1.2) (0.6) (1,017) (1.1) (0.6) (966) (0.8) (0.4) Eastern 39,813 2.1 0.6 30,961 5.9 2.1 34.834 4.3 1.5 (1,594) (0.6) (0.2) (1,391) (1.1) (0.5) (1,188) (0.7) (0.3) Zagreb 57,241 0.5 0.2 40,694 0.8 0.2 53,818 0.6 0.2 (1,794) (0.2) (0.1) (2,773) (0.2) (0.1) (1,627) (0.2) (0.1) Adriatic North 47,964 0.6 0.2 42,772 0.6 0.1 46,240 0.6 0.2 (2,035) (0.3) (0.1) (2,109) (0.3) (0.0) (1,573) (0.2) (0.1) Adriatic South 47,030 0.6 0.1 39,087 2.9 0.9 43,088 1.7 0.5 (1,427) (0.2) (0.0) (1,783) (0.7) (0.2) (1,310) (0.4) (0.1) CROATIA 48,583 1.2 0.4 35,560 4.2 1.6 42,761 2.6 1.0 (915) (0.2) (0.1) (773) (0.5) (0.2) (622) (0.3) (0.1) 79 Table B3.7: Poverty Rates by Age Urban Rural Overall Poverty Gap Poverty Gap Poverty Gap Age Headcount Poverty Gap Headcount Poverty Gap Headcount Poverty Gap Squared Squared Squared 0-15 4.8 1.1 0.4 12.4 2.5 0.8 8.6 1.8 0.6 (1.7) (0.4) (0.2) (2.0) (0.6) (0.3) (1.3) (0.3) (0.2) 16-30 4.0 0.9 0.3 12.6 3.0 1.1 7.7 1.8 0.6 (0.9) (0.3) (0.1) (1.9) (0.6) (0.3) (1.0) (0.3) (0.1) 31-49 3.3 0.8 0.3 12.4 2.9 1.0 7.4 1.8 0.6 (0.7) (0.2) (0.1) (1.6) (0.5) (0.2) (0.8) (0.2) (0.1) 50-64 5.4 1.0 0.3 13.5 3.1 1.1 9.3 2.0 0.7 (1.1) (0.2) (0.1) (1.9) (0.5) (0.2) (1.0) (0.3) (0.1) 65+ pensioner 10.8 1.9 0.5 29.0 7.3 2.7 19.5 4.5 1.6 (1.6) (0.3) (0.1) (2.8) (0.9) (0.5) (1.7) (0.5) (0.2) 65+ w/o pension 23.8 8.0 3.5 46.4 15.1 7.1 39.3 12.9 6.0 (4.9) (1.9) (1.1) (4.6) (2.1) (1.3) (3.6) (1.6) (1.0) CROATIA 5.7 1.2 0.4 17.0 4.2 1.6 11.1 2.6 1.0 (0.8) (0.2) (0.1) (1.5) (0.5) (0.2) (0.9) - (0.86) Notes: Standard errors are in parentheses, corrected for the sample design. Source: Author's estimates on HBS 2004. 80 Table B3-8: Poverty Rates by Age and Gender Urban Rural Croatia Poverty Gap Poverty Gap Poverty Gap Headcount Poverty Gap Headcount Poverty Gap Headcount Poverty Gap Squared Squared Squared FEMALE 0-15 5.3 1.5 0.6 11.4 2.6 1.1 8.2 2.0 0.8 (2.2) (0.6) (0.3) (2.3) (0.8) (0.5) (1.6) (0.5) (0.3) 16-30 3.4 0.7 0.2 12.4 3.0 1.1 7.2 1.7 0.6 (1.1) (0.2) (0.1) (2.2) (0.7) (0.3) (1.1) (0.3) (0.1) 31-49 3.3 0.7 0.3 11.4 2.6 0.8 6.7 1.5 0.5 (0.8) (0.2) (0.1) (1.6) (0.4) (0.2) (0.8) (0.2) (0.1) 50-64 6.0 1.1 0.3 15.1 3.7 1.3 10.2 2.3 0.8 (1.3) (0.3) (0.1) (2.1) (0.6) (0.3) (1.2) (0.3) (0.1) 65+ pensioner 12.0 2.0 0.5 30.1 7.0 2.4 20.3 4.3 1.4 (2.0) (0.4) (0.1) (3.4) (1.0) (0.4) (2.0) (0.5) (0.2) 65+ w/o pension 22.0 7.3 3.2 47.0 15.4 7.1 38.8 12.7 5.8 (4.8) (1.9) (1.1) (4.5) (2.0) (1.2) (3.6) (0.2) (1.0) All Female 6.3 1.4 0.5 18.5 4.7 1.8 11.9 2.9 1.1 (0.9) (0.2) (0.1) (0.2) (0.6) (0.3) (0.9) (0.3) (0.1) MALE 0-15 4.4 0.7 0.2 13.3 2.4 0.6 9.0 1.6 0.4 (1.6) (0.3) (0.1) (2.4) (0.5) (0.2) (1.4) (0.3) (0.1) 16-30 4.6 1.1 0.4 12.8 3.0 1.1 8.3 1.9 0.7 (1.2) (0.4) (0.2) (2.0) (0.6) (0.3) (1.1) (0.3) (0.2) 31-49 3.3 0.8 0.3 13.3 3.2 1.2 8.2 2.0 0.7 (1.0) (0.3) (0.1) (1.8) (0.6) (0.3) (1.0) (0.3) (0.2) 50-64 4.8 0.9 0.3 11.8 2.5 0.8 8.2 1.7 0.6 (1.1) (0.3) (0.1) (2.2) (0.6) (0.2) (1.2) (0.3) (0.1) 65+ pensioner 9.3 1.8 0.5 27.7 7.6 3.0 18.4 4.7 1.8 (2.0) (0.4) (0.2) (3.1) (1.1) (0.7) (2.0) (0.6) (0.4) 65+ w/o pension 49.9 18.5 8.1 42.3 13.6 6.9 43.7 14.5 7.1 (20.5) (8.9) (4.6) (10.6) (5.1) (3.6) (9.5) (0.4) (0.3) All Male 5.1 1.1 0.4 15.6 3.7 1.4 10.2 2.4 0.9 (0.8) (0.2) (0.1) (0.2) (0.5) (0.2) (0.9) (0.3) (0.1) Notes: Standard errors are in parentheses, corrected for the sample design. Source: Author's estimates on HBS 2004. 81 Table B3.9: Poverty Rates by Employment Status Urban Rural Overall Poverty Gap Poverty Gap Poverty Gap Employment Status Headcount Poverty Gap Headcount Poverty Gap Headcount Poverty Gap Squared Squared Squared Employee 2.2 0.4 0.1 5.8 0.9 0.3 3.7 0.6 0.2 (0.6) (0.1) (0.1) (0.9) (0.2) (0.1) (0.5) (0.1) (0.0) Self-employed 2.2 0.6 0.2 18.6 4.8 1.8 14.5 3.8 1.4 (1.1) (0.3) (0.1) (3.0) (0.9) (0.4) (2.3) (0.7) (0.3) Unemployed 12.8 3.2 1.1 23.8 6.7 2.8 17.7 4.8 1.9 (3.0) (1.1) (0.5) (4.2) (1.6) (0.9) (2.4) (0.9) (0.5) Retired 8.9 1.6 0.5 25.2 6.2 2.3 15.6 3.5 1.2 (1.2) (0.3) (0.1) (2.4) (0.7) (0.3) (1.3) (0.4) (0.2) Other inactive 7.0 1.8 0.7 24.0 6.7 2.8 15.1 4.2 1.7 (1.3) (0.4) (0.2) (2.5) (0.9) (0.5) (1.4) (0.5) (0.3) Aged less than 15 4.8 1.1 0.4 12.4 2.5 0.8 8.6 1.8 0.6 (1.7) (0.4) (0.2) (2.0) (0.6) (0.3) (1.3) (0.3) (0.2) Overall 5.7 1.2 0.4 17.0 4.2 1.6 11.1 2.6 1.0 (0.8) (0.2) (0.1) (1.5) (0.5) (0.2) (0.9) (0.3) (0.1) 82 Table B3.10: Poverty Rates by Educational Attainment Urban Rural Overall Poverty Gap Poverty Gap Poverty Gap Headcount Poverty Gap Headcount Poverty Gap Headcount Poverty Gap Squared Squared Squared Unfinished Primary 24.5 5.5 1.8 34.1 9.7 4.0 31.3 8.5 3.4 (3.7) (1.2) (0.5) (2.6) (1.0) (0.6) (2.2) (0.8) (0.4) Primary 10.1 2.2 0.8 23.6 5.8 2.1 18.1 4.3 1.6 (1.7) (0.4) (0.2) (2.5) (0.8) (0.4) (1.7) (0.5) (0.2) Vocational Secondary 6.5 1.2 0.4 9.8 1.9 0.6 8.1 1.6 0.5 (1.3) (0.3) (0.1) (1.5) (0.4) (0.2) (1.0) (0.2) (0.1) General Secondary 1.9 0.4 0.1 4.2 0.9 0.4 2.6 0.6 0.2 (0.5) (0.1) (0.1) (1.0) (0.3) (0.2) (0.5) (0.1) (0.1) Post Secondary 0.2 0.0 0.0 1.0 0.4 0.2 0.4 0.1 0.0 (0.2) (0.0) (0.0) (0.8) (0.3) (0.1) (0.2) (0.1) (0.0) Aged less than 15 4.8 1.1 0.4 12.4 2.5 0.8 8.6 1.8 0.6 (1.7) (0.4) (0.2) (2.0) (0.6) (0.3) (1.3) (0.3) (0.2) Overall 5.7 1.2 0.4 17.0 4.2 1.6 11.1 2.6 1.0 (0.8) (0.2) (0.1) (1.5) (0.5) (0.2) (0.9) (0.3) (0.1) 83 Table B3-11: Relative Poverty Risks by Region and Urban/Rural Areas (0) SIMULATED Relative Poverty Risk after controlling for the partial effect of: ACTUAL (1) (2) (3) (4) (5) (6) Relative Poverty Region Risk Household All Education Employment status Region Age size factors URBAN Central 1.25 1.60 1.08 0.87 1.33 1.17 1.02 (0.28) (0.42) (0.29) (0.29) (0.32) (0.27) (0.31) Eastern 0.98 0.96 0.97 0.86 0.94 1.03 0.80 (0.26) (0.27) (0.26) (0.25) (0.28) (0.28) (0.23) Zagreb 0.23 0.18 0.23 0.50 0.20 0.25 0.92 (0.07) (0.07) (0.06) (0.09) (0.06) (0.07) (0.21) Adriatic North 0.29 0.61 0.24 0.41 0.25 0.26 0.51 (0.15) (0.21) (0.10) (0.18) (0.15) (0.11) (0.25) Adriatic South 0.30 0.49 0.25 0.59 0.37 0.33 1.09 (0.10) (0.15) (0.09) (0.15) (0.13) (0.10) (0.23) RURAL Central 1.85 1.77 1.93 1.50 1.85 1.89 1.14 (0.27) (0.30) (0.25) (0.27) (0.27) (0.28) (0.29) Eastern 2.17 2.06 2.15 1.58 2.24 2.11 1.25 (0.32) (0.35) (0.31) (0.27) (0.32) (0.33) (0.39) Zagreb 0.53 0.37 0.78 0.96 0.43 0.60 0.66 (0.18) (0.12) (0.22) (0.27) (0.15) (0.15) (0.26) Adriatic North 0.63 0.20 0.52 1.08 0.43 0.58 0.52 (0.26) (0.14) (0.25) (0.41) (0.20) (0.27) (0.33) Adriatic South 1.28 1.22 1.32 1.55 1.34 1.24 1.53 (0.27) (0.29) (0.25) (0.31) (0.27) (0.26) (0.37) Notes: Standard errors are in parentheses, corrected for sample design effects. Poverty risks for Croatia are in Table 3.4. 84 Table B3.12: Relative Poverty Risks by Household Size (0) SIMULATED Relative Poverty Risk after controlling for the partial effect of: ACTUAL (1) (2) (3) (4) (5) (6) Relative Poverty Household Size Risk Household All Education Employment status Region Age size factors Urban 1 1.01 0.82 0.05 1.09 0.89 1.16 0.82 (0.19) (0.19) (0.19) (0.21) (0.19) (0.20) (0.21) 2 0.97 0.83 0.88 1.07 0.84 1.04 1.19 (0.18) (0.16) (0.14) (0.17) (0.17) (0.18) (0.23) 3 0.29 0.58 0.25 0.37 0.36 0.27 0.72 (0.10) (0.15) (0.10) (0.11) (0.11) (0.10) (0.22) 4 0.05 0.25 0.08 0.16 0.10 0.06 0.73 (0.04) (0.09) (0.04) (0.07) (0.05) (0.04) (0.19) 5 0.55 0.78 0.32 0.58 0.55 0.35 0.83 (0.10) (0.28) (0.16) (0.23) (0.21) (0.17) (0.33) 6+ 1.11 1.39 1.06 1.45 1.31 1.14 1.30 (0.54) (0.59) (0.51) (0.62) (0.56) (0.55) (0.61) Rural 1 3.48 2.80 3.45 3.41 3.26 3.85 1.93 (0.30) (0.34) (0.31) (0.34) (0.30) (0.33) (0.35) 2 2.04 1.57 2.20 1.83 1.83 2.24 1.22 (0.23) (0.23) (0.24) (0.24) (0.21) (0.25) (0.24) 3 1.39 1.16 1.45 1.22 1.40 1.32 0.85 (0.28) (0.28) (0.26) (0.26) (0.28) (0.27) (0.30) 4 0.91 0.84 0.75 0.70 0.92 0.82 0.76 (0.17) (0.19) (0.16) (0.17) (0.18) (0.17) (0.23) 5 1.02 1.40 1.34 1.17 1.37 1.00 1.16 (0.27) (0.38) (0.28) (0.31) (0.30) (0.27) (0.33) 6+ 1.48 1.57 1.47 1.41 1.47 1.26 1.37 (0.34) (0.40) (0.32) (0.36) (0.34) (0.31) (0.46) Notes: Standard errors are in parentheses, corrected for sample design effects. 85 Table B3.12: Relative Poverty Risks by Age of the Head of the Household (0) SIMULATED Relative Poverty Risk after controlling for the partial effect of: ACTUAL (1) (2) (3) (4) (5) (6) Relative Poverty Age of Household Head Risk Household All Education Employment status Region Age size factors Urban 16 ­ 30 0.00 0.00 0.00 0.00 0.13 0.00 0.92 (0.00) (0.00) (0.00) (0.00) (0.13) (0.00) (0.46) 31 ­ 49 0.27 0.40 0.14 0.38 0.34 0.20 0.71 (0.07) (0.11) (0.05) (0.09) (0.09) (0.06) (0.14) 50 ­ 64 0.49 0.77 0.47 0.58 0.51 0.53 1.08 (0.11) (0.17) (0.11) (0.12) (0.11) (0.11) (0.18) 65 or above 1.11 1.01 1.18 1.22 0.94 1.19 0.91 (0.19) (0.18) (0.18) (0.20) (0.18) (0.19) (0.21) Rural 16 ­ 30 0.90 1.05 0.52 0.60 1.94 0.92 0.72 (0.51) (0.70) (0.41) (0.47) (0.72) (0.53) (0.69) 31 ­ 49 1.03 1.20 1.12 1.05 1.20 1.01 1.01 (0.17) (0.19) (0.16) (0.17) (0.17) (0.17) (0.23) 50 ­ 64 1.17 1.07 0.99 0.94 1.21 1.07 1.24 (0.21) (0.24) (0.20) (0.21) (0.21) (0.21) (0.29) 65 or above 2.69 2.13 2.99 2.57 2.33 2.81 1.18 (0.25) (0.24) (0.24) (0.25) (0.22) (0.26) (0.20) Notes: Standard errors are in parentheses, corrected for sample design effects. 86 Table B3:14: Relative Poverty Risks by Educational Attainment of the Head of the Household (0) SIMULATED Relative Poverty Risk after controlling for the partial effect of: ACTUAL (1) (2) (3) (4) (5) (6) Educational Attainment of Relative Poverty the Head of the Household Risk Household All Education Employment status Region Age size factors Urban Unfinished Primary 2.82 1.56 2.88 3.01 2.49 3.00 1.05 (0.47) (0.46) (0.46) (0.54) (0.48) (0.48) (0.36) Primary 1.08 0.78 0.96 1.11 1.04 1.15 0.64 (0.26) (0.24) (0.19) (0.24) (0.24) (0.25) (0.21) Vocational secondary 0.49 0.70 0.34 0.64 0.54 0.43 0.95 (0.13) (0.17) (0.10) (0.13) (0.14) (0.11) (0.21) General Secondary 0.17 0.40 0.15 0.27 0.23 0.17 0.73 (0.06) (0.11) (0.06) (0.08) (0.07) (0.06) (0.16) Post Secondary 0.01 0.47 0.07 0.07 0.01 0.01 1.05 (0.01) (0.13) (0.04) (0.04) (0.01) (0.01) (0.22) Rural Unfinished Primary 2.87 1.89 2.82 2.68 2.55 2.97 1.21 (0.29) (0.29) (0.27) (0.30) (0.28) (0.30) (0.30) Primary 2.26 1.96 2.45 2.07 2.28 2.28 1.42 (0.26) (0.29) (0.26) (0.27) (0.26) (0.28) (0.29) Vocational secondary 0.75 1.08 0.76 0.74 0.92 0.68 1.09 (0.15) (0.21) (0.16) (0.15) (0.18) (0.14) (0.24) General Secondary 0.43 0.74 0.45 0.29 0.49 0.35 0.73 (0.19) (0.29) (0.19) (0.16) (0.20) (0.18) (0.32) Post Secondary 0.05 0.35 0.04 0.05 0.05 0.05 0.61 (0.04) (0.18) (0.04) (0.50) (0.05) (0.05) (0.32) Notes: Standard errors are in parentheses, corrected for sample design effects. 87 Table B3.15: Relative Poverty Risks by Employment Status of the Head of the Household (0) SIMULATED Relative Poverty Risk after controlling for the partial effect of: ACTUAL (1) (2) (3) (4) (5) (6) Employment Status of Relative Poverty Household Head Risk Employment Household All Education Region Age status size factors Urban Employee 0.22 0.43 0.13 0.22 0.30 0.16 0.56 (0.08) (0.12) (0.05) (0.08) (0.09) (0.06) (0.13) Self-employed 0.18 0.23 0.25 0.34 0.18 0.19 1.23 (0.11) (0.13) (0.12) (0.15) (0.11) (0.11) (0.33) Unemployed 2.31 3.04 0.99 3.13 2.32 2.37 2.24 (0.59) (0.75) (0.48) (0.71) (0.60) (0.61) (0.84) Retired 0.83 0.80 0.92 0.96 0.73 0.91 1.08 (0.13) (0.14) (0.14) (0.14) (0.12) (0.13) (0.19) Other inactive 0.94 0.95 0.74 1.21 0.94 0,96 0.85 (0.40) (0.45) (0.35) (0.47) (0.40) (0.41) (0.53) Rural Employee 0.52 0.54 0.46 0.39 0.59 0.52 0.60 (0.14) (0.15) (0.13) (0.13) (0.14) (0.14) (0.21) Self-employed 1.66 1.80 1.96 1.71 1.76 1.77 1.53 (0.26) (0.31) (0.26) (0.27) (0.27) (0.27) (0.33) Unemployed 2.51 1.70 0.45 2.17 2.64 2.08 0.38 (0.69) (0.67) (0.27) (0.66) (0.70) (0.67) (0.28) Retired 2.10 1.71 2.48 1.95 1.95 2.05 1.21 (0.21) (0.20) (0.22) (0.21) (0.19) (0.21) (0.21) Other inactive 4.27 3.57 2.67 3.69 4.11 4.13 2.83 (0.58) (0.68) (0.51) (0.62) (0.57) (0.59) (0.82) Notes: Standard errors are in parentheses, corrected for sample design effects. 88 FigureB3.16: Actual versus Simulated Expenditure Distributions EDUCATION LABOR FORCE PARTICIPATION 1 Actual Actual 1 Simulated Simulated .8 .8 .6 .6 .4 .4 .2 .2 0 0 9 9.5 POVERTY LINE 10.5 11 11.5 12 9 9.5 POVERTY LINE 10.5 11 11.5 12 Log of per adult equivalent expenditure Log of per adult equivalent expenditure REGION DEPEN CY DEN RATIOS 1 Actual Actual .8 Simulate Simulated .8 .6 .6 .4 .4 .2 .2 0 0 9 9.5 POVERTY LINE 10.5 11 11.5 12 9 9.5 POVERTY LINE 10.5 11 11.5 12 Log of per adult equivalent expenditure Log of per adult equivalent expenditure H SEH OU OLD SIZE ALL FACTORS 1.5 .8 Actual Actual Simulated Simulated .6 1 .4 .5 .2 0 0 9 9.5 POVERTY LINE 10.5 11 11.5 12 9 9.5 POVERTY LINE 10.5 11 11.5 12 Log of per adult equivalent expenditure Log of per adult equivalent expenditure 89 BACKGROUND PAPER #4 REGIONAL DEVELOPMENT AND SOCIAL INDICATORS IN CROATIA Zeljko Lovrincevi Davor Mikuli INRODUCTION Systematic analysis of living standards, poverty, inequality and regional development are not performed on a regular basis in Croatia. Comprehensive profile of living standards and poverty has not been derived since the last World Bank report in 2001, while regional growth and social profile have not been examined at all. Therefore, the aim of this research is to provide a comprehensive profile of social and economic characteristics of Croatia's regions at NUTS III level. Regional profile of government's social transfers to households is also analyzed. In this paper, demographic and economic structure of Croatian economy is analyzed, as well as the process of secondary distribution of income in Croatia on the regional level. According to data availability limitation, the analysis was restricted to the period 2001-2003. We also tried to assess effectiveness of government social transfers to households given the regional inequality profile. Also, sources of growth on the regional level and growth prospects were identified. Final draft includes two appendices. The first appendix presents the regional GDP by counties for period 2001-2003, and the second appendix presents preliminary data on gross disposable income of household sector in Croatia. REGIONAL TRENDS IN THE NMS10 COUNTRIES AND EU15 Prior to the development analysis of the Croatian counties, we briefly outline the experiences of the new NMS10 members as well as EU countries in terms of the regional development level differences. According to the GDP dynamics, employment, unemployment and population figures in the NUTS II regions of the new member states (NMS1023), four groups of NUTS II regions can be identified in terms of the convergence process towards the EU15:24: ˇ Regions with high convergence potential ˇ Regions with moderate convergence potential ˇ Regions with moderate divergence risk ˇ Regions with high divergence risk. 23 Cyprus, the Check Republic, Estonia, Latvia, Lithuania, Hungary, Malta, Poland, Slovakia and Slovenia. 24 Revue élargissement no. 75, 11th April, 2005. 91 Figure 4.1 shows the average annual GDP growth rates in the various EU25 NUTS II regions. In the period 1995 - 2002, 31 out of the total of 41 regions on the NUTS II level of the new member states have recorded reductions in the difference in economical development, according to the GDP p.c. PPS with respect to the EU25 average. On average, the annual GDP p.c. PPS growth rate for the NMS10 amounted to 5.6 percent, while the EU15 countries recorded annual growth of 4 percent. Six out of the ten NUTS II regions in the NMS10, which have grown at a slower rate in comparison to the EU25 average, are in the Check Republic, which is, along with Cyprus, the only country with a recorded slower rate of growth in comparison to the average of the older EU15 members. In the first group of NUTS II regions with high convergence potential, the three Baltic countries (Estonia, Lithuania and Latvia) and Slovenia are included, representing small economies which have been classified as one NUTS II region despite being nation states. Also, the three Hungarian regions are included here, located in the area between Vienna and Budapest, and the eastern region surrounding Debrecen, as well as the two Slovakian regions: Bratislava and Eastern Slovakia (Kosice). All of these regions are characterized by a high GDP growth levels, reductions in unemployment levels, gradual reductions or increases in the number of employees, and favorable demographic trends (increases or small reductions in the population figures). NUTS II regions with moderate convergence potential are characterized by relatively dynamic growth (average GDP per capita growth rate, according to PPS above 4.5 percent), but in conditions of gradual increases in unemployment or decreases in employment. In these regions, it is expected that the positive components of the "creative destruction" process will overwhelm the negative ones, thus the continuation should record positive trends towards further convergence. In the NUTS II regions with moderate divergence risk, which are, in addition to Cyprus, found mainly along the borders of Check Republic and Germany, slower growth has been recorded (less or around the NMS10 average), however with increases in the unemployment rates. This group also includes the regions with slower GDP growth, but with reductions in the unemployment rates (Plzen, Karl. Vary). The high divergence risk group includes 11 NUTS II regions, where the average annual increases of the GDP per capita PPS are more than one standard deviation lower than the NMS10 average. Apart from the slower growth, these regions are characterized by increases in the unemployment figures which, in addition to the reductions of the employment levels, lead to significant increases of the unemployment rate. Regions in this group also exhibit the worst demographic trends (reduction in population numbers). The NUTS II region classification according to the economic growth potential is shown in Table 4.1. It should be emphasized that, for the benefit of easier location of the various regions on the map, the central cities of specific regions have been stated, and not the official name of the region. 92 Figure 4.1: Average Annual Real GDP Growth Rates of EU25 NUTS II Regions, 1995-2002, in % Source: Eurostat DG REGIO. 93 Table 4.1: Average Annual GDP p.c. PPS (1995-2002), Unemployment (1999-2003), Employment (1999-2003) and Population (1995-2002) Growth Rates NUTS II level regions GDP growth Unemployment Employment Population Regions with high convergence potential Budapest 8.1 -5.1 0.6 -0.3 Talinn 7.8 -2.9 0.5 -0.7 Riga 7.7 -5.3 0.7 -0.8 Bratislava 7.4 -0.8 -0.1 -0.4 Vilnius 7.1 -1.5 -0.8 -0.6 Gyor 6.5 0.9 0.2 0.0 Tatabanya 6.1 -5.5 1.2 0.0 Kosice 6.0 0.5 0.4 0.3 Ljubljana 5.5 -2.0 0.3 0.1 Debrecen 5.1 -7.8 1.5 0.1 Regions with moderate convergence potential Warsaw 8.2 9.8 -3.0 0.2 Prague 6.5 1.0 -0.4 -0.6 Poznan 6.4 11.8 -0.3 0.1 Zilina 6.2 2.0 -0.1 0.0 Trnava 5.2 2.3 0.6 -0.1 La Valette 5.0 0.5 0.6 0.8 Pecs 4.9 -1.0 0.5 -0.1 Miskolc 4.6 -3.5 1.4 -0.1 Kladno 4.5 -8.3 0.9 0.2 Regions with moderate divergence risk Bialystok 5.8 7.7 -3.2 -0.1 Wroclaw 5.7 11.9 -4.4 -0.3 Kielce 5.6 7.7 -2.1 -0.3 Lodz 5.5 10.1 -1.1 -0.4 Krakow 5.4 14.1 -1.3 0.2 Gdansk 5.4 13.1 -2.1 0.1 Nikosia 3.8 -5.0 3.2 1.1 Jihaliva 3.6 -2.8 -0.2 -0.2 Hradec 3.4 -3.3 -0.1 -0.1 Plzen 3.1 -4.0 0.1 -0.1 Karl. Vary 1.5 -3.7 0.1 -0.1 Regions with high divergence risk Szcecin 5.0 5.2 -2.9 -0.1 Olsztyn 4.9 4.2 -1.0 -0.2 Rzeszow 4.8 7.0 -0.8 0.0 Bydgoszcz 4.6 10.6 -0.5 -0.1 Katowice 4.4 12.7 -2.1 -0.4 Zielona G. 4.4 7.8 -0.7 -0.3 Lublin 4.1 8.5 -0.5 -0.1 Szeged 3.9 2.3 -0.4 0.0 Opole 3.4 5.4 -3.1 -0.3 Brno 2.6 -2.2 -0.2 -0.1 Ostrava 2.2 2.5 -0.6 -0.3 Source: Revue élargissement, no. 75, 11th April 2005. It is important to notice that during the period 1995-2002, in the group of new member states there was a worsening of the ratio between the most developed and least developed region in all of the 4 analyzed countries with defined several NUTS II regions (Table 4.2).25 25 Malta, Cyprus, Estonia, Latvia, Lithuania and Slovenia are simultaneously state and NUTS II regions. 94 Table 4.2: Ratio between NUTS II Region with Highest/Lowest GDP p.c. PPS, 1995-2002 Country 1995 1997 1999 2001 2002 NMS Check Republic 2.4 2.5 2.6 2.9 2.9 Hungary 2.0 2.2 2.4 2.4 2.6 Poland 1.7 1.8 2.1 2.1 2.2 Slovakia 2.8 2.8 2.9 3.0 3.1 EU15 Belgium 3.0 3.1 3.2 3.2 3.1 Germany 2.9 2.9 2.8 2.9 2.8 Greece 2.1 1.9 1.8 1.9 1.9 Spain 2.1 2.1 2.1 2.1 2.1 France 3.0 3.0 2.9 3.0 3.1 Ireland 1.4 1.5 1.5 1.5 1.6 Italy 2.4 2.3 2.3 2.3 2.4 Netherlands 1.6 1.7 1.7 1.7 1.7 Austria 2.3 2.2 2.2 2.2 2.1 Portugal 1.9 2.0 1.9 1.9 1.8 Finland 1.6 1.7 1.9 1.9 1.9 Sweden 1.5 1.6 1.6 1.6 1.6 Great Britain 4.1 4.2 4.4 4.3 4.3 Source: Eurostat. The most significant regional development differences are noted in Slovakia and the Check Republic, while the relatively highest increases in inequalities are recorded in Hungary. Usually, this increasing divergence is consequence of the above average growth of regions comprising the capital city and the surrounding regions. That phenomenon occurs as a result of the so called "gateway" effect, where almost all of the capital cities in the transition countries represent the entry point for foreign investments. That implies a concentration of primarily financial services, telecommunications, IT and other logistic activities in the capital cities. The process is clear and present despite the efforts by the governments in the transition countries to achieve balanced regional development. On the other hand, old member countries (EU15) are clearly experiencing an end of the trend of further centralization of economic activities in the most developed areas. The primary reason for this is the planned policy of balanced regional development supported by the European structural funds. Therefore in the NMS group, an end of the trend of centralization of economic activities and an emphasis on a more balanced regional development can be expected in the long run. In the continuation, the dynamic of the economic structure changes in the European NUTS II regions are analyzed. Table 4.3 shows the growth rates of different activities, as classified in the National classification of economic activities (NACE), for the EU25 countries. It is clear that throughout the period, the GDP of new members grew at a faster rate in comparison to the old members, as expected since with the EU accession process, real convergence process begun, thus there is a so called low basis effect present. On average, the real growth of the new members was faster by 0.87 percentage points. However, when analyzing the growth according to activities, it is noticeable that the highest average growth among the new member states is recorded in the retail, hotel, restaurant and transportation sector, (G, H and I) and the business and financial services sector (J and K). Industry (C, D and E) and construction (F) are growing slightly faster than the total GDP, with noticeable seasonal pattern over the year, while the activities of public administration, education, health and other personal and community services (L, M, N and O) are growing at a slower rate in comparison to the average. The slowest growth (and in some cases real decreases) have been recorded in agriculture and fisheries (A and B). Such trends indicate that the economic structure greatly influences the growth potential of certain NUTS II regions. A more favorable current economic structure (a higher proportion of propulsive service sectors and a smaller proportion of agriculture and government services) ensures higher growth rates in the middle run. Thus, this influenced further increases in the differences between the developed and less developed regions in the NMS10, since the most favorable economic structure is found in the most developed regions. 95 Table 4.3: Real GDP Growth Rates of the EU25 According to the NACE Classification Activities, 2000-2003 GDP A, B C, D, E F G, H, I J, K L, M, N , 00 01 02 03 00 01 02 03 00 01 02 03 00 01 02 03 00 01 02 03 00 01 02 03 00 01 02 03 euro- 3.5 1.6 0.9 0.5 -0.6 -2.4 0.6 -3.8 4.3 0.6 0.2 0.0 2.7 0.0 -0.5 -0.4 4.4 3.2 1.2 0.5 4.9 2.9 0.8 1.4 2.3 1.8 2.2 0.6 zone EU25 3.6 1.7 1.1 0.9 -0.6 -2.4 1.3 -3.3 4.2 0.2 0.1 0.4 2.5 0.2 0.0 0.2 4.6 3.2 1.6 1.0 4.9 3.1 1.0 1.9 2.3 1.8 2.2 0.7 EU15 3.6 1.7 1.0 0.8 -0.5 -2.8 1.4 -3.4 4.1 0.2 0.0 0.1 2.5 0.4 0.0 0.2 4.6 3.1 1.5 0.9 4.9 3.1 1.0 1.9 2.3 1.8 2.2 0.7 Members BE 3.9 0.7 0.9 1.3 1.0 -11.3 12.7 -3.2 5.0 -0.3 -0.2 -0.3 7.7 1.2 -1.5 -0.4 2.9 2.3 3.5 0.8 2.3 1.6 0.1 3.0 2.6 1.3 0.9 1.6 CZ 3.9 2.6 1.5 3.7 5.7 -7.0 2.6 -1.0 7.1 -5.0 7.8 7.1 -0.3 -8.2 3.1 -0.5 1.7 9.1 -1.0 -0.9 4.8 11.9 -3.9 5.3 3.8 1.4 0.9 1.6 DK 2.8 1.6 1.0 0.4 6.4 -1.5 -4.3 3.3 3.3 0.0 -0.4 -0.2 1.6 4.0 0.6 -2.7 7.3 2.9 1.9 1.5 3.9 3.9 1.3 0.4 -0.2 1.3 1.8 0.6 DE 2.9 0.8 0.1 -0.1 -0.8 0.3 -1.8 -0.7 4.6 -1.3 -0.4 0.5 -2.3 -5.6 -4.8 -4.4 3.4 3.9 1.4 0.9 4.8 3.8 0.1 0.5 2.1 0.2 1.6 -0.4 EE 7.8 6.4 7.2 5.1 -0.8 -5.4 0.1 -1.5 13.9 8.5 12.9 8.0 13.8 4.3 20.6 7.7 7.2 9.1 4.2 7.2 8.3 5.9 5.4 2.1 2.6 4.1 3.9 1.8 EL 4.5 4.3 3.6 4.5 -3.7 -3.8 -1.2 -4.0 5.3 3.0 2.6 2.6 5.7 14.4 0.8 11.2 7.6 7.0 3.6 6.1 5.1 4.2 -0.2 4.3 1.4 3.1 8.0 3.4 ES 4.4 2.8 2.2 2.5 2.7 -2.9 1.7 -1.4 3.9 2.5 0.7 1.3 6.1 5.3 5.2 4.3 3.8 3.5 1.9 1.6 5.9 6.1 0.4 1.3 4.4 2.5 2.8 3.3 FR 3.8 2.1 1.2 0.5 -2.2 -4.0 4.7 -7.4 4.2 2.8 0.8 -0.1 7.1 1.5 -0.1 -1.4 4.5 2.4 0.4 -0.8 4.8 0.8 1.6 3.1 2.4 3.3 3.4 -0.1 IE 9.9 6.0 6.1 3.7 IT 3.0 1.8 0.4 0.3 -2.9 -0.5 -3.9 -5.7 2.3 -0.2 -0.3 -1.0 3.5 3.1 2.5 2.5 5.6 3.5 -0.1 0.0 5.1 2.8 1.8 1.2 1.0 2.1 1.2 0.6 CY 5.0 4.1 2.1 1.9 -5.9 3.8 5.4 4.5 3.4 0.3 0.4 0.6 -1.2 4.0 4.7 4.4 8.8 5.2 -0.3 -0.7 6.4 5.5 3.5 2.7 3.6 3.1 4.0 4.3 LV 6.9 8.0 6.4 7.5 11.5 6.4 4.4 1.0 4.4 9.7 8.1 7.8 8.2 6.1 10.8 13.7 8.2 10.2 8.1 10.4 12.3 12.2 5.5 4.3 1.4 2.0 2.8 3.3 LT 3.9 6.4 6.8 9.7 6.4 -4.6 8.2 2.2 5.4 13.9 4.6 15.8 -18.2 7.4 12.7 22.0 6.7 8.1 9.3 9.1 5.0 5.6 6.6 6.2 4.3 -0.9 2.9 2.4 LU 9.0 1.5 2.5 2.9 -7.2 -15.1 0.1 -2.3 6.9 1.0 2.0 2.6 5.1 8.0 5.3 4.3 9.4 6.3 3.3 3.4 9.4 -0.5 2.5 1.7 2.9 4.7 1.0 3.6 HU 5.2 3.8 3.5 3.0 -7.4 23.4 -12.1 -4.0 6.4 0.4 1.3 5.4 19.2 5.2 12.9 1.2 0.8 5.0 4.7 4.3 8.3 4.4 6.3 -1.3 3.3 3.4 2.3 3.0 MT 6.4 -2.4 2.6 -0.3 NL 3.5 1.4 0.6 -0.9 1.5 -3.6 -1.6 -1.9 3.5 0.5 -1.0 -2.1 4.2 2.1 -3.2 -3.0 6.3 0.7 0.8 -1.3 2.9 1.6 -0.4 -0.1 1.6 2.8 2.8 2.3 AT 3.4 0.7 1.2 0.8 -3.0 0.6 -0.5 -1.3 6.2 2.5 1.7 0.2 1.7 -3.5 0.4 5.1 3.2 2.0 2.5 1.8 6.3 0.7 1.6 0.6 1.3 -0.7 -1.3 -0.5 PL 4.0 1.0 1.4 3.8 -7.9 9.2 2.0 2.1 6.5 -0.3 -0.2 6.3 0.3 -7.9 -6.8 -2.9 4.0 3.6 5.9 2.3 4.5 2.7 -0.1 5.8 2.5 0.4 0.2 3.4 PT 3.4 1.7 0.4 -1.2 -4.0 -0.3 5.7 -6.7 2.7 1.9 -1.0 -0.5 4.9 2.8 -3.8 -11.4 4.1 4.0 1.0 0.1 7.1 4.3 0.9 2.0 3.5 2.1 1.5 -1.4 SI 3.9 2.7 3.3 2.5 0.8 -12.1 15.4 -15.3 8.4 5.0 4.7 3.3 0.9 -2.2 0.6 3.4 2.1 3.7 3.4 3.1 1.2 4.7 4.0 4.3 4.8 3.3 2.3 3.0 SK 2.0 3.8 4.6 4.0 1.9 4.9 -1.6 4.4 0.8 1.4 -0.3 9.5 0.2 -0.5 9.3 6.9 2.2 9.7 -3.2 -2.4 2.5 1.9 17.5 8.9 2.8 14.0 16.6 5.5 FI 5.1 1.1 2.3 2.0 10.6 -4.8 3.6 0.6 11.0 0.3 2.1 0.9 -2.6 -2.9 2.5 1.3 5.3 2.8 2.3 3.3 5.3 0.7 2.6 3.6 2.1 3.0 1.8 0.5 SE 4.3 1.0 2.0 1.5 2.8 4.3 2.8 1.2 8.2 -1.6 4.5 1.9 0.7 5.2 -0.4 0.8 3.7 0.5 1.7 2.3 4.9 1.8 -0.3 1.9 1.7 1.4 1.9 1.3 UK 3.9 2.3 1.8 2.2 -0.6 -9.1 11.9 -2.6 1.9 -1.6 -2.5 -0.2 1.3 1.8 3.8 5.2 5.1 2.9 3.6 2.5 5.0 4.6 2.1 4.6 3.2 2.4 2.6 1.3 Source: Eurostat DG REGIO. 96 DEMOGRAPHIC AND ECONOMIC CHARACTERISTICS OF THE CROATIAN REGIONS After the experience of the NMS10, in this section demographic and economic features of Croatian regions are analyzed. Because of some doubts regarding final regional breakdown of Croatia instead of preliminary NUTS II region, we rather used breakdown based on so-called analytical regions. According to main features Croatia could be divided in 5 analytical regions: Zagreb region, Central Croatia, Eastern Croatia, Adriatic North and Adriatic South. Besides analytical regions, all data are presented using current administrative breakdown on counties. Such administrative division of Croatia fulfils all of the EU criteria for NUTS III breakdown. Demographic structure Tables 4.4 and 4.5 show the demographic structure of the Croatian regions. The working population comprises 64 percent of the total population in Croatia, i.e. the male population between the age 15 and 64 and female population between 15 and 59. Senior population accounts for little less than 19 percent-and the remaining 17 percent of the total population are children. Table 5.4.4: Demographic Structure of Croatia, by County (NUTS III) Demographic structure Demographic structure, % Senior Senior County Working Working population population Children contingent Children contingent (F above (F above (0-14) (F 15-59, (0-14) (F 15-59, 60, M 60, M M 15-64) M 15-64) above 65) above 65) Zagreb 53 822 202 003 51 866 17.5 65.7 16.9 Krapina-Zagorje 24 293 89 662 28 142 17.1 63.1 19.8 Sisak-Moslavina 29 948 114 647 40 331 16.2 62.0 21.8 Karlovac 20 521 86 853 33 496 14.6 61.7 23.8 Varazdin 31 807 118 247 34 061 17.3 64.2 18.5 Koprivnica-Krizevci 21 064 78 410 24 604 17.0 63.2 19.8 Bjelovar-Bilogora 22 805 82 283 27 544 17.2 62.0 20.8 Primorje-Gorski kotar 42 835 201 527 59 482 14.1 66.3 19.6 Lika-Senj 8 200 30 896 14 315 15.4 57.8 26.8 Virovitica-Podravina 16 962 57 820 18 059 18.3 62.3 19.5 Pozega-Slavonia 16 966 52 097 16 420 19.8 60.9 19.2 Slavonski Brod- 34 728 108 692 32 353 19.8 61.8 18.4 Posavina Zadar 29 496 101 242 30 336 18.3 62.9 18.8 Osijek-Baranja 58 719 210 882 60 036 17.8 64.0 18.2 Sibenik-Knin 18 953 67 375 26 063 16.9 59.9 23.2 Vukovar-Sirmium 39 359 128 317 36 119 19.3 63.0 17.7 Split-Dalmatia 85 585 296 386 79 531 18.5 64.2 17.2 Istria 31 177 135 445 38 984 15.2 65.9 19.0 Dubrovnik-Neretva 22 467 76 565 23 282 18.4 62.6 19.0 Meimurje 21 964 76 703 19 484 18.6 64.9 16.5 City of Zagreb 122 963 512 580 140 381 15.8 66.1 18.1 Croatia 754 634 2 828 632 834 889 17.1 64.0 18.9 Source: Census 2001, CBS. 97 The demographic structure of certain counties (NUTS III level) and 5 analytical regions significantly differ from the Croatian average. Therefore, according to the ratio of children, the two "youngest" counties are the County of Pozega-Slavonia and Slavonski Brod-Posavina, in which children account for almost a fifth of the total population. If one considers that County of Vukovar-Sirmium can also be included in this group of counties with a high proportion of children, it can be concluded that Eastern Croatia is the youngest region in Croatia. The highest proportion of the working population is present in three most developed counties (according to the GDP per capita levels ­ see Table 4.8), i.e. County of Primorje-Gorski kotar, County of Istria and the City of Zagreb (66 percent). The Zagreb region has the highest proportion of the working population in the overall population, followed by Adriatic North. Central Croatia is the most senior region in Croatia, with more than 20 percent of senior population, closely followed by the Adriatic North. The counties with the highest proportion of seniors are County of Lika-Senj (26.8 percent), County of Karlovac (23.8 percent) and County of Sibenik-Knin (23.2 percent). Table 4.5: Demographic Structure of Croatia by Analytical Region Demographic structure Demographic structure, % Senior Senior Region Working Working population population Children contingent Children contingent (F above (F above 60, (0-14) (F 15-59, (0-14) (F 15-59, 60, M M above M 15-64) M 15-64) above 65) 65) Zagreb region* 176 785 714 583 192 247 16.3 65.9 17.7 Central Croatia** 172 402 646 805 207 662 16.8 63.0 20.2 Adriatic North*** 82 212 367 868 112 781 14.6 65.4 20.0 Adriatic South**** 156 501 541 568 159 212 18.3 63.2 18.6 Eastern Croatia***** 166 734 557 808 162 987 18.8 62.8 18.4 Croatia 754 634 2 828 632 834 889 17.1 64.0 18.9 *the City of Zagreb, the County of Zagreb ** Counties of Krapina-Zagorje, Sisak-Moslavina, Karlovac, Varazdin, Koprivnica-Krizevci, Bjelovar-Bilogora, Meimurje *** Counties of Istria, Primorje-Gorski kotar, Lika-Senj, ****Counties of Zadar, Sibenik-Knin, Split-Dalmatia, Dubrovnik-Neretva *****Vukovar-Sirmium, Osijek-Baranja, Slavonski Brod-Posavina, Pozega-Slavonia, Virovitica-Podravina Source: Census 2001, CBS. Tables 4.6 and 4.7 show the education structure of the Croatian regions. The inhabitants of certain counties and regions have been classified according to the education level into three groups: ˇ Primary education (no school and elementary school) ˇ Secondary education (high school) ˇ Tertiary education (higher education and university level). The most prominent education level in Croatia is the high school or secondary education level (47.4 percent), with the lowest share of the higher school education and university levels of education (12.0 percent). The regions with the poorest education structure are the Central and Eastern Croatia, where almost half of the population has no school or have elementary education only. Higher and university educated population in these regions comprise 7 percent of total population. The extremely unfavorable education structure is recorded in County of Koprivnica-Krizevci (58.2 percent of the population with primary education and 7.0 percent with higher school and university level education) and the County of Virovitica-Podravina (56.8 percent of the population with primary education and 5.8 percent with higher school and university level education). The highest proportion of the tertiary education level can be found in the Zagreb region (18.5 percent) and the Adriatic North (13.5 percent). 98 Table 4.6: Education by County, 2001, in % County of Primary Secondary Tertiary Zagreb 43.6 48.5 7.9 Krapina-Zagorje 53.0 41.1 5.8 Sisak-Moslavina 48.7 43.7 7.6 Karlovac 47.3 44.1 8.7 Varazdin 46.0 45.6 8.4 Koprivnica-Krizevci 58.2 34.8 7.0 Bjelovar-Bilogora 54.9 38.5 6.6 Primorje-Gorski kotar 30.2 54.5 15.3 Lika-Senj 52.2 40.7 7.1 Virovitica-Podravina 56.8 37.4 5.8 Pozega-Slavonia 53.4 39.9 6.8 Slavonski Brod-Posavina 49.7 43.4 6.9 Zadar 40.8 48.4 10.8 Osijek-Baranja 46.6 44.2 9.2 Sibenik-Knin 42.8 47.6 9.5 Vukovar-Sirmium 51.5 41.8 6.7 Split-Dalmatia 34.0 52.4 13.6 Istria 36.9 50.5 12.6 Dubrovnik-Neretva 34.5 51.4 14.0 Meimurje 48.3 45.2 6.6 City of Zagreb 25.1 52.3 22.6 Croatia 40.6 47.4 12.0 Source: Census 2001, CBS. Table 4.7: Education by region, 2001, in % Analytical regions Primary Secondary Tertiary Zagreb region 30.3 51.2 18.5 Central Croatia 50.5 42.1 7.3 34.7 51.7 13.5 Adriatic North 36.5 50.9 12.6 Adriatic South Eastern Croatia 50.0 42.4 7.6 Croatia 40.6 47.4 12.0 Source: Census 2001, CBS. Economic Structure The three most developed counties according to the GDP per capita are the City of Zagreb, the County of Istria and the County of Primorje-Gorski kotar. Tables 4.8 and 4.9 show the development level of the Croatian counties according to that indicator. In addition to the City of Zagreb, County of Istria and County of Primorje-Gorski kotar, only County of Koprivnica-Krizevci in 2001 and 2002 and County of Lika-Senj in 2003 have reached the GDP per capita levels above the Croatian average. 99 At the regional level, the Zagreb region and Adriatic North have above average GDP p.c. and along with Adriatic South generate the highest increases in GDP levels. The least developed are Counties of Vukovar-Sirmium and Slavonski Brod-Posavina, where the GDP per capita levels reach less then 60 percent of the Croatian national average. It should be noted that some of the less developed counties generate below average growth levels of GDP (Counties of Krapina-Zagorje, Pozega-Slavonia and Slavonski Brod-Posavina), and are therefore lagging even further in relation to the other counties (the last three columns in Table 4.8). Table 4.8: Gross Domestic Product per capita, by County, Croatia =100 Index Index Index County of 2001 2002 2003 2002/2001 2003/2002 2003/2001 Zagreb 67.9 77.1 74.2 125.1 106.5 133.3 Krapina-Zagorje 79.0 74.6 72.6 102.6 105.8 108.6 Sisak-Moslavina 86.8 81.2 77.0 101.7 103.2 104.9 Karlovac 84.9 85.4 77.7 109.1 98.7 107.7 Varazdin 95.1 98.1 94.2 112.5 104.6 117.6 Koprivnica-Krizevci 103.5 101.9 95.8 107.2 102.5 109.9 Bjelovar-Bilogora 78.5 79.8 74.7 110.4 101.7 112.2 Primorje-Gorski kotar 117.5 112.5 118.1 104.6 114.8 120.2 Lika-Senj 80.2 90.9 103.4 122.9 124.1 152.5 Virovitica-Podravina 80.0 78.1 75.4 106.2 105.1 111.6 Pozega-Slavonia 73.9 71.2 72.2 105.2 110.6 116.3 Slavonski Brod-Posavina 61.0 60.1 57.5 107.6 104.5 112.5 Zadar 72.1 73.4 80.1 112.4 120.7 135.8 Osijek-Baranja 77.6 80.1 75.3 112.7 102.6 115.6 Sibenik-Knin 63.6 65.8 69.7 113.2 116.2 131.6 Vukovar-Sirmium 58.0 58.3 57.5 109.3 107.3 117.3 Split-Dalmatia 75.8 75.1 75.3 109.2 110.4 120.5 Istria 134.5 135.6 137.5 110.7 111.6 123.6 Dubrovnik-Neretva 90.2 86.8 88.4 105.7 111.8 118.2 Meimurje 83.1 84.9 80.2 111.8 103.3 115.5 City of Zagreb 176.4 174.8 179.2 108.4 112.3 121.8 Croatia 100.0 100.0 100.0 109.4 109.5 119.8 Source: Project CBS-EIZG, Regional GDP preliminary results. Table 4.9: Gross Domestic Product per capita by regions, Croatia = 100 Index Index Index Analytical regions 2001 2002 2003 2002/2001 2003/2002 2003/2001 Zagreb region 145.5 146.8 148.9 110.7 111.4 123.3 Central Croatia 87.5 86.7 81.9 107.8 102.9 110.9 Adriatic North 120.2 118.9 123.8 108.3 114.1 123.6 Adriatic South 75.5 75.2 77.3 109.6 113.2 124.1 Eastern Croatia 69.7 70.0 67.4 109.6 104.9 115.0 Croatia 100.0 100.0 100.0 109.4 109.5 119.8 Source: Project CBS-EIZG, Regional GDP preliminary results. Tables 4.10 and 4.11 describe the employment structure according to the economic activities and unemployment rates. The economic activities have been separated into agriculture, i.e. the primary 100 sector (activities A and B), industry, i.e. the secondary sector (activities C, D and E) and services, i.e. the tertiary sector, further separated into the public sector (activities L, M and N) and other service sector (activities F, G, H, I, J, K, O and P). At the national level, the highest employment can be found in the service sector (45.7 percent), and lowest employment in agriculture (10.4 percent). Table 4.10: Employment Structure by Economic Activity and Unemployment Rate, by County, 2003, in % F, G, H, I, J, Unemployment County of A, B C, D, E L, M, N K, O, P rate* Zagreb 15.9 27.6 41.7 14.8 22.0 Krapina-Zagorje 25.5 28.8 28.6 17.1 16.3 Sisak-Moslavina 12.6 33.2 33.1 21.1 31.9 Karlovac 12.9 28.2 39.1 19.9 29.8 Varazdin 11.2 36.1 33.9 18.8 16.9 Koprivnica-Krizevci 28.5 30.7 25.9 14.9 18.6 Bjelovar-Bilogora 35.0 22.6 26.4 16.0 26.0 Primorje-Gorski kotar 2.4 22.0 54.6 20.9 16.5 Lika-Senj 18.2 13.6 44.5 23.7 26.7 Virovitica-Podravina 26.0 27.1 29.0 17.8 30.9 Pozega-Slavonia 20.8 27.7 30.1 21.4 25.4 Slavonski Brod-Posavina 18.1 25.3 37.0 19.6 32.7 Zadar 7.0 15.8 53.1 24.2 30.1 Osijek-Baranja 14.3 22.5 41.6 21.6 29.5 Sibenik-Knin 5.6 22.5 48.1 23.8 35.4 Vukovar-Sirmium 24.5 16.1 38.1 21.3 36.2 Split-Dalmatia 3.9 20.7 52.9 22.5 28.4 Istria 3.9 23.5 54.1 18.6 11.3 Dubrovnik-Neretva 6.3 11.8 60.4 21.5 22.7 Meimurje 25.0 29.2 32.0 13.9 15.6 City of Zagreb 1.4 19.9 55.0 23.7 12.5 Croatia 10.4 23.3 45.7 20.7 21.9 * registered unemployment, not ILO definition. Source: CBS. Croatia is characterized by significant differences in regional unemployment levels, outlined by the data in the last column of Table 4.10. On one hand, there are most developed counties with relatively low unemployment levels in comparison with the Croatian national average (21.9 percent in 2002.). Thus, the County of Istria has 11.3 percent unemployment rate, and the City of Zagreb 12.5 percent. On the other hand, certain counties are recording high unemployment levels, namely County of Vukovar-Sirmium (36.2 percent), County of Sibenik-Knin (35.4 percent) and County of Slavonski Brod-Posavina (32.7 percent). Table 4.11: Employment Structure by Economic Activities, by Analytical Regions, 2003, in % F, G, H, I, J, Total Analytical regions A, B C, D, E L, M, N K, O, P employment Zagreb region 3.9 21.2 52.7 22.2 100.0 Central Croatia 20.9 30.3 31.4 17.5 100.0 Adriatic North 4.0 22.1 53.8 20.1 100.0 Adriatic South 5.0 18.6 53.6 22.8 100.0 Eastern Croatia 18.9 22.8 37.5 20.7 100.0 Croatia 10.4 23.3 45.7 20.7 100.0 Source: CBS. 101 The highest proportion of employment in agriculture can be found in Central Croatia (20.9 percent), where the leaders are Counties of Bjelovar-Bilogora (35.0 percent), Koprivnica-Krizevci (28.5 percent), Krapina-Zagorje (25.5 percent) and Meimurje (25.0 percent). The high proportion of employment in agriculture is also present in the traditional agriculture counties in Eastern Croatia. The Zagreb region, Adriatic North and Adriatic South have single-digit agriculture employment levels, with the lowest percentage recorded in the City of Zagreb (1.4 percent). Central Croatia is also characterized with the relatively highest proportion of employment in the industry sector (C, D, E). In this case, Counties of Varazdin and Sisak-Moslavina are the leaders with over a third of total employment in the industry sector. Other regions have similar proportions of industry employment (ranging from 18.6 percent to 22.8 percent). The lowest proportion of employment in the industry sector has been recorded in the County of Dubrovnik-Neretva (11.8 percent) and the County of Lika-Senj (13.6 percent). The Zagreb region, the Adriatic North and the Adriatic South have the highest share of the service sector. Over a half of total employment in these regions is in the service sector. The highest proportions have been recorded in the County of Dubrovnik-Neretva (60.4 percent) and the City of Zagreb (55.0 percent). The most equalized employment level can be found in the public sector. The highest proportions are in the County of Zadar (24.2 percent), County of Sibenik-knin (23.8 percent), County of Lika-Senj and the City of Zagreb (both with 23.7 percent). Table 4.12 shows the approximation of labor productivity according to the regions in Croatia, measured by the relation of gross value added per employee. It should be emphasized that different data sources are used regarding the value added (national accounts data) and number of person employed (administrative data). Productivity calculated in that way therefore is treated only as indication of real productivity. Methodological issues and different data sources can influence the reliability of productivity indicator. Because of that, productivity indicator is not presented on the county, but only on region level. The only two regions with above-average productivity are the Zagreb region (17.4 percent above average) and North Adriatic (5.8 percent above average). The Zagreb region has above average labor productivity in all sectors, except the primary sector. Eastern and Adriatic North and South recorded above-average labor productivity in the primary sector. Table 4.12: Estimation of Labor Productivity by Economic Activities, by Analytical Regions, 2003, Croatia = 100 F, G, H, I, J, Analytical regions A, B C, D, E L, M, N Total K, O, P Zagreb region 93.5 133.9 115.1 103.9 117.4 Central Croatia 76.9 82.8 96.7 95.8 87.1 Adriatic North 144.8 121.4 97.4 97.4 105.8 Adriatic South 137.9 83.0 85.9 100.0 91.0 Eastern Croatia 120.9 67.2 86.7 99.2 88.8 Croatia 100.0 100.0 100.0 100.0 100.0 Source: Project CBS-EIZG, Regional GDP preliminary results. Table 4.13 shows the road and water infrastructure indicators for Croatian counties. Infrastructure equipment is an important prerequisite for economic development. The end of this section will show the correlation analysis results, attempting to identify the various factors relevant in explaining the regional development differences, i.e. identify the factors showing statistically significant correlation with the regional GDP per capita levels. GDP per capita has been alternatively linked with specific variables from various groups of explanatory factors. As data on GDP are available for period 2001- 2003 only, it is not possible to construct a strong econometric model implying causality in GDP and explanatory variables trends. Correlation coefficients are to be used as indication whether GDP and other variables are moving in the same or opposite direction. 102 Table 4.13: Infrastructure Development Indicators Waste waters Total water Road density in from public Number of inhabitants delivered to County of relation to surface sewage per road km users area (km/km2) (m3/household) (m3/household) 2001 2003 2001 2003 2001 2001 Zagreb 160.351 165.033 0.616 0.612 59.538 113.30 Krapina-Zagorje 151.231 153.034 0.763 0.758 32.949 10.69 Sisak-Moslavina 90.722 90.713 0.455 0.463 43.698 34.84 Karlovac 86.604 86.860 0.447 0.449 50.533 33.99 Varazdin 165.971 160.616 0.879 0.909 59.343 54.76 Koprivnica-Krizevci 112.898 105.655 0.632 0.672 38.614 52.23 Bjelovar-Bilogora 90.706 88.652 0.540 0.564 25.392 26.35 Primorje-Gorski kotar 199.091 203.837 0.426 0.417 110.320 74.99 Lika-Senj 28.319 28.169 0.345 0.355 93.794 18.27 Virovitica-Podravina 102.418 101.996 0.446 0.451 34.747 26.03 Pozega-Slavonia 116.637 118.388 0.398 0.397 37.322 29.98 Slavonski Brod- 192.001 191.209 0.444 0.455 29.105 23.16 Posavina Zadar 95.860 95.032 0.455 0.463 74.237 34.46 Osijek-Baranja 199.961 205.656 0.394 0.389 43.600 37.25 Sibenik-Knin 92.696 94.293 0.404 0.403 80.236 38.06 Vukovar-Sirmium 196.658 200.969 0.411 0.413 39.315 18.43 Split-Dalmatia 185.232 186.813 0.545 0.546 93.797 63.74 Istria 110.839 112.308 0.660 0.650 121.130 52.75 Dubrovnik-Neretva 124.868 125.912 0.548 0.548 89.693 32.87 Meimurje 212.866 214.286 0.748 0.756 39.880 16.01 City of Zagreb 1..039.215 1037.543 1.158 1.172 105.450 112.50 Croatia 155.792 155.689 0.497 0.501 71.688 - Source: CBS. The first group comprises structural variables, defined as the proportion of employed population in various economy sectors for each specific county. Sectors are defined in the same manner used in the explanations accompanying Tables 4.10 and 4.11. As shown in Table 4.14, it is evident that there is a statistically significant correlation between the proportion of employed in the primary sector and the size of GDP per capita in a specific county, a correlation with medium intensity and a negative sign (-0.46 for period 2001- 2003). In other words, less developed counties on average have higher employment levels in the primary sector. The proportion of employment in the secondary sector has not been identified as significant as an explanation for regional development, same as in the case of tertiary sector employment. However, the latter variable does show statistical correlation with the regional GDP per capita. When excluding the public sector employees from the sample, the correlation coefficient becomes significant, with a medium intensity positive sign (0.41). The correlation coefficient increases even further following the exclusion of tourist sector employees (0.47). Furthermore, the county infrastructure equipment levels also have a statistically significant correlation with the regional GDP per capita. In this analysis, all of the infrastructure variables included have shown a positive correlation with the level of regional GDP per capita, with medium and strong intensity. The correlation coefficient between the GDP per capita and the road density is 0.63, while in the case of the correlation with the total water delivered to users, the resulting value is 0.62. The education structure in specific counties has also been found as significant variable in explaining regional development differences.26 The correlation analysis has established a statistically significant correlation between the regional GDP per capita and the proportion of population with primary and tertiary education levels. The correlation coefficient in the first example has a negative sign and the 26 The education structure of the population in specific Croatian regions has been shown by Tables 4.6 and 4.7. 103 value of -0.59, while in the second case, an even stronger positive relationship has been established, with the coefficient value of 0.75. Table 4.14: Overview of Correlation Coefficients between County GDP p. c. and Various Variables Correlation between GDP p.c. and: 2001 2002 2003 2001-2003 Structural variables -0.45 -0.44 -0.50 -0.46 Employed in primary sector (-2.26)* (-2.22) (-2.61) (-4.09) Employed in tertiary sector (without 0.37 0.40 0.48 0.41 public service) (1.8) (1.93) (2.44) (3.61) Employed in tertiary sector (without 0.44 0.46 0.52 0.41 tourism) (2.19) (1.93) (2.74) (3.61) Infrastructure variables 0.67 0.67 0.61 0.63 Road intensity (4.06) (4.07) (3.42) (6.50) 0.59 0.62 0.71 0.62 Water (3.26) (3.54) (4.48) (6.37) Educational structure -0.59 -0.59 -0.63 -0.59 Primary education (-3.31) (-3.26) (-3.65) (-5.81) 0.77 0.76 0.79 0.75 High education (5.40) (5.18) (5.73) (9.01) Demographic structure -0.57 -0.59 -0.60 -0.57 Children (-3.10) (-3.28) (-3.31) (-5.51) 0.55 0.53 0.46 0.49 Working contingent (2.92) (2.79) (2.33) (4.53) *Note: in parentheses are t-statistics Source: Own calculations. Finally, the demographic structure is equally important in explaining of regional development discrepancies. A statistically significant correlation with the regional GDP per capita have the variables of the proportion of children in the total county population levels, and the level of working population, while the ratio of senior population has not been shown as significant. The correlation between the regional GDP per capita and the proportion of children (as an indicator of economically supported population) has a negative sign, with medium to strong intensity (-0.57). On the other hand, the working population has shown a positive relationship with the regional GDP per capita, also with a medium to strong intensity (0.49). SOURCES OF GROWTH OF THE CROATIAN COUNTIES ­ REGIONAL ECONOMY STRUCTURE ACCORDING TO SECTORS, ENTERPRISE SIZE AND OWNERSHIP This section analyses the relationship between the economic growth of certain counties on one hand, and the regional economy structure according to sectors, enterprise size and ownership on the other. The findings of this section research, in addition to the previously illustrated education and demographic indicators, as well as infrastructure capabilities indicators, provide a complete picture concerning the growth potential of certain Croatian counties. Economy Structure of Croatian Counties The experiences of the new EU member states show that the economy structure has a significant influence on the economy growth rates. The regions with a higher share of the tertiary sector (excluding the public sector) in the process of EU accession and transition into a market economy, are 104 experiencing higher rates of GDP growth. In regards to industry sector27 (industries C, D and E of NACE classification), the growth rates are dependant on the internal industrial structure (export orientation, technology transfer), rather than the total share of industry sector. The areas with a significant share of agriculture and public sector mainly experience slower growth in comparison to the average. This section analyses the relationship between the economy structure of the Croatian counties and the average rate of nominal growth in the period 2001-2003. The regional accounts system has started developing in Croatia only recently; therefore the data on the gross value added and GDP are available only in current prices. The real growth rates on the county level are not calculated at this point even for experimental purposes. Due to the differing regional economy structure not only in terms of sector distribution, but also the difference regional market conditions, the use of the national GDP deflator which does not reflect the differences in the levels and regional price trends in specific activities would not be justified. However, the analysis of the nominal GDP growth in various sectors brings us to some conclusion on the relationship between the economy structure and the growth rate. For analytical purposes, the activities have been grouped into five sectors. The primary sector includes agriculture (A) and fisheries (B). The secondary sector includes the processing industry with mining, manufacturing and electricity distribution (C, D and E). The tertiary sector has been separated into three sub-sectors. The first is comprised of services similar to manufacturing industry (construction, F) or activities that are closely related to product distribution (trade, G, transport and communications, I, and hotels and restaurants, H). The second sub-sector within the tertiary sector comprise financial services (J), business services (K), other personal services (O), and private households (P). The third service sub-sector in general comprises the government units, and includes public administration (L), education (M) and health (N). Table 4.15 shows the industry structure of gross value added in various counties in Croatia in the period 2001 - 2003. On the level of the total economy, the service sector (F, G, H and I) is dominant and comprises 33.3 percent of total value added. In all counties, apart from the Counties of Varazdin, Koprivnica- krizevci, Sisak-Moslavina and Meimurje, the proportion of this sector is the most significant. If financial, business, personal services and private households (industries J, K, O and P) are added, it is clear that the total tertiary sector (excluding public services) comprises 48.3 percent of total value added. As a rule, a higher proportion of the tertiary sector in the gross value added (GVA) is found in developed counties. In addition, some less developed counties close to the Adriatic Sea (for example, Counties of Zadar and Sibenik-Knin) have a high share of tertiary sector (tourism)28. The development level of various counties is well illustrated by the proportion of the primary sector (A and B) in a way that less developed counties have a higher proportion of the primary sector in total value added. Primarily, this refers to the counties in Eastern and Central Croatia. Significant differences between counties in GVA proportions have been recorded in industry sector (C, D and E). Therefore the smallest proportion, just above 10 percent has been recorded in the County of Zadar (13.1 percent) and Dubrovnik-Neretva (11.7 percent), while the highest proportions have been recorded in the County of Meimurje (35.7 percent), Koprivnica-Krizevci (35.5 percent) and Sisak-Moslavina (32.4 percent). The relationship between the share of this sector and total development levels are not clearly identifiable. 27 Industry sector comprises mining and quarrying (C), manufacturing (D) and electricity supply (E). 28 More detailed sectoral breakdown of value added is presented in Appendix 1. 105 Table 4.15: Regional Distribution of GVA, Current Prices in the Period 2001-2003, in % County of A, B C, D, E F, G, H, I J, K, O, P L, M, N TOTAL Zagreb 15.0 30.5 35.9 7.1 11.5 100.0 Krapina-Zagorje 12.8 30.1 30.4 7.1 19.6 100.0 Sisak-Moslavina 12.0 32.4 28.3 8.0 19.3 100.0 Karlovac 10.0 25.4 37.5 8.7 18.4 100.0 Varazdin 11.8 32.1 28.5 9.5 18.1 100.0 Koprivnica-Krizevci 21.8 35.5 22.6 6.9 13.2 100.0 Bjelovar-Bilogora 27.9 19.8 24.6 9.0 18.6 100.0 Primorje-Gorski kotar 2.3 23.5 41.7 14.8 17.6 100.0 Lika-Senj 17.0 19.0 37.2 6.4 20.3 100.0 Virovitica-Podravina 28.6 22.5 26.5 6.3 16.1 100.0 Pozega-Slavonia 21.9 21.5 25.3 6.9 24.4 100.0 Slavonski Brod-Posavina 18.7 21.0 28.9 10.0 21.4 100.0 Zadar 10.1 13.1 40.0 14.4 22.3 100.0 Osijek-Baranja 19.3 18.4 29.6 12.3 20.3 100.0 Sibenik-Knin 8.4 16.4 36.9 13.6 24.7 100.0 Vukovar-Sirmium 27.2 13.4 30.4 7.2 21.8 100.0 Split-Dalmatia 4.0 20.8 37.7 15.8 21.6 100.0 Istria 5.2 30.9 35.9 14.1 14.0 100.0 Dubrovnik-Neretva 8.4 11.7 39.9 17.3 22.7 100.0 Meimurje 15.8 35.7 23.6 10.0 14.8 100.0 City of Zagreb 0.4 25.2 32.7 23.2 18.4 100.0 Croatia 8.7 24.6 33.3 15.0 18.4 100.0 Source: Author's calculations. In continuation, Table 4.16 presents the average nominal GDP growth rates in the period 2001 - 2003 according to various sectors and counties. It is clear that the fastest growth has been recorded in the tertiary sector (except public administration). Thus, fastest growth, 19.0 percent, has been recorded in the business, financial and personal services. The sub- sector comprising of activities F, G, H and I in the analyzed period has increased at a very high average nominal rate of 16.9 percent. On the other hand, the nominal decrease in gross value added has been recorded in agriculture and fisheries (-2.4 percent). Slow nominal growth has been recorded in the public sector (L, M and N, 4.8 percent). In industry, the average nominal growth amounted to relative low 6.8 percent, but with significant differences across counties. Here, industrial production growth rates higher than 20 percent have been recorded in the Counties of Vukovar-Sirmium, Zagreb, Pozega-Slavonia, Lika-Senj and Sibenik- Knin, and on the other hand the largest decreases have been recorded in the County of Sisak- Moslavina (-7.5 percent) and Split-Dalmatia (-4.2 percent). It is evident that in industry sector the county structure of industrial manufacture is crucial since industry encompasses at the same time rapidly developing companies (publishing, part of capital products manufacture) but also decreasing traditional activities under the influence of the growing international competition (textiles, metal industry, etc.). 106 Table 4.16: Gross Value Added Average Nominal Growth Rate, Current Prices, in the Period 2001-2003, in % County of A, B C, D, E F, G, H, I J, K, O, P L, M, N TOTAL Zagreb 2.5 28.2 15.0 18.8 11.6 16.7 Krapina-Zagorje -4.4 8.4 5.9 15.5 2.8 5.3 Sisak-Moslavina -4.9 -7.5 19.2 4.2 6.9 3.5 Karlovac 3.6 10.2 0.4 15.8 2.3 4.9 Varazdin -1.5 0.0 26.2 20.7 6.3 9.6 Koprivnica-Krizevci -1.3 4.0 12.4 21.1 4.7 6.0 Bjelovar-Bilogora -1.7 4.2 18.3 22.2 2.9 7.0 Primorje-Gorski kotar -0.2 0.6 16.4 20.8 6.6 10.8 Lika-Senj -2.7 24.9 52.6 26.1 4.0 24.8 Virovitica-Podravina -4.0 10.5 16.2 16.8 2.3 6.7 Pozega-Slavonia -4.6 27.3 11.0 17.7 3.2 9.0 Slavonski Brod-Posavina -4.5 1.0 21.2 11.3 5.3 7.2 Zadar -2.0 13.7 31.9 17.5 6.5 17.8 Osijek-Baranja -3.2 5.7 19.5 16.8 3.9 8.7 Sibenik-Knin -2.6 28.0 22.0 18.2 5.8 15.9 Vukovar-Sirmium -5.6 33.7 16.4 13.6 5.4 9.5 Split-Dalmatia -5.2 -4.2 23.1 16.7 5.3 10.9 Istria 2.3 14.0 10.7 22.3 7.8 12.4 Dubrovnik-Neretva -2.7 1.9 15.2 19.0 3.1 9.9 Meimurje -4.6 8.4 17.0 18.5 5.0 8.6 City of Zagreb -0.3 6.0 15.3 19.9 3.3 11.5 Croatia TOTAL -2.4 6.8 16.9 19.0 4.8 10.6 Source: Author's calculations. The highest average nominal growth rate has been recorded in the County of Lika-Senj (24.8 percent). It should be noted that during the analyzed period, this county benefited from intensive motorway construction, which had positive impacts not only on the construction industry, but on other industries, either directly (construction material manufacture, transport, wholesale retailing) or indirectly through increased spending by the temporary labor force (retail, hotels and restaurants, personal and business services). However, as motorway construction moves towards the Adriatic South, it is to be expected that the positive effect will be transferred to the southern counties (Zadar, Sibenik-Knin, Split-Dalmatia), and that the County of Lika-Senj will experience the fate of the County of Karlovac which recorded very low growth rates after the finalization of the transport routes. Apart from the County of Lika-Senj, the other fastest growing counties were in central Dalmatia (Zadar 17.8 percent, Sibenik-Knin 15.9 percent), and the County of Zagreb (16.7 percent). In the observed period, the first two counties benefited, apart from motorway construction, from rapid tourism growth. In the case of the County of Zagreb, there is a favorable economic structure and a favorable position surrounding the capital city. The proximity of the City of Zagreb and the cost aspects (lower property prices, lower tax burdens) have influenced entrepreneurship growth which is growing more rapidly in the proximity of Zagreb than in the city itself. The three most developed counties (the City of Zagreb, the Counties of Istria and Primorje-Gorski kotar) with their favorable economic structure have secured high and stable growth rates above Croatian average (10.6 percent). The highest average growth rate in that group was recorded in Istria (12.4 percent), following with the City of Zagreb (11.5 percent), and Primorje-Gorski kotar (10.8 percent). However, if the temporary motorway construction effects (the County of Lika-Senj), and the low basis phenomenon (tourism in central and southern Dalmatia) are excluded, it is clear that the most developed counties will continue to increase the difference in the development level in comparison to the rest of Croatia. 107 On the other hand, the less favorable economic structure (relatively high proportion of agriculture and companies not yet restructured) in some counties of Central (Counties of Krapina-Zagorje, Sisak- Moslavina and Karlovac), and Eastern Croatia (primarily Counties of Virovitica-Podravina and Slavonski Brod-Posavina, but others as well) represent a significant risk of further lagging behind. Table 4.17 shows the contribution to the GDP growth across various activity groups in Croatia in the period 2001-2003. Table 4.17: Contribution to Total Gross Value Added Growth, Current Prices, in the Period 2001-2003, in % of Total Nominal GVA Increase County of A, B C, D, E F, G, H, I J, K, O, P L, M, N TOTAL Zagreb 2.29 48.31 32.93 8.02 8.45 100.00 Krapina-Zagorje -10.82 47.21 33.91 19.37 10.33 100.00 Sisak-Moslavina -17.03 -73.75 143.61 9.58 37.58 100.00 Karlovac 7.12 54.22 2.79 27.10 8.77 100.00 Varazdin -1.91 1.63 71.40 19.84 12.30 100.00 Koprivnica-Krizevci -4.89 24.21 46.96 23.12 10.59 100.00 Bjelovar-Bilogora -6.96 12.27 59.85 26.98 7.87 100.00 Primorje-Gorski kotar -0.04 1.37 61.27 26.56 10.84 100.00 Lika-Senj -2.00 18.18 73.87 6.42 3.53 100.00 Virovitica-Podravina -17.33 36.20 60.62 14.89 5.62 100.00 Pozega-Slavonia -11.45 57.31 32.24 12.93 8.97 100.00 Slavonski Brod-Posavina -12.20 3.05 78.45 14.88 15.83 100.00 Zadar -1.20 10.83 67.96 13.89 8.52 100.00 Osijek-Baranja -7.53 12.30 62.77 23.16 9.29 100.00 Sibenik-Knin -1.43 25.67 50.78 15.45 9.53 100.00 Vukovar-Sirmium -16.92 43.15 51.14 10.03 12.61 100.00 Split-Dalmatia -2.03 -8.32 76.11 23.58 10.66 100.00 Istria 0.98 33.45 32.32 24.27 8.97 100.00 Dubrovnik-Neretva -2.32 2.35 61.23 31.47 7.28 100.00 Meimurje -8.87 34.98 44.86 20.10 8.92 100.00 City of Zagreb -0.01 13.76 42.32 38.44 5.49 100.00 Croatia -2.04 16.12 51.61 25.80 8.52 100.00 Source: Author's calculations. It is clear that more than 50 percent of nominal gross value added increases originates from the growth of the sector which includes construction, retail, hotels and restaurants, and transport and communications (F, G, H and I). In all counties, except for Counties of Zagreb, Pozega-Slavonia and Karlovac, the growth contribution of this sector is the most significant. If financial, business and personal services and private households are added (industries J, K, O and P), it is clear that the total tertiary sector (except public administration) comprises more than 75 percent of the growth of total gross value added. The negative contribution to the nominal increases to value added stems from the primary sector (A and B). This sector positively influences the increases of value added only in Counties of Karlovac, Istria and Zagreb. In terms of the secondary sector (C, D and E), it is clear that there are significant differences between various counties. On one side, there are counties with a favorable industry sector structure and a strong growth potential, while on the other side, there are non-restructured industrial manufacturers in certain counties even creating nominal reductions in total industrial manufacture. The high contribution of industrial manufacture to the total increases of gross value added has been recorded in the counties of Zagreb, Pozega-Slavonia, Karlovac, Krapina-Zagorje and Vukovar-Sirmium. It should not be forgotten that in certain counties this is a result of low base phenomenon. On the other hand, the significant nominal reduction of gross value added in the industry sector has been recorded in the Counties of Sisak-Moslavina and Split-Dalmatia. 108 In the period observed, the sector containing the public units activities (L, M and N) has recorded low nominal growth (on the level of total economy even real decreases), and thus total nominal growth contributions are low. Clear demonstration of the relationship between the economy structure and the growth potential is given by Figure 4.2. It can be seen that the counties with a significant proportion of the tertiary sector (except public administration) have recorded on average higher nominal gross value added growth rates. The highest proportion of the tertiary sector (higher than 50 percent) has been recorded in the three most developed counties (the City of Zagreb, Istria and Primorje-Gorski kotar counties) and the counties in Adriatic South (Zadar, Split-Dalmatia, Sibenik-Knin and Dubrovnik-Neretva). On the other hand, the lowest proportion of the tertiary sector (with a high proportion of the primary sector) has been recorded in the counties of Eastern and Central Croatia which are underdeveloped. Such economy structure in the middle run within the EU accession process could contribute to the widening of the inequalities among the Croatian counties. Figure 4.2: The Relationship between the Proportion of Tertiary Sector (except public administration) in gross value added and the nominal growth index in the period 2001-2003 60,0 55,0 Services, in % of GVA 50,0 45,0 40,0 35,0 30,0 25,0 20,0 0,0 5,0 10,0 15,0 20,0 25,0 30,0 Nominal growth rate In continuation, other factors influencing the differences in growth rates of the Croatian counties are investigated. Primarily, this refers to the effects of small and medium entrepreneurship growth and the differences in the ownership structure. Economy Growth and Enterprise Size According to Counties This section analyses the economy structure of certain Croatian counties according to enterprise size. Economy structure data, according to the unit size are not officially calculated and published in Croatia even on the level of the national economy. Therefore, the author's estimates based on various information sources are presented.29 For analytical purposes, the estimation of the economy structure has been formulated according to size and separated into four sectors. These are the small and medium enterprise sector, large entrepreneurs sector, financial institutions sector and the general government administration units sector. The small and medium enterprise sector has been separated into two sub-sectors: small and medium market producers (crafts, small and medium legal units), and 29 Primarily, this is the data set covering the FINA (financial agency) surveys, used in the official GDP calculations. The authors would like to thank the colleagues from DZS, and particularly Mrs. Maja Gorjan Breges for the help and assistance in data processing for the purposes of this research. 109 non-market producers (owner occupied dwellings and individual agricultural producers). Table 4.18 shows the average proportions of various sectors in the period 2001 - 2003. Table 4.18: Average Proportions in the GVA of Various Unit Groups According to Size in the Period 2001-2003 SME County of Large entrepreneurs Financial institutions Government units Total Market Non-market Zagreb 42.5 16.7 28.4 2.2 10.2 100.0 Krapina- 35.9 15.8 28.1 3.1 17.1 100.0 Zagorje Sisak- 19.7 15.5 43.7 3.2 17.9 100.0 Moslavina Karlovac 30.2 13.0 36.9 3.1 16.8 100.0 Varazdin 34.9 13.3 32.0 3.6 16.2 100.0 Koprivnica- 18.0 24.2 42.9 3.2 11.7 100.0 Krizevci Bjelovar- 29.0 27.3 22.6 4.2 16.9 100.0 Bilogora Primorje- 39.4 6.2 33.7 4.6 16.1 100.0 Gorski kotar Lika-Senj 20.7 15.9 42.1 2.7 18.6 100.0 Virovitica- 18.1 25.0 39.0 3.1 14.7 100.0 Podravina Pozega- 22.1 19.8 32.5 3.1 22.5 100.0 Slavonia Slavonski 30.8 20.2 25.6 3.7 19.7 100.0 Brod- Posavina Zadar 35.0 11.2 28.5 4.8 20.5 100.0 Osijek- 24.8 15.5 36.8 4.3 18.6 100.0 Baranja Sibenik- 30.0 12.6 29.4 5.1 22.9 100.0 Knin Vukovar- 25.5 24.1 27.9 2.3 20.1 100.0 Sirmium Split- 41.9 7.6 25.8 4.6 20.1 100.0 Dalmatia Istria 39.9 8.8 34.1 4.5 12.8 100.0 Dubrovnik- 36.0 12.2 24.6 6.0 21.2 100.0 Neretva Meimurje 40.0 16.8 26.3 3.8 13.1 100.0 City of 35.0 5.1 35.0 7.9 16.9 100.0 Zagreb Croatia 33.8 11.2 32.9 5.1 16.9 100.0 Source: Author's calculations. The largest proportion of GVA is created by the small and medium market producers (33.8 percent). This is followed by the large entrepreneurs sector with 32.9 percent, and government units with 16.9 percent proportion. The smallest proportion, amounting to 5.1 percent in total gross value added, is recorded for financial institutions (banks and insurance) and the non market small producers (11.2 percent). Market oriented small and medium entrepreneurs have the most significant proportion in the County of Zagreb (42.5 percent), followed by the County of Split-Dalmatia (41.9 percent), Meimurje (40.0 percent) and Istria (39.9 percent). On the other hand, the smallest proportion of this group has been recorded in the County of Koprivnica-Krizevci (18.0 percent), Virovitica-Podravina (18.1 percent), and Sisak-Moslavina (19.7 percent). The low share of this sector (below 30 percent) has been 110 recorded in almost all of the counties in Eastern Croatia. All three most developed counties (The City of Zagreb, Counties of Istria and Primorje-Gorski kotar) have above-average share of this sector. The differences in non-market small producers' proportions are mainly due to the differences in the proportions of agriculture producers, considering that the proportion of household owners (imputed dwelling rent) does not vary significantly across counties. Thus, the highest proportions have been noted in the Eastern and Central Croatia (Counties of Bjelovar-Bilogora, 27.3 percent, Virovitica- Podravina, 25.0 percent, Vukovar-Sirmium, 24.1 percent and Koprivnica-Krizevci, 24.2 percent). Expectedly, the lowest share of this sector has been recorded in the three most developed counties. Due to the fact that the large enterprises includes public companies (INA, HEP, HP, HT) which are equally active in all parts of Croatia, low levels of proportional variation have been recorded in comparison to the small manufacturers. However, depending on the geographical position of the other large entrepreneurs (private or state owned) certain differences do exists. Thus, the smallest proportion of large entrepreneurs has been estimated in the County of Bjelovar-Bilogora (22.6 percent) and the largest for the County Sisak-Moslavina (43.7 percent). On the other hand, significant differences in the county proportions in the GVA have been registered in the financial sector (banks and insurance). The economic development differences, business profitability but also disposable income of households has a significant influence of the regional distribution of the financial services covering households and companies. Thus the highest proportions of this sector have been recorded in the City of Zagreb (7.9 percent) and the smallest in the County of Zagreb (2.2 percent). In this example, suggested explanation is that the business branches of the financial institutions in the City of Zagreb simultaneously serve households and entrepreneurs from the surrounding counties as well. In principle, it can be stated that there is a strong positive relationship between the proportion of the financial institutions sector and the development of a certain county. In addition, the counties in the costal regions have even higher financial sector proportion than indicated by their development levels. An explanation is that financial institutions try to service the needs of the domicile units and foreign tourists as well and therefore collect the foreign exchange deposits originating from tourism income. In terms of government administration units, it should be noted that the relative differences between the proportions of this sector in the GVA in certain counties are relatively smaller in comparison to the case of small entrepreneurs and financial institutions. Table 4.19 shows the average nominal growth rates of various sectors in the period 2001 - 2003. It is clear that on the total economy level, the highest nominal growth rate has been recorded in the financial institutions sector (19.2 percent), followed by sector of market small and medium enterprises (13.6 percent). The financial institutions sector has recorded double-figure nominal average growth rates in all counties except the Counties of Sisak-Moslavina (7.4 percent) and Slavonski Brod- Posavina (8.3 percent). The highest growth rate of this sector has been found in Counties of Split- Dalmatia (27.9 percent), Bjelovar-Bilogora (21.9 percent), and the City of Zagreb (21.8 percent), which also has the highest proportion of gross value added in this sector. The small market enterprises in the analyzed period have been experiencing stable high growth rates. The highest nominal growth (50.0 percent) has been recorded in County of Lika-Senj where it is evident that motorway construction has stimulated additional activities, with small and medium sector as the most flexible in that respect. According to growth rates, the County of Lika-Senj is followed by Counties of Sibenik-Knin (23.6 percent) and Vukovar-Sirmium (23.3 percent). Small and medium entrepreneurs generated the least momentum in the County Virovitica-Podravina even with average nominal reductions, amounting to 0.6 percent annually. Low growth rates in this sector have also been recorded in Counties of Osijek-Baranja (4.5 percent) and Karlovac (7.2 percent). In the analysis of these indicators, the above average presence of the underground economy in the SME sector should be expected. Therefore the reported value added and hence the increases are most likely underestimated. The counties with a significant proportion of areas of special state concern with 111 significant tax incentives have recorded the highest value added growth rates in this sector (Counties of Lika-Senj, Vukovar-Sirmium and Sibenik-Knin). Table 4.19: Average Nominal GVA Growth Rates in the Period 2001 - 2003, in % SME Large Financial Government County of Non- Total Market entrepreneurs institutions units market Zagreb 17.6 2.6 26.0 18.1 10.8 16.4 Krapina-Zagorje 12.7 -7.3 4.7 12.3 1.1 5.1 Sisak-Moslavina 17.7 1.1 -2.4 7.4 6.1 3.5 Karlovac 7.2 4.5 4.8 13.2 0.4 5.0 Varazdin 9.1 -1.6 16.3 16.8 5.0 9.5 Koprivnica-Krizevci 8.8 0.6 7.1 21.7 4.0 5.9 Bjelovar-Bilogora 10.6 2.0 10.0 21.9 2.0 7.0 Primorje-Gorski kotar 12.9 1.4 11.0 15.3 5.5 10.4 Lika-Senj 50.0 1.0 34.3 14.6 3.6 24.1 Virovitica-Podravina -0.6 -0.4 16.1 17.4 1.5 6.6 Pozega-Slavonia 8.8 -6.6 23.5 16.7 2.1 8.7 Slavonski Brod- 11.1 -2.6 12.0 8.3 4.4 7.0 Posavina Zadar 11.8 -7.4 47.7 16.9 5.7 17.1 Osijek-Baranja 4.5 3.5 16.7 11.6 2.9 8.6 Sibenik-Knin 23.6 4.9 21.1 17.2 4.7 15.4 Vukovar-Sirmium 23.3 -2.3 11.4 16.3 4.4 9.3 Split-Dalmatia 16.3 -2.9 8.2 27.9 4.4 10.6 Istria 10.3 -0.9 18.5 17.8 6.8 11.9 Dubrovnik-Neretva 20.3 -3.5 7.3 15.9 2.1 9.6 Meimurje 7.8 -2.1 19.9 14.1 2.9 8.4 City of Zagreb 14.9 4.8 11.2 21.8 1.6 11.2 Croatia 13.6 0.3 12.9 19.2 3.5 10.4 Source: Author's calculations. Above average nominal growth rate has also been recorded in the large entrepreneurs sector. On one hand, the growth of income of large entrepreneurs is a result of the infrastructure cycle which is mainly undertaken by large entrepreneurs (construction), but also consolidation and value added growth rate (price liberalization) in the case of public companies (INA, HEP, HT). On the other hand, the counties with a significant proportion of the large, unstructured manufacturers (textiles industry, metal processing, etc.) have recorded low growth rates or even decreases of gross value added in the case of large entrepreneurs. Consequently, the largest value added increases among large entrepreneurs have been noted in the County of Zadar (47.7 percent) and the County of Lika-Senj (34.3 percent). Large entrepreneurs have demonstrated the worst results in the County of Sisak- Moslavina (nominal reduction of 2.4 percent), and low nominal growth rates were also recorded in Counties of Krapina-Zagorje (4.7 percent), Karlovac (4.8 percent), and Koprivnica-krizevci (7.1 percent). Slow nominal growth (real reduction) on total economy level has been noted in the government administration units sector, and the small non-market manufacturers sector, where agriculture is the most significant. Such trends in these sectors are not unexpected. In the analyzed period, the bulk of the structural adjustments involved the consolidation of public finances, which explains the gross value added trends in the government administration units sector. The expected public sector reform, as well as further transition process towards market economy will definitely influence the slow growth of the government sector in the following period. In terms of non-market manufacturers, reduction of their proportions is to be expected, and partly a transformation into competitive market manufacturers. 112 From the above mentioned, it can be concluded that there is a positive relationship between the proportion of the market small and medium entrepreneurs in total gross value added and economy growth rates (see Figure 4.3). Financial institutions sector contributed to the rapid growth in the analyzed period, while counties with a higher proportion of government administration units and non- market producers have recorded lower rates of growth on average. Figure 4.3: Relationship between he Average Nominal Growth rate of the Total GVA and the proportion of Market Small and Middle Entrepreneurs in Total Gross Value Added 45,0 Market SME, in % of total GVA 40,0 35,0 30,0 25,0 20,0 15,0 0,0 5,0 10,0 15,0 20,0 25,0 30,0 Avarage nominal growth rate of total GVA The Economic Growth and Ownership Structure According to Counties This section analyses the influence of ownership structure on the growth at the county level. The data concerning the ownership structure are not officially published; therefore the data used in this analysis represents the author's estimates based on available data sources. For analytical purposes, the ownership structure has been classified into four sectors. These are the private ownership sector, public ownership sector, mixed ownership sector with the majority private stake and the mixed ownership sector with the majority public stake. Private property also includes the crafts sector, property owners and residents, privately-owned companies, and private financial institutions. The analysis used the data from FINA research, and since there is no statistical registry in Croatia which contains updated data concerning the ownership structure, this estimate should be considered as estimation only and not entirely accurate. Apart from the lack of a satisfactory statistical registry, an additional problem should be mentioned, concerning the related companies where it is very difficult to establish the final ownership structure. This is especially important in cases relating to state property where the registered owners are often various units owned by the government. In addition, the unsolved cases of property returns present further problems as well. Therefore, it remains questionable to what extent do the units filling in the statistical forms actually have full knowledge of the complete ownership structure. However, such estimates enable an insight into the relative importance of the private and the state property in various counties. Table 4.20 shows the estimated proportions of private and state ownership shown as the proportion of total gross value added in a certain county. The total proportion of private property on the total national economy level has been estimated at 69.0 percent, while the proportion of gross value added created by the state owned units in the same period amounted to 31.0 percent. The largest proportion of private property has been estimated in the County of Meimurje, 79.1 percent in the period 2001 - 113 2003. The County of Zagreb had a slightly smaller proportion, 78.9 percent, followed by the County of Koprivnica-Krizevci with 78.6 percent. On the other hand, the smallest proportion of private property has been estimated in the county of Lika-Senj (49.6 percent), followed by the Counties of Sibenik-Knin (59.2 percent), Osijek-Baranja (66.0 percent) and Pozega-Slavonia (66.1 percent). Table 4.20: Estimated GVA Proportion According to the Ownership Structure in the Period 2001-2003, in % of Total GVA for the County Private property State property Mixed, County of Mixed, mainly Full mainly Total Full Total state private Zagreb 73.5 5.4 78.9 20.0 1.1 21.1 Krapina-Zagorje 66.6 9.0 75.6 23.1 1.2 24.4 Sisak-Moslavina 48.2 20.4 68.6 28.4 3.0 31.4 Karlovac 63.2 9.1 72.4 25.3 2.3 27.6 Varazdin 63.0 10.4 73.4 24.6 2.0 26.6 Koprivnica-Krizevci 55.5 23.1 78.6 19.9 1.5 21.4 Bjelovar-Bilogora 69.1 6.1 75.2 24.3 0.6 24.8 Primorje-Gorski kotar 61.2 6.9 68.1 28.9 3.0 31.9 Lika-Senj 44.1 5.6 49.6 48.5 1.8 50.4 Virovitica-Podravina 59.1 14.9 74.0 22.5 3.5 26.0 Pozega-Slavonia 58.9 7.3 66.1 29.8 4.1 33.9 Slavonski Brod-Posavina 59.3 7.7 66.9 28.2 4.9 33.1 Zadar 64.0 6.1 70.1 29.1 0.9 29.9 Osijek-Baranja 57.4 8.6 66.0 31.2 2.8 34.0 Sibenik-Knin 51.0 8.2 59.2 32.8 8.0 40.8 Vukovar-Sirmium 56.9 7.2 64.1 31.3 4.6 35.9 Split-Dalmatia 61.1 6.8 67.9 28.8 3.2 32.1 Istria 64.5 10.3 74.8 21.8 3.4 25.2 Dubrovnik-Neretva 52.8 9.4 62.2 31.2 6.6 37.8 Meimurje 72.9 6.2 79.1 19.3 1.6 20.9 City of Zagreb 56.1 9.8 65.9 32.3 1.7 34.1 Croatia 59.6 9.4 69.0 28.5 2.5 31.0 Source: Author's calculations. Table 4.21 shows the average annual nominal growth of gross value added according to the ownership structure in the period 2001 - 2003. It is clear that gross value added in the units with private owners (12.5 percent) grew at a much faster rate in comparison to the state owned units (5.7 percent). However, the interpretation of these estimates should be read with caution since there is a distinct possibility that a part of the units initially owned by the state have transferred ownership into private hands within the privatization process. In addition, it is possible that is some cases the failure to meet the payments to the government results in the government increasing its ownership share. In that way, some proportion of the difference in the growth rate most probably originates from the statistical reclassification. However, it can be concluded that in all counties, apart from the County of Zagreb, the gross value added of the units in private property is growing at a faster rate in comparison to the units in state ownership. 114 Table 4.21: Average Annual Nominal GVA Growth According to the Ownership Structure in the Period 2001-2003 Private ownership Government ownership County of Mixed, mainly Mixed, mainly Full TOTAL Full TOTAL private state Zagreb 17.9 -14.0 15.4 19.1 54.5 20.5 Krapina-Zagorje 14.4 -34.7 6.7 0.6 0.0 0.6 Sisak-Moslavina 7.9 4.6 6.9 2.2 -41.2 -3.2 Karlovac 7.4 4.1 7.0 0.1 -3.2 -0.1 Varazdin 12.9 20.0 13.7 0.6 -19.4 -1.1 Koprivnica-Krizevci 10.0 -1.9 6.2 4.4 10.4 4.8 Bjelovar-Bilogora 10.0 -10.3 8.2 2.9 27.4 3.3 Primorje-Gorski kotar 14.0 -5.3 11.9 7.4 6.6 7.3 Lika-Senj 31.0 19.5 29.5 19.2 19.9 19.3 Virovitica-Podravina 16.1 -4.8 11.5 4.9 -57.2 -6.0 Pozega-Slavonia 11.1 4.9 10.4 5.6 6.0 5.6 Slavonski Brod-Posavina 10.5 -15.2 7.3 4.7 16.4 6.4 Zadar 20.0 17.8 19.8 10.9 22.4 11.2 Osijek-Baranja 11.0 -0.6 9.4 6.9 8.4 7.0 Sibenik-Knin 22.8 8.3 20.8 7.5 11.6 8.2 Vukovar-Sirmium 11.4 5.8 10.7 1.0 51.8 7.0 Split-Dalmatia 17.8 -9.6 14.7 0.4 25.1 2.7 Istria 14.3 0.7 12.2 3.5 112.1 11.0 Dubrovnik-Neretva 16.1 16.2 16.1 0.9 -5.4 -0.2 Meimurje 11.2 -3.4 10.0 3.7 -6.3 2.8 City of Zagreb 18.8 -10.6 14.0 5.6 14.1 6.0 Croatia 15.4 -4.4 12.5 5.2 11.2 5.7 Source: Author's calculations. The highest growth rate of private sector is recorded in the County of Lika-Senj with average annual nominal growth of 29.5 percent. It is followed by the County of Sibenik-Knin (20.8 percent) and the County of Zadar (19.8 percent). The private sector exhibited the slowest growth in the Counties of Krapina-Zagorje (6.7 percent) and Koprivnica-krizevci (6.2 percent). In the case of the ownership structure it can be concluded that faster growth of private ownership in comparison to state ownership can be expected, but it cannot be concluded that higher share of the private sector in itself ensures higher growth rates. Figure 4.4 shows that for instance, 15 percent growth rates are simultaneously achieved by the County of Zagreb with the private sector share of almost 80 percent and the County of Sibenik-Knin with the proportion of private sector lower than 60 percent. Figure 4.4: The Relationship between the Gross Value Added Growth Rate and Private Sector Share 30 Avarage nom growth rate of GVA 25 20 15 inal 10 5 0 45,0 50,0 55,0 60,0 65,0 70,0 75,0 80,0 85,0 Private sector, in % GVA 115 In conclusion, in terms of the structural characteristics of the Croatian counties, it can be stated that economic growth primarily depends on the economy structure according to the activities in a way that favorable economy structure, with a high share of the tertiary sector (except public administration) ensures the generation of high economic growth. Growth is significantly stimulated by propulsive small and medium market oriented entrepreneurs, and not own-account producers. A higher proportion of private property in the analyzed period has not shown itself to be the factor which individually ensures high growth rates. The probable cause is the fact that within private property there are two basic divisions, privatized and new-private. The majority of the research so far discusses the problems with a part of the privatized property (tycoonisation), while the situation is significantly better in the case of new property. However, on the aggregate level there are no clear conclusions. In terms of the structural characteristics the following can be concluded: ˇ The most developed counties, the City of Zagreb, the County of Istria and Primorje-Gorski kotar according to the economy structure characterised by a high proportion of the tertiary sector and a relatively propulsive small and medium entrepreneurs' sector, as well as a solid group of large entrepreneurs in both private and state ownership, have above average growth potential which could lead to further increases of regional development inequalities. ˇ The large infrastructure projects contributes significantly to county growth where these are implemented, but the largest direct and indirect impacts are felt during the period wherein these projects are underway, followed by expectedly much slower growth rates. The best examples for this claim are the Counties of Karlovac and Lika-Senj. In the observed period, the County of Karlovac was in the final part of the infrastructural motorway construction cycle, and exhibited very low nominal increases, while motorway construction through the County of Lika-Senj was at the most intensive stage, which corresponds with the highest growth rates in this county. Therefore the effects of the projects can be considered to be temporary only. ˇ The counties in Adriatic Croatia, in addition to the propulsive small and medium enterprises sector, as well as demographic and educational structure, also have extremely favorable economic structure which ensures above average growth in the future. However, it should be noted that the growth of the gross value added and the development of the SME sector is largely linked with the extremely positive tourism results in Croatia in the analyzed period. This also represents a certain level of risk in light of changing consumer preferences, but also this is a source of threat as tourism is volatile on a global level, therefore the economies of these areas are potentially in jeopardy. ˇ The county which definitely has an enormous growth potential is the County of Zagreb, characterized by a favorable economic structure and extremely propulsive SME sector. In the analyzed period, this county was only in the initial phases of complete exploitation of its geographical position as the "ring" surrounding the capital city. The relatively favorable cost aspects (property values, commercial spaces, housing, and lower taxes) as well as the proximity of the City of Zagreb make this area extremely attractive for entrepreneurs. ˇ The analytical regions Central and Eastern Croatia are exposed to the largest risk of further lagging behind the most developed counties. The reasons for the smaller growth potential are relatively unfavorable economic structure (with high agriculture and non-market manufacturers share), war-related damages, and the slow developments in the process of restructuring the large companies (Table 4.22). 116 Table 4.22: Average Annual Growth Rates: GDP p.c. (2001-03), Population (2001-03) and Employment (2001-2003) County of GDP Population Employment Counties with high convergence potential Zadar 17.8 1.0 2.4 Zagreb 16.7 0.9 3.2 Sibenik-Knin 15.9 0.2 2.1 City of Zagreb 11.5 0.1 2.6 Split-Dalmatia 10.9 0.7 3.4 Istria 12.4 0.5 2.0 Counties with moderate convergence potential Lika-Senj 24.8 -0.6 10.4 Varazdin 9.6 -0.4 3.2 Dubrovnik-Neretva 9.9 0.3 -0.4 Primorje-Gorski kotar 10.8 -0.1 1.8 Meimurje 8.6 0.0 3.1 Counties with moderate divergence risk Vukovar-Sirmium 9.5 -0.6 1.9 Osijek-Baranja 8.7 -0.2 0.9 Pozega-Slavonia 9.0 -0.3 1.2 Bjelovar-Bilogora 7.0 -0.8 0.6 Koprivnica-Krizevci 6.0 -0.5 1.4 Counties with high divergence risk Slavonski Brod-Posavina 7.2 -0.2 0.4 Virovitica-Podravina 6.7 -0.6 -1.9 Karlovac 4.9 -0.9 0.4 Krapina-Zagorje 5.3 -0.6 1.1 Sisak-Moslavina 3.5 -0.7 0.2 Source: CBS, author's estimates. Underground Economy and Regional Development This section shows the results of the underground economy calculations (lower estimation boundary ­ Eurostat approach) in various counties in Croatia. Considering the fact that the regional GDP calculations are relatively new, therefore the data available only refers to the period 2001-2003, the underground economy estimation refers to this period as well. The results are given by Table 4.23. The methodological framework for the calculation of the underground economy on county level is influenced by the data availability. On one hand, there is data concerning regional gross value added and regional GDP separated according to activities, while on the other hand there is data describing the proportion of the underground economy according to activities, but only on the national level. It is for that reason that the regional division of the underground economy can only be executed by applying the proportions according to activity on a national level to all of the counties in Croatia. Figure 4.5 shows the size of the underground economy in various counties, i.e. the proportion of the underground economy in the GDP of various Croatian counties. Figure 4.5: Underground Economy Size According to County, 2002 (Eurostat Approach) 18,0 17,0 16,0 15,0 14,0 13,0 12,0 11,0 Vukovar-Sirmium Sisak-Moslavina Bjelovar-Bilogora Virovitica-Podravina Slavonski Brod-Posavina Split-Dalmatia Primorje-Gorski kotar Sibenik-Knin Koprivnica-Krizevci Lika-Senj Osijek-Baranja Meimurje Varazdin Pozega-Slavonia Dubrovnik-Neretva City of Zagreb Karlovac Krapina-Zagorje Istria Croatia Zagreb Zadar 117 Table 4.23: Underground Economy Size in Various Counties in Croatia (lower estimation boundary ­ Eurostat approach), 2002, in HRK thousands Methodological Underground Total Correction percentage changes to the Total Correction County of Official GDP economy (N1- corrected including housing housing rent corrections percentage, in % N7) GDP rent, as % of GDP calculations Zagreb 9 839 1 128 457 1 584 11 423 11.5 16.1 Krapina-Zagorje 4 305 441 179 620 4 925 10.2 14.4 Sisak-Moslavina 6 097 643 293 936 7 033 10.6 15.4 Karlovac 4 895 584 228 812 5 707 11.9 16.6 Varazdin 7 371 780 318 1 098 8 469 10.6 14.9 Koprivnica-Krizevci 5 146 522 257 778 5 924 10.1 15.1 Bjelovar-Bilogora 4 296 433 210 643 4 939 10.1 15.0 Primorje-Gorski kotar 14 021 1 696 695 2 391 16 412 12.1 17.1 Lika-Senj 1 974 187 95 281 2 255 9.5 14.2 Virovitica-Podravina 2 955 296 138 434 3 389 10.0 14.7 Pozega-Slavonia 2 490 231 112 342 2 832 9.3 13.8 Slavonski Brod-Posavina 4 332 432 187 619 4 951 10.0 14.3 Zadar 4 916 569 237 806 5 722 11.6 16.4 Osijek-Baranja 10 777 1 119 517 1 636 12 413 10.4 15.2 Sibenik-Knin 3 043 306 140 446 3 489 10.0 14.7 Vukovar-Sirmium 4 847 471 206 677 5 524 9.7 14.0 Split-Dalmatia 14 350 1 536 585 2 122 16 472 10.7 14.8 Istria 11 481 1 446 554 2 000 13 481 12.6 17.4 Dubrovnik-Neretva 4 379 452 195 647 5 026 10.3 14.8 Meimurje 4 107 440 197 637 4 744 10.7 15.5 City of Zagreb 55 610 6 392 2 442 8 834 64 444 11.5 15.9 Croatia 181 231 20 102 8 242 28 344 209 575 11.1 15.6 Source: Author's calculations. 118 In relation to the 2002 Croatian average of 15.6 percent of the underground economy in the total GDP, above average proportions of the underground economy have been recorded in the Counties of Istria (17.4 percent), Primorje-Gorski kotar (17.1 percent), Karlovac (16.6 percent), Zadar (16.4 percent), Zagreb (16.1 percent) and the City of Zagreb (15.9 percent). All other counties have recorded below average proportions of the underground economy in the official GDP. The lowest proportion has been observed in the Counties of Pozega-Slavonia (13.8 percent) and Vukovar-Sirmium (14.0 percent). In continuation, we show the results of the correlation analysis which aimed to establish the factors influencing or linking the county differences in the underground economy size. These factors have been grouped in various groups. The first is the demographic structure of the counties. The proportions of various age categories have been taken for each county, i.e. children (population younger than 14 years of age), working contingent (female population between 15 and 59 years of age and male population between 15 and 64) and senior population (females older than 60, males older than 65). The second group of factors is the educational population structure. Depending on the education level, the population in each county has been separated into those with no education and with finished primary education, with finished secondary education and those with finished higher education programs and universities and masters levels, called tertiary education. The third group of factors is the economic structure of the economies in the counties. The economic structure is comprised on the basis of the proportions of employment in various activities in the total employment figures, and contains the employment in the primary (agriculture and fisheries, A+B), secondary (industry, C+D+E) and the tertiary sectors (services, G+H+I+J+K+L+M+N+O+P). The tertiary sector has been additionally separated into the group containing the public employees, those employed in education and health services, and others in the other group. The county differences in the size of the underground economy have been examined in comparison with some other factors. Primarily these are the total tax incomes of the local authorities, as the approximation of the tax burden across counties. Another analysis involved examining the link between the underground economy according to counties and the development levels (measured by the regional GDP), and the county differences in the unemployment levels. The correlation analysis results are given in Table 4.24. Table 4.24: The Correlation Coefficients between Certain Variables and County Proportions of the Underground Economy in the GDP Figure Variable Correlation coefficient Demographic structure Children -0.53* Working contingent 0.67* Seniors -0.22 Education structure Lower qualification level -0.75* Middle qualification level 0.76* High and higher qualification level 0.65* Economic structure Employment ratio in the primary sector -0.70* Employment ratio in the secondary sector -0.08 Employment ratio in the tertiary sector 0.56* Employment ratio in the tertiary sector, except government services 0.66* Employment ratio in the tertiary sector, except tourism and governmnt. services 0.69* Total local authority tax income as a % of GVA 0.71* Per capita GDP 0.55* Note: * 5 percent significance level. Source: Author's calculations. 119 Regarding demographic structure, statistically significant relationship can be found in the case of variables children and the working contingent. With the educational structure of the Croatian counties, all of the variables have shown a statistically significant and strong relationship with the county differences in the size of the underground economy. The middle, high and the highest qualifications have positive signs, indicating that on average, the Croatian counties with a more favorable education structure have higher levels of the underground economy. The most interesting results refer to the relationship between the underground economy and the economy structure of certain Croatian counties. The counties with higher ratios of population employed in the primary sector, on average exhibit lower proportions of the underground economy. The relationship between employment in the secondary sector and the size of the underground economy is weak and not statistically significant. However, the relationship between the tertiary sector and the size of the underground economy is significant and medium strong positive sign. This implies that the counties with higher shares of the service sector have on average higher proportions of the underground economy. This relationship is strengthened if the scope of the tertiary sector excludes the public services, education and health (correlation coefficient 0.66), and if in addition, one excludes tourism as well (0.69). The tax income is also statistically significantly correlated with the size of the underground economy in various Croatian counties. The correlation coefficient is 0.71 with a positive sign, confirming the theoretical assumption that one of the main reasons for the growth of the underground economy is the increases in the tax burden. The more developed counties on average have higher proportions of the underground economy. The correlation coefficient between the county development level, measured by the per capita GDP and the size of the underground economy is 0.55. SECONDARY DISTRIBUTION OF INCOME Tables 4.25, 4.26 and 4.27 show gross disposable income (GDI), primary income, and the disposable income and GDP ratios per county and regions.30 These indicators are significant for the identification of the sectoral distribution of regional value added. The influence of the redistribution policies is also presented. As can be seen from Table 4.25 process of income redistribution significantly reduces the difference measured by GDP p.c. between the most developed county (City of Zagreb) and the least developed counties (County of Vukovar-Sirmium in 2001 and 2002, County of Slavonski Brod ­ Posavina in 2003). The ratio between the most and the least developed county in terms of GDP p.c. was 3.04 (2001) to 3.08 (2003), and in terms of household GDI p.c. the ratio was 1.59 (2001) to 1.69 (2003). The higher ratio between the disposable and primary income of the household sector indicates higher net income transfers to the region (increased by the social and other transfers and reduced by taxes on income and social security contributions) within the secondary income distribution process. In the Croatian economy, the highest indicator has been recorded in Counties of Vukovar-Sirmium (117.9 in 2001, 116.0 in 2002 and 115.1 in 2003), Slavonski Brod-Posavina (114.8, 112.9 and 113.0), Lika-Senj (113.4, 111.6 and 106.6), and Sibenik-Knin (116.9, 113.4 and 112.6). This is an 30 Primary income records gross wages and salaries, gross operating surplus (consumption of fixed capital included), mixed income and property income. Disposable income is derived from primary income after adding and subtracting all secondary distribution transaction (social transfers, other transfers, taxes and social security contributions). More about the concept of primary and disposable income see in ESA 1995 or SNA 1993. 120 expected result, since these are the less developed counties, but also counties with the highest proportion of the population in areas of special state concern status. On the other hand, the highest net provider is the City of Zagreb. The lowest ratio between the disposable income and primary income has been recorded in the City of Zagreb (87.3 in 2003), followed by Counties of Zagreb (91.3), Istria (92.6) and Primorje-Gorski kotar (95.9). Table 4.25: Gross Disposable Income of Households per capita, by county, Croatia =100 County of 2001 2002 2003 2002/2001 2003/2002 2003/2001 Zagreb 100.4 101.2 100.9 105.8 107.6 113.8 Krapina-Zagorje 93.3 93.0 91.7 104.6 106.3 111.2 Sisak-Moslavina 95.0 95.3 94.4 105.3 106.8 112.5 Karlovac 93.3 96.5 96.3 108.5 107.6 116.8 Varazdin 95.6 97.3 95.4 106.8 105.7 112.9 Koprivnica-Krizevci 102.8 103.2 102.8 105.3 107.4 113.1 Bjelovar-Bilogora 95.4 95.3 96.3 104.8 109.0 114.3 Primorje-Gorski kotar 109.3 109.5 109.8 105.1 108.2 113.7 Lika-Senj 96.8 100.0 104.3 108.3 112.6 121.9 Virovitica-Podravina 91.9 90.1 89.0 102.8 106.5 109.5 Pozega-Slavonia 88.7 87.7 84.5 103.7 103.9 107.8 Slavonski Brod-Posavina 79.4 77.7 76.5 102.7 106.2 109.0 Zadar 89.9 88.1 88.1 102.9 107.8 110.9 Osijek-Baranja 90.7 90.0 89.7 104.0 107.5 111.9 Sibenik-Knin 87.9 88.4 88.8 105.6 108.4 114.5 Vukovar-Sirmium 81.6 81.1 80.4 104.4 106.9 111.5 Split-Dalmatia 88.1 87.3 87.2 103.9 107.7 111.9 Istria 115.5 116.5 114.4 105.8 106.0 112.1 Dubrovnik-Neretva 96.3 96.4 94.5 105.1 105.7 111.1 Meimurje 86.2 85.9 86.2 104.5 108.3 113.1 City of Zagreb 126.7 126.6 129.0 104.9 109.9 115.3 Croatia 100.0 100.0 100.0 104.9 107.9 113.2 The disposable income, as the broadest measure of household current purchasing power, in relation to GDP, apart from the aspect of secondary income distribution, encompasses the aspect of value added distribution among the residential and non-residential households inhabiting the county. It also encompasses the aspect of value added distribution between the various institutional sectors, households, government, entrepreneurs and the foreign sector. The lowest ratio between disposable income and GDP has been recorded in the City of Zagreb. There are numerous factors influencing the value of this indicator. The first is the already mentioned distribution policy. The second factor is the significant number of residents of neighboring counties employed in the City of Zagreb, thus according to the residential rules, GDP is recorded in the residence of the producers, and the household income in the region where the residential employee household is located. The third factor relates to the distribution of value added between the households and the entrepreneurs. The City of Zagreb is the location of domestic and foreign companies with high profits (banks, insurance 121 companies, large state companies, successful foreign-owned companies), therefore a significant proportion of value added is not allocated to the household sector, but is allocated to non-household owners (government, entrepreneurs, abroad) through the income distribution. Apart from the City of Zagreb, the below-average levels of this indicator have been recorded in other counties with the above average GDP per capita: Counties of Istria and Primorje-Gorski kotar. On the other hand, from the household viewpoint, the most favorable ratio has been recorded in Counties of Vukovar-Sirmium, Sibenik-Knin, Slavonski Brod-Posavina and Zagreb. Table 4.26: Some Derivative Indicators on Disposable Income by Households, Primary Income and GDP by counties, ratio in % Disposable income/primary income Disposable income/GDP County of 2001 2002 2003 2001 2002 2003 Zagreb 94.1 92.2 91.3 96.4 82.1 83.8 Krapina-Zagorje 102.0 100.2 99.5 76.9 77.9 77.8 Sisak-Moslavina 108.6 106.5 106.6 71.3 73.4 75.5 Karlovac 107.1 104.3 103.2 71.6 70.6 76.3 Varazdin 100.7 98.3 97.9 65.5 62.0 62.4 Koprivnica-Krizevci 100.0 98.9 98.9 64.8 63.3 66.1 Bjelovar-Bilogora 105.4 104.5 103.8 79.2 74.7 79.4 Primorje-Gorski kotar 98.5 96.4 95.9 60.6 60.9 57.3 Lika-Senj 113.4 111.6 106.6 78.7 68.7 62.1 Virovitica-Podravina 108.3 107.5 109.0 74.9 72.1 72.7 Pozega-Slavonia 108.2 105.4 106.1 78.2 77.0 72.1 Slavonski Brod-Posavina 114.8 112.9 113.0 84.8 80.9 81.9 Zadar 107.5 105.9 104.4 81.3 75.1 67.7 Osijek-Baranja 107.4 105.4 104.7 76.2 70.3 73.4 Sibenik-Knin 116.9 113.4 112.6 90.0 84.0 78.6 Vukovar-Sirmium 117.9 116.0 115.1 91.6 87.0 86.1 Split-Dalmatia 106.5 102.9 101.8 75.8 72.7 71.4 Istria 94.3 92.5 92.6 56.0 53.7 51.3 Dubrovnik-Neretva 104.8 101.5 100.0 69.6 69.4 65.8 Meimurje 101.7 99.9 99.3 67.6 63.2 66.3 City of Zagreb 90.4 88.2 87.3 46.8 45.3 44.4 100.6 98.4 97.6 65.2 62.5 61.6 Croatia Source: Project CBS-EIZG, Regional GDP preliminary results. At the level of analytical regions, it is evident that the indicator of the ratio between disposable income and primary income indicate the development level of a specific region to a great extent, meaning that a higher indicator implies lower regional development. Therefore, the indicator is the highest in the case of Eastern Croatia, and the lowest for Zagreb region. The strong link can also be established between the development level and the ratio of disposable income and GDP. 122 In comparison to the EU member countries, Croatia, according to the estimated indicators, has significantly high ratio between the total disposable income and primary income (97.6 perecnt in 2003). The EU average in 2001 was 87 percent.31 Mainly, this originates from the high value of the estimated transfers from abroad. If the value of transfers received from abroad were to be excluded from the calculation, then this indicator would be slightly above the EU average at 91.5 percent. Due to the same reason, the ratio between the disposable income of the household sector and GDP is also higher than the EU average, with 65.2 percent in Croatia and 61 percent in the EU (both for 2001). However, this indicator in the EU member countries is demonstrating same characteristics as evident in the case of Croatia. The more developed regions have noted lower values of ratios between disposable income and primary income, as well as disposable income and GDP. Table 4.27: Some Derivative Indicators on Disposable Income by Households, Primary Income and GDP by analytical regions, ratio in % Disposable income/primary income Disposable income/GDP Analytical regions 2001 2002 2003 2001 2002 2003 Zagreb region 91.3 89.1 88.2 53.4 50.8 50.0 Central Croatia 103.7 101.8 101.3 70.5 68.8 71.3 Adriatic North 97.9 96.0 95.5 59.9 58.4 55.2 Adriatic South 107.6 104.5 103.3 77.3 73.9 70.6 Eastern Croatia 111.0 109.1 108.9 80.7 76.1 77.1 Croatia 100.6 98.4 97.6 65.2 62.5 61.6 Source: Project CBS-EIZG, Regional GDP preliminary results. Tables 4.28 and 4.29 show the regional data on the social transfer ratios (subcategory at the secondary income distribution account) in the household disposable income, primary income and GDP. Social transfers include pensions, social care, child allowance, health insurance compensations and the unemployment benefits32. According to the sum of all social transfers, the County of Sibenik-Knin has highest share of social transfers received. This county is followed by the County of Vukovar-Sirmium, thus it can be concluded that all of the indicators observed are in an inverse proportional relationship with the economic development level. The Spearman's rank correlation analysis between the indicators of the ratios of social transfers in GDP and GDP per capita of certain counties has shown a negative relationship (at the 5 percent significance level, the resultant coefficient was significant, with the value of -0.70). The significance of the negative correlation has been confirmed by the correlation analysis between the ratio of social transfers in primary income and GDP per capita (-0.42), while in the third case the correlation is insignificant, albeit with a negative sign (-0.22). In addition to Counties of Sibenik-Knin and Vukovar-Sirmium, the highest ratio of social transfers in the primary income can be found in Counties of Slavonski Brod-Posavina, Lika-Senj, Split- Dalmatia and Zadar. It leads to the conclusion that according to this criterion, the regions with the highest social transfer income are the Eastern and Adriatic Croatia. The lowest ratio of social 31 According to Eurostat: Income of private households and gross domestic products in Europe's regions, Statistics in focus, 2003. 32 Structure of disposable income in more detail is presented in Appendix 2. 123 transfers in primary income in 2003 was identified in Counties of Zagreb (17.6 percent), Koprivnica-Krizevci (17.9 percent) and Meimurje (20.0 percent). The three last mentioned counties have also recorded the lowest proportion of social transfers in disposable income. According to this indicator, the order changes somewhat in relation to the previous indicator. The first is still the County of Sibenik-Knin (32.1 percent in 2003), followed by Counties of Split-Dalmatia (28.6 percent), Vukovar-Sirmium (28.3 percent), Zadar (27.8 percent), Slavonski Bro-Posavina (27.6 percent and Lika-Senj (27.4 percent). This clarifies the conclusion that Adriatic Croatia has the highest total proportion of social transfers in disposable income. It is interesting to note that the City of Zagreb has also recorded an above average proportion of social transfers in disposable income (25.3 percent), despite the highest GDP per capita level in Croatia. Primarily, this is a consequence of the significant proportion of pension supplements, which are mostly not dependant on the social status of the recipients. Finally, the highest proportion of social transfers in GDP is found in Eastern Croatia (20.4 percent in 2003), where Counties of Vukovar-Sirmium and Slavonski Brod-Posavina county have notably high proportions. High proportions have also been recorded in certain counties of the Adriatic South (County of Sibenik-knin, 25.2 percent). The lowest proportion of social transfers in GDP has been noted in the the City of Zagreb (11.2 percent), the County of Istria (11.5 percent), County of Koprivnica-Krizevci (12.0 percent in 2003) and the County of Meimurje (13.3 percent). Table 4.28: Social Transfers Relations to Primary Income, Disposable Income and GDP by counties, in % Social transfers/ primary Social transfers/ Social transfers/GDP County of income disposable income 2001 2002 2003 2001 2002 2003 2001 2002 2003 Zagreb 19.2 18.4 17.6 20.4 20.0 19.3 19.6 16.4 16.2 Krapina-Zagorje 23.8 23.3 22.7 23.3 23.2 22.8 17.9 18.1 17.7 Sisak-Moslavina 30.5 29.6 29.2 28.1 27.8 27.4 20.1 20.4 20.7 Karlovac 31.2 29.5 28.3 29.1 28.2 27.4 20.8 19.9 20.9 Varazdin 23.9 22.5 22.3 23.7 22.9 22.7 15.5 14.2 14.2 Koprivnica-Krizevci 18.6 18.1 17.9 18.6 18.3 18.1 12.0 11.6 12.0 Bjelovar-Bilogora 23.1 22.9 21.8 21.9 21.9 21.0 17.4 16.3 16.7 Primorje-Gorski kotar 27.5 26.8 25.5 27.9 27.8 26.6 16.9 16.9 15.2 Lika-Senj 33.9 32.6 29.2 29.9 29.2 27.4 23.5 20.1 17.0 Virovitica-Podravina 24.8 24.8 25.5 22.9 23.1 23.4 17.2 16.7 17.0 Pozega-Slavonia 27.2 25.8 26.3 25.1 24.5 24.8 19.7 18.9 17.9 Slavonski Brod- 32.3 31.4 31.2 28.2 27.8 27.6 23.9 22.5 22.6 Posavina Zadar 30.8 30.2 29.0 28.6 28.5 27.8 23.3 21.4 18.8 Osijek-Baranja 29.3 28.3 27.3 27.3 26.8 26.1 20.8 18.9 19.2 Sibenik-Knin 39.6 37.6 36.1 33.9 33.2 32.1 30.5 27.9 25.2 Vukovar-Sirmium 34.3 33.3 32.6 29.1 28.7 28.3 26.7 24.9 24.4 Split-Dalmatia 32.5 30.6 29.1 30.5 29.7 28.6 23.2 21.6 20.4 Istria 21.7 21.3 20.8 23.0 23.0 22.5 12.9 12.4 11.5 Dubrovnik-Neretva 28.1 26.4 25.8 26.8 26.0 25.7 18.7 18.1 16.9 Meimurje 21.2 20.8 20.0 20.9 20.8 20.1 14.1 13.2 13.3 City of Zagreb 24.3 23.6 22.1 26.9 26.8 25.3 12.6 12.1 11.2 Croatia 26.5 25.5 24.5 26.3 26.0 25.1 17.1 16.2 15.5 Source: Project CBS-EIZG, Regional GDP preliminary results. 124 The negative correlation between social transfers, as source of disposable income of households and the level of economic development can be seen in Figure 4.6. Figure 4.6: The Relationship between Social Transfers and Economic Development of Croatian Counties 190 170 GDP p.c., Croatia = 100 150 130 110 90 70 50 10 15 20 25 30 social transfers, as % of GDP Table 4.29: Social Transfers Relations to Primary Income, Disposable Income and GDP by Analytical Regions, in % Social transfers/ primary Social transfers/ Social transfers/GDP income disposable income Analytical regions 2001 2002 2003 2001 2002 2003 2001 2002 2003 Zagreb region 23.1 22.4 21.0 25.3 25.1 23.8 13.5 12.8 11.9 Central Croatia 24.9 24.1 23.4 24.0 23.6 23.1 16.9 16.3 16.5 Adriatic North 25.7 25.0 24.0 26.2 26.1 25.1 15.7 15.2 13.9 Adriatic South 32.3 30.7 29.4 30.0 29.4 28.5 23.2 21.7 20.1 Eastern Croatia 30.1 29.2 28.8 27.1 26.8 26.4 21.9 20.4 20.4 26.5 25.5 24.5 26.3 26.0 25.1 17.1 16.2 15.5 Croatia Source: Project CBS-EIZG, Regional GDP preliminary results. The largest category within the social transfers is pensions. Table 4.30 shows the proportion of pensions in the total pension amount in Croatia, as well as the significance of pension income in the total disposable income of a certain county. Apart from the demographic structure, the proportion of pensions in the total disposable income depends considerably on the proportion of disabled and retired war veterans in the county's total population. Furthermore, it depends on the structural problems of certain counties influencing the lower levels of primary income (wages, mixed income, proprietor's income) which influences the proportion of pension income in the total disposable income. 125 Table 4.30: Pension Income Regional Distribution in Croatia Proportion of county's pensions in the Proportion of pension income in total total national pensions disposable income County of 2001 2002 2003 2001 2002 2003 Zagreb 5.2 5.3 5.3 12.4 12.3 12.1 Krapina-Zagorje 2.6 2.6 2.6 14.2 14.4 14.4 Sisak-Moslavina 4.2 4.2 4.2 17.6 17.7 17.8 Karlovac 3.3 3.3 3.3 18.5 18.1 17.8 Varazdin 3.4 3.4 3.4 14.3 14.0 14.1 Koprivnica-Krizevci 1.8 1.8 1.9 10.5 10.6 10.7 Bjelovar-Bilogora 2.1 2.2 2.2 12.5 12.7 12.5 Primorje-Gorski kotar 8.7 8.7 8.6 19.3 19.3 18.6 Lika-Senj 1.4 1.4 1.4 20.2 19.9 18.8 Virovitica-Podravina 1.5 1.4 1.5 12.5 12.8 13.3 Pozega-Slavonia 1.5 1.5 1.5 14.3 14.4 14.9 Slavonski Brod-Posavina 3.0 3.0 3.0 15.7 16.1 16.3 Zadar 3.5 3.5 3.6 18.0 18.1 18.0 Osijek-Baranja 6.6 6.6 6.6 16.2 16.4 16.2 Sibenik-Knin 2.8 2.8 2.8 20.7 20.4 20.0 Vukovar-Sirmium 4.0 4.0 4.0 17.8 17.9 17.9 Split-Dalmatia 10.6 10.5 10.4 19.0 18.9 18.4 Istria 5.0 5.1 5.0 15.5 15.5 15.2 Dubrovnik-Neretva 2.7 2.7 2.7 16.6 16.5 16.6 Meimurje 1.5 1.5 1.5 10.9 11.2 10.9 City of Zagreb 24.6 24.6 24.4 18.4 18.4 17.5 Croatia 100.0 100.0 100.0 16.6 16.6 16.3 Source: Project CBS-EIZG, Regional GDP preliminary results. As evident, the lowest proportion of pensions in the total disposable income has been recorded in Counties of Koprivnica-Krizevci (10.7 percent in 2003), Meimurje (10.9 percent) and Zagreb (12.1 percent). On the other hand, the proportion of pension in the total disposable income of the household income sector is the highest in Counties of Sibenik-Knin (20.0 percent) and Lika-Senj (18.8 percent). As opposed to total social transfers, pensions are positively correlated to economic development (Figure 4.7). It can be explained by the way how individual pension is determined. The amount of individual pension primarily depends on period in which contributions are paid, as well as the wage level. As the average wage was higher in the more developed regions, the pensions are also above average in the most developed regions. Figure4.8 presents even stronger negative correlation of economic development of Croatian counties and the share of social transfers in disposable income, when pensions are excluded from social transfers. 126 Figure 4.7: The Relationship between Figure 4.8: The Relationship between Pensions and Economic Development of Social Transfers, without Pensions and Croatian Counties Economic Development of Croatian Counties 190 190 170 170 GDP p.c., Croatia = 100 GDP p.c., Croatia = 100 150 150 130 130 110 110 90 90 70 70 50 50 6 8 10 12 10 15 20 25 social transfers, without pensions, as % of pensions, as % of disposable income disposable income Apart from the pensions and health insurance compensations, the most significant items of social transfers in Croatia are social welfare (social care allowances), unemployment benefits and child allowances. Tables 4.31 and 4.32 show the number of social welfare beneficiaries. Table 4.31: Number of Social Welfare Beneficiaries, by counties, 2003 Number of social Proportion of social welfare County of Population welfare beneficiaries beneficiaries in the total population (%) Zagreb 316 011 3 518 1.1 Krapina-Zagorje 140 521 1 517 1.1 Sisak-Moslavina 182 838 9 079 5.0 Karlovac 139 113 6 464 4.6 Varazdin 183 214 3 551 1.9 Koprivnica-Krizevci 123 169 3 299 2.7 Bjelovar-Bilogora 130 836 4 301 3.3 Primorje-Gorski kotar 305 139 3 125 1.0 Lika-Senj 52 988 1 276 2.4 Virovitica-Podravina 92 200 4 753 5.2 Pozega-Slavonia 85 414 3 200 3.7 Slavonski Brod- 176 221 9 098 5.2 Posavina Zadar 165 757 4 496 2.7 Osijek-Baranja 328 803 14 454 4.4 Sibenik-Knin 113 644 12 785 11.3 Vukovar-Sirmium 202 488 7 232 3.6 Split-Dalmatia 471 017 8 200 1.7 Istria 208 627 1 336 0.6 Dubrovnik-Neretva 123 863 1 967 1.6 Meimurje 118 429 5 797 4.9 City of Zagreb 780 019 12 067 1.5 Croatia 4 440 311 121 515 2.7 Source: Ministry of Health and Social Welfare. 127 Table 4.32: Number of Social Welfare Beneficiaries, by Analytical Regions, 2003 Proportion of social Number of social Analytical regions Population welfare beneficiaries in welfare beneficiaries the total population (%) Zagreb region 1 096 030 15 585 1.4 Central Croatia 1 018 120 34 008 3.3 Adriatic North 566 754 5 737 1.0 Adriatic South 874 281 27 448 3.1 Eastern Croatia 885 126 38 737 4.4 4 440 311 121 515 2.7 Croatia Source: Ministry of Health and Social Welfare. It is visible that, according to the proportion in the total number of inhabitants, the largest number of beneficiaries is present in the County of Sibenik-Knin (11.3 percent of the population). A high proportion of population receiving benefits has also been recorded in Counties of Slavonski Brod- Posavina (5.2 percent), Virovitica-Podravina (5.2 percent), Sisak-Moslavina (5.0 percent) and Meimurje (4.9 percent). As expected, the lowest number of social welfare beneficiaries has been recorded in the most developed counties. Hence, the smallest number of social welfare beneficiaries has been observed in the County of Istria (0.6 percent of the population), followed by the County of Primorje-Gorski kotar (1.0 percent), the City of Zagreb (1.5 percent), and the Counties of Zagreb and Krapina- Zagorje (1.1 percent). Eastern Croatia has the largest proportion of beneficiaries in total number of inhabitants (4.4 percent), while Zagreb region has the lowest (1.4 percent). The average for Croatia is 2.7 percent. The number of beneficiaries of unemployment benefits, apart from the development levels, additionally indicates the structural problems of certain counties, faced with the problems of restructuring companies in the area and the consequent unemployment (Tables 4.33 and 4.34). Consequently, the proportion of beneficiaries of unemployment benefits in the total population has grown significantly in 2002, but then again in 2003 came to the same level as in 2001. The highest proportion in 2003 has been recorded in Counties of Karlovac (2.08 percent of the population), Virovitica-Podravina (2.05 percent) and Dubrovnik-Neretva (1.93 percent). In 2002, except these three counties levels above 2 percent of the beneficiaries of unemployment benefits in the total population have also been recorded in Counties of Split-Dalmatia, Bjelovar-Bilogora, Slavonski Brod-Posavina, Sisak-Moslavina, Osijek-baranja and Zadar. The lowest proportion of beneficiaries of unemployment benefit in 2003 has been noted in the City of Zagreb (1.12 percent), Counties of Pozega-Slavonia, Varazdin, Zagreb and Lika-Senj. Croatian average of the proportion of beneficiaries of unemployment benefits in 2003 was 1.53 percent. Only Zagreb region was below that average (1.17 percent), while all other analytical regions where above average, with highest proportion in Eastern Croatia (1.76 percent), and Adriatic South (1.75 percent). 128 Table 4.33: Number of Beneficiaries of Unemployment Benefits by Counties Proportion of total population in % County of 2001 2002 2003 2001 2002 2003 Zagreb 4 236 5 050 4 147 1.34 1.60 1.31 Krapina-Zagorje 2 560 2 717 2 156 1.82 1.93 1.53 Sisak-Moslavina 3 401 3 713 2 681 1.86 2.03 1.47 Karlovac 2 707 3 422 2 899 1.95 2.46 2.08 Varazdin 2 623 2 910 2 513 1.43 1.59 1.37 Koprivnica-Krizevci 1 778 2 081 1 822 1.44 1.69 1.48 Bjelovar-Bilogora 2 347 2 960 2 167 1.79 2.26 1.66 Primorje-Gorski kotar 4 936 4 858 4 446 1.62 1.59 1.46 Lika-Senj 700 912 698 1.32 1.72 1.32 Virovitica-Podravina 1 696 2 127 1 891 1.84 2.31 2.05 Pozega-Slavonia 1 175 1 322 1 166 1.38 1.55 1.37 Slavonski Brod- 3 026 3 755 3 146 1.72 2.13 1.79 Posavina Zadar 2 607 3 378 2 827 1.57 2.04 1.71 Osijek-Baranja 6 279 6 987 5 904 1.91 2.12 1.80 Sibenik-Knin 1 607 1 971 1 760 1.41 1.73 1.55 Vukovar-Sirmium 2 861 3 619 3 455 1.41 1.79 1.71 Split-Dalmatia 8 604 10 295 8 320 1.83 2.19 1.77 Istria 3 061 3 376 3 020 1.47 1.62 1.45 Dubrovnik-Neretva 2 181 2 886 2 396 1.76 2.33 1.93 Meimurje 1 683 1 888 1 837 1.42 1.59 1.55 City of Zagreb 10 302 10 569 8 728 1.32 1.35 1.12 70 370 80 796 67 979 1.58 1.82 1.53 Croatia Source: Croatian Employment Service. Table 4.34: Number of Beneficiaries of Unemployment Benefits by Analytical Regions Proportion of total population in % Analytical regions 2001 2002 2003 2001 2002 2003 Zagreb region 14 538 15 619 12 875 1.33 1.43 1.17 Central Croatia 17 099 19 691 16 075 1.68 1.93 1.58 Adriatic North 8 697 9 146 8 164 1.53 1.61 1.44 Adriatic South 14 999 18 530 15 303 1.72 2.12 1.75 Eastern Croatia 15 037 17 810 15 562 1.70 2.01 1.76 70 370 80 796 67 979 1.58 1.82 1.53 Croatia Source: Croatian Employment Service. 129 A significant category of government social transfers relates to child allowance. However, this category correlates more significantly with the demographic characteristics, rather than development level. On the basis of the data from Tables 4.35 and 4.36, it is clear that the highest proportion of child allowance in the total Croatian child allowance amount has been provided in Eastern Croatia (28.0 percent), while the smallest proportion of child allowance has been recorded in Adriatic Norht (7.7 percent). The county with the highest share of child allowance is the County of Split-Dalmatia (11.9 percent in 2002), and the smallest share is recorded in the County of Lika- Senj (1.1 percent). Table 4.35: Child Allowance, by counties, 2001-2003 in HRK structure, in % County of 2001 2002 2003 2001 2002 2003 Zagreb 143 965 208 101 998 569 95 437 563 6.0 6.1 5.9 Krapina-Zagorje 74 231 490 55 295 196 51 899 654 3.1 3.3 3.2 Sisak-Moslavina 101 198 122 71 824 648 69 970 009 4.2 4.3 4.3 Karlovac 66 339 838 44 167 597 41 265 683 2.7 2.6 2.6 Varazdin 104 989 622 75 650 531 71 451 482 4.3 4.5 4.4 Koprivnica-Krizevci 72 375 928 54 448 343 53 441 480 3.0 3.2 3.3 Bjelovar-Bilogora 83 303 955 64 677 047 62 833 740 3.4 3.9 3.9 Primorje-Gorski kotar 100 076 650 67 353 594 62 212 662 4.1 4.0 3.9 Lika-Senj 26 245 348 17 720 786 16 887 273 1.1 1.1 1.0 Virovitica-Podravina 63 873 625 49 303 532 49 099 047 2.6 2.9 3.0 Pozega-Slavonia 65 944 945 46 315 726 44 833 073 2.7 2.8 2.8 Slavonski Brod- 142 731 863 101 863 007 100 529 838 5.9 6.1 6.2 Posavina Zadar 96 735 939 67 761 166 64 800 471 4.0 4.0 4.0 Osijek-Baranja 225 715 452 155 987 161 149 824 176 9.3 9.3 9.3 Sibenik-Knin 71 050 325 49 106 011 47 363 124 2.9 2.9 2.9 Vukovar-Sirmium 144 465 304 107 585 457 106 417 957 6.0 6.4 6.6 Split-Dalmatia 302 789 050 199 827 242 196 315 922 12.5 11.9 12.2 Istria 67 717 663 48 211 912 45 001 534 2.8 2.9 2.8 Dubrovnik-Neretva 81 552 933 51 253 878 49 514 309 3.4 3.1 3.1 Meimurje 72 267 614 55 060 947 53 135 887 3.0 3.3 3.3 City of Zagreb 307 300 937 193 774 020 179 365 169 12.7 11.5 11.1 Croatia 2 414 871 812 1 679 186 373 1 611 600 053 100.0 100.0 100.0 Source: Croatian Institute for Pension Insurance ­ HZMO. Table 4.36: Child Allowance, by Analytical Regions, 2001-2003 in HRK structure, in % Analytical regions 2001 2002 2003 2001 2002 2003 Zagreb region 451 266 145 295 772 589 274 802 731 18.7 17.6 17.1 Central Croatia 574 706 570 421 124 310 403 997 935 23.8 25.1 25.1 Adriatic North 194 039 661 133 286 293 124 101 469 8.0 7.9 7.7 Adriatic South 552 128 246 367 948 297 357 993 827 22.9 21.9 22.2 Eastern Croatia 642 731 190 461 054 884 450 704 091 26.6 27.5 28.0 Croatia 2 414 871 812 1 679 186 373 1 611 600 053 100.0 100.0 100.0 Source: Croatian Institute for Pension Insurance ­ HZMO. As conclusion, Table 4.37 presents correlation coefficients between social transfers and the level of economic development in terms of both, GDP p.c., as well as gross disposable income per capita in period 2001-2003. According to expectations social transfers are negatively correlated with development variables, meaning that more developed counties have a lower share of social transfers 130 in GDP (GDI). Coefficients are negative and significant at 5 percent significance level for all presented types of social transfer except pensions. Those results confirm the hypothesis that income redistribution process significantly reduces the inequality in welfare of Croatian counties, but the impact of various types of transfers is different. Transfers in the scope of obligatory social security system are not significantly correlated with development level of individual county. Table 4.37: The Correlation Coefficients between Social Transfer Variables and the Level of Economic Development Variable GDP p.c. GDI p.c. Total social transfers -0.17 -0.23 Pensions 0.08 0.05 Social transfers without pensions -0.61* -0.73* Unemployment benefits -0.67* -0.73* Social welfare benefits -0.56* -0.64* Child allowances -0.62* -0.72* Note: * 5 percent significance level. Source: Author's calculations. 131 APPENDIX A3.I: GDP IN CROATIA BY COUNTIES IN PERIOD 2001-2003 Table A3.1. Gross domestic product in current prices, 2001 Counties GDP, mil. GDP, mil. GDP, Structure. GDP p.c. Index kn EUR mil. USD in % kn EUR USD Croatia = 100) Croatia 165 639 22 177 19 863 100.0 37 309 4 995 4 474 100.0 Zagreb 7 863 1 053 943 4.7 25 334 3 392 3 038 67.9 Krapina-Zagorje 4 194 561 503 2.5 29 485 3 948 3 536 79.0 Sisak-Moslavina 5 997 803 719 3.6 32 375 4 335 3 882 86.8 Karlovac 4 486 601 538 2.7 31 693 4 243 3 801 84.9 Varazdin 6 553 877 786 4.0 35 490 4 752 4 256 95.1 Koprivnica- 4 801 643 576 2.9 38 598 5 168 4 629 103.5 Krizevci Bjelovar- 3 893 521 467 2.4 29 304 3 923 3 514 78.5 Bilogora Primorje-Gorski 13 399 1 794 1 607 8.1 43 853 5 871 5 259 117.5 kotar Lika-Senj 1 606 215 192 1.0 29 934 4 008 3 590 80.2 Virovitica- 2 783 373 334 1.7 29 834 3 994 3 578 80.0 Podravina Pozega-Slavonia 2 367 317 284 1.4 27 567 3 691 3 306 73.9 Slavonski Brod- 4 026 539 483 2.4 22 768 3 048 2 730 61.0 Posavina Zadar 4 372 585 524 2.6 26 899 3 601 3 226 72.1 Osijek-Baranja 9 565 1 280 1 147 5.8 28 955 3 877 3 472 77.6 Sibenik-Knin 2 687 360 322 1.6 23 747 3 179 2 848 63.6 Vukovar- 4 434 594 532 2.7 21 648 2 898 2 596 58.0 Sirmium Split-Dalmatia 13 146 1 760 1 576 7.9 28 272 3 785 3 390 75.8 Istria 10 368 1 388 1 243 6.3 50 174 6 718 6 017 134.5 Dubrovnik- 4 142 555 497 2.5 33 642 4 504 4 034 90.2 Neretva Meimurje 3 673 492 440 2.2 31 010 4 152 3 719 83.1 City of Zagreb 51 284 6 866 6 150 31.0 65 820 8 812 7 893 176.4 132 Table A3.2. Gross DomesticPproduct in Current Prices, 2002. Counties GDP, mil. GDP, mil. GDP, Structure, GDP p.c. Index kn EUR mil. USD in % kn EUR USD Croatia = 100) Croatia 181 231 24 468 23 047 100.0 40 814 5 510 5 190 100.0 Zagreb 9 839 1 328 1 251 5.4 31 456 4 247 4 000 77.1 Krapina-Zagorje 4 305 581 548 2.4 30 453 4 112 3 873 74.6 Sisak-Moslavina 6 097 823 775 3.4 33 127 4 472 4 213 81.2 Karlovac 4 895 661 623 2.7 34 873 4 708 4 435 85.4 Varazdin 7 371 995 937 4.1 40 051 5 407 5 093 98.1 Koprivnica- Krizevci 5 146 695 654 2.8 41 577 5 613 5 287 101.9 Bjelovar- Bilogora 4 296 580 546 2.4 32 564 4 397 4 141 79.8 Primorje-Gorski kotar 14 021 1 893 1 783 7.7 45 903 6 197 5 837 112.5 Lika-Senj 1 974 267 251 1.1 37 116 5 011 4 720 90.9 Virovitica- Podravina 2 955 399 376 1.6 31 873 4 303 4 053 78.1 Pozega-Slavonia 2 490 336 317 1.4 29 041 3 921 3 693 71.2 Slavonski Brod- Posavina 4 332 585 551 2.4 24 521 3 311 3 118 60.1 Zadar 4 916 664 625 2.7 29 958 4 045 3 810 73.4 Osijek-Baranja 10 777 1 455 1 371 5.9 32 675 4 411 4 155 80.1 Sibenik-Knin 3 043 411 387 1.7 26 839 3 624 3 413 65.8 Vukovar- Sirmium 4 847 655 616 2.7 23 797 3 213 3 026 58.3 Split-Dalmatia 14 350 1 937 1 825 7.9 30 636 4 136 3 896 75.1 Istria 11 481 1 550 1 460 6.3 55 335 7 471 7 037 135.6 Dubrovnik- Neretva 4 379 591 557 2.4 35 429 4 783 4 505 86.8 Meimurje 4 107 554 522 2.3 34 650 4 678 4 406 84.9 City of Zagreb 55 610 7 508 7 072 30.7 71 355 9 634 9 074 174.8 133 Table A3.3. Gross domestic product in current prices, 2003. Counties GDP, mil. GDP, mil. GDP, Structure, GDP p.c. Index kn EUR mil. USD in % kn EUR USD Croatia = 100) Croatia 198 422 26 235 29 609 100.0 44 689 5 909 6 669 100.0 Zagreb 10 480 1 386 1 564 5.3 33 165 4 385 4 949 74.2 Krapina-Zagorje 4 556 602 680 2.3 32 427 4 287 4 839 72.6 Sisak-Moslavina 6 290 832 938 3.2 34 409 4 549 5 135 77.0 Karlovac 4 831 639 721 2.4 34 730 4 592 5 183 77.7 Varazdin 7 709 1 019 1 150 3.9 42 080 5 564 6 279 94.2 Koprivnica- Krizevci 5 275 697 787 2.6 42 817 5 661 6 389 95.8 Bjelovar- Bilogora 4 367 577 652 2.2 33 387 4 414 4 982 74.7 Primorje-Gorski kotar 16 100 2 129 2 402 8.1 52 770 6 977 7 874 118.1 Lika-Senj 2 449 324 365 1.2 46 208 6 109 6 895 103.4 Virovitica- Podravina 3 105 411 463 1.5 33 677 4 453 5 025 75.4 Pozega-Slavonia 2 754 364 411 1.4 32 248 4 264 4 812 72.2 Slavonski Brod- Posavina 4 528 599 676 2.3 25 698 3 398 3 835 57.5 Zadar 5 936 785 886 3.0 35 802 4 734 5 342 80.1 Osijek-Baranja 11 059 1 462 1 650 5.6 33 634 4 447 5 019 75.3 Sibenik-Knin 3 536 468 528 1.8 31 127 4 115 4 645 69.7 Vukovar- Sirmium 5 203 688 776 2.6 25 694 3 397 3 834 57.5 Split-Dalmatia 15 839 2 094 2 364 8.0 33 628 4 446 5 018 75.3 Istria 12 814 1 694 1 912 6.5 61 429 8 122 9 167 137.5 Dubrovnik- Neretva 4 896 647 731 2.5 39 516 5 225 5 897 88.4 Meimurje 4 241 561 633 2.1 35 819 4 736 5 345 80.2 City of Zagreb 62 454 8 257 9 320 31.5 80 069 10 586 11 948 179.2 134 Table A3.4. Sectoral Structure of GDP, 2001, by counties L, M, Counties A, B C, D, E F G H I J, K TOTAL N, O Croatia 9.5 25.7 5.2 12.1 3.6 10.5 10.6 22.8 100.0 Zagreb 16.6 27.1 4.8 20.0 2.1 10.0 5.8 13.6 100.0 Krapina-Zagorje 13.6 29.3 4.8 13.8 2.5 9.5 4.9 21.6 100.0 Sisak-Moslavina 12.7 36.2 5.5 6.6 1.6 10.5 6.4 20.4 100.0 Karlovac 9.8 25.4 19.3 6.9 3.2 8.2 5.6 21.5 100.0 Varazdin 12.8 35.2 4.8 10.1 1.9 7.5 6.8 21.0 100.0 Koprivnica- Krizevci 22.7 36.0 3.5 10.0 1.6 6.7 5.1 14.4 100.0 Bjelovar- Bilogora 29.6 20.8 5.3 6.4 2.1 8.0 6.6 21.2 100.0 Primorje-Gorski kotar 2.5 26.8 5.4 11.0 8.1 14.9 10.3 21.0 100.0 Lika-Senj 20.8 18.1 10.6 4.1 4.9 11.4 4.2 26.0 100.0 Virovitica- Podravina 30.9 22.7 3.2 13.4 1.2 6.2 4.3 18.0 100.0 Pozega-Slavonia 24.1 17.4 4.7 12.0 1.5 8.0 4.5 27.8 100.0 Slavonski Brod- Posavina 20.6 23.4 5.3 8.6 1.2 9.8 6.7 24.4 100.0 Zadar 11.6 14.3 7.2 10.6 5.4 12.3 10.6 27.9 100.0 Osijek-Baranja 21.4 19.2 5.3 10.6 1.6 9.1 8.5 24.3 100.0 Sibenik-Knin 9.5 13.8 5.8 11.4 5.2 13.3 9.0 32.0 100.0 Vukovar- Sirmium 30.8 10.9 8.4 11.2 1.3 7.8 4.7 24.8 100.0 Split-Dalmatia 4.6 23.5 4.9 12.6 4.4 12.1 11.4 26.4 100.0 Istria 5.6 29.4 5.6 10.0 12.7 9.2 9.6 17.8 100.0 Dubrovnik- Neretva 9.0 12.5 4.5 7.2 9.0 18.2 11.8 27.9 100.0 Meimurje 17.6 35.9 5.9 7.5 1.4 7.0 7.7 16.9 100.0 City of Zagreb 0.5 27.1 3.6 15.1 2.0 10.7 16.2 25.0 100.0 135 Table A3.5. Sectoral structure of GDP, 2002., by counties L, M, Counties A, B C, D, E F G H I J, K TOTAL N, O Croatia 9.1 24.2 5.6 13.6 3.8 10.3 11.5 22.0 100.0 Zagreb 15.6 31.7 5.1 18.6 2.8 8.5 5.6 12.0 100.0 Krapina-Zagorje 13.6 29.8 7.5 9.8 2.8 9.6 5.8 21.0 100.0 Sisak-Moslavina 12.5 32.1 5.5 8.2 2.9 11.9 6.0 20.9 100.0 Karlovac 10.7 22.7 20.1 7.6 4.3 8.4 6.4 19.8 100.0 Varazdin 12.4 32.1 6.5 12.8 2.5 7.1 6.8 19.7 100.0 Koprivnica- Krizevci 23.1 35.7 3.5 10.4 1.7 5.8 5.5 14.2 100.0 Bjelovar- Bilogora 29.0 19.0 4.6 9.2 3.1 8.5 7.4 19.2 100.0 Primorje-Gorski kotar 2.5 21.8 6.5 13.1 7.9 14.9 12.0 21.5 100.0 Lika-Senj 17.7 20.8 12.6 7.0 5.2 9.6 4.4 22.7 100.0 Virovitica- Podravina 29.9 20.4 4.1 15.2 1.9 5.8 4.9 17.8 100.0 Pozega-Slavonia 23.0 23.4 4.6 8.7 1.9 7.3 5.0 26.0 100.0 Slavonski Brod- Posavina 19.4 19.0 5.9 12.1 1.8 10.0 7.3 24.5 100.0 Zadar 10.8 11.8 9.5 13.1 6.2 11.1 10.9 26.6 100.0 Osijek-Baranja 19.7 17.9 5.0 13.9 1.9 9.4 8.9 23.2 100.0 Sibenik-Knin 8.9 18.4 6.7 10.6 5.3 12.8 9.2 28.1 100.0 Vukovar- Sirmium 27.9 12.9 7.6 10.7 2.3 9.3 4.6 24.6 100.0 Split-Dalmatia 4.1 21.3 5.7 14.2 4.6 12.8 11.4 26.0 100.0 Istria 5.4 32.9 6.1 10.0 11.2 6.3 10.8 17.3 100.0 Dubrovnik- Neretva 9.1 11.8 5.1 8.0 7.2 18.0 13.9 27.0 100.0 Meimurje 16.2 35.5 6.7 9.0 2.2 6.0 8.8 15.6 100.0 City of Zagreb 0.4 24.0 3.3 17.2 2.1 10.8 18.0 24.1 100.0 136 Table A3.6. Sectoral structure of GDP, 2003., by counties L, M, Counties A, B C, D, E F G H I J, K TOTAL N, O Croatia 7.4 24.0 6.6 14.5 4.0 10.1 12.7 20.8 100.0 Zagreb 12.8 32.7 6.0 18.0 3.1 8.7 6.1 12.6 100.0 Krapina-Zagorje 11.2 31.1 6.7 11.1 3.0 10.2 6.3 20.5 100.0 Sisak-Moslavina 10.7 28.9 5.9 10.2 3.1 13.1 6.4 21.8 100.0 Karlovac 9.6 28.1 10.8 9.4 5.1 9.1 7.3 20.7 100.0 Varazdin 10.3 29.0 9.5 12.9 2.5 7.2 8.4 20.0 100.0 Koprivnica- Krizevci 19.7 34.7 4.3 12.3 2.0 5.9 6.8 14.2 100.0 Bjelovar- Bilogora 25.0 19.7 5.9 9.8 3.4 7.4 9.0 19.7 100.0 Primorje-Gorski kotar 2.0 22.1 8.2 13.3 7.5 14.4 12.5 20.0 100.0 Lika-Senj 12.6 18.1 25.9 5.6 4.7 10.1 3.9 19.1 100.0 Virovitica- Podravina 25.0 24.3 4.5 16.7 1.9 5.5 5.3 16.8 100.0 Pozega-Slavonia 18.4 23.8 6.2 11.3 2.8 6.9 5.5 25.1 100.0 Slavonski Brod- Posavina 16.3 20.7 6.8 12.4 2.5 10.2 7.6 23.5 100.0 Zadar 8.0 13.3 14.3 12.9 6.8 10.6 10.7 23.4 100.0 Osijek-Baranja 17.0 18.1 5.6 15.2 2.0 9.3 10.3 22.5 100.0 Sibenik-Knin 6.7 16.9 10.5 11.5 5.3 12.3 10.0 26.9 100.0 Vukovar- Sirmium 22.9 16.3 8.6 11.9 2.5 9.6 4.9 23.4 100.0 Split-Dalmatia 3.4 17.5 7.5 16.3 4.8 13.2 12.8 24.4 100.0 Istria 4.7 30.2 7.1 11.0 11.5 6.9 12.2 16.4 100.0 Dubrovnik- Neretva 7.1 10.8 7.3 9.6 8.1 17.7 14.5 25.0 100.0 Meimurje 13.6 35.8 7.3 9.3 2.3 6.4 9.2 16.1 100.0 City of Zagreb 0.4 24.5 3.9 17.8 2.0 9.7 19.8 22.0 100.0 137 APPENDIX B3. GROSS DISPOSABLE INCOME OF HOUSEHOLD SECTOR IN CROATIA BY COUNTIES IN PERIOD 2001-2003 Table B3.1: Primary, Secondary and Total Gross Disposable Income of Household Sector in Croatia in 2001, by counties in millions of kunas Structure, as % of county disposable Total Structure, income Primary Secondary Counties disposable as % of total income income Total income income Primary Secondary disposable income income income Croatia 107 294 681 107 975 100 99.4 0.6 100.0 Zagreb 8 056 -477 7 579 7.0 106.3 -6.3 100.0 Krapina-Zagorje 3 164 63 3 227 3.0 98.1 1.9 100.0 Sisak-Moslavina 3 938 339 4 278 4.0 92.1 7.9 100.0 Karlovac 2 999 213 3 211 3.0 93.4 6.6 100.0 Varazdin 4 259 32 4 291 4.0 99.3 0.7 100.0 Koprivnica- Krizevci 3 110 1 3 111 2.9 100.0 0.0 100.0 Bjelovar- Bilogora 2 924 159 3 082 2.9 94.9 5.1 100.0 Primorje-Gorski kotar 8 246 -125 8 121 7.5 101.5 -1.5 100.0 Lika-Senj 1 115 149 1 263 1.2 88.2 11.8 100.0 Virovitica- Podravina 1 927 159 2 086 1.9 92.4 7.6 100.0 Pozega-Slavonia 1 712 140 1 851 1.7 92.4 7.6 100.0 Slavonski Brod- Posavina 2 976 439 3 415 3.2 87.1 12.9 100.0 Zadar 3 305 248 3 552 3.3 93.0 7.0 100.0 Osijek-Baranja 6 790 500 7 290 6.8 93.1 6.9 100.0 Sibenik-Knin 2 069 349 2 418 2.2 85.6 14.4 100.0 Vukovar- Sirmium 3 446 617 4 063 3.8 84.8 15.2 100.0 Split-Dalmatia 9 362 604 9 966 9.2 93.9 6.1 100.0 Istria 6 155 -353 5 803 5.4 106.1 -6.1 100.0 Dubrovnik- Neretva 2 751 131 2 882 2.7 95.5 4.5 100.0 Meimurje 2 442 42 2 484 2.3 98.3 1.7 100.0 City of Zagreb 26 551 -2 548 24 002 22.2 110.6 -10.6 100.0 138 Table B3.2. Primary, secondary and total gross disposable income of household sector in Croatia in 2002, by counties in millions of kunas Structure, as % of county disposable Total Structure, income Primary Secondary Counties disposable as % of total income income Total income income Primary Secondary disposable income income income Croatia 115 146 -1 840 113 306 100 101.6 -1.6 100.0 Zagreb 8 766 -687 8 079 7.1 108.5 -8.5 100.0 Krapina-Zagorje 3 350 5 3 355 3.0 99.8 0.2 100.0 Sisak-Moslavina 4 204 273 4 477 4.0 93.9 6.1 100.0 Karlovac 3 314 142 3 456 3.1 95.9 4.1 100.0 Varazdin 4 651 -81 4 570 4.0 101.8 -1.8 100.0 Koprivnica- Krizevci 3 294 -35 3 259 2.9 101.1 -1.1 100.0 Bjelovar- Bilogora 3 070 138 3 207 2.8 95.7 4.3 100.0 Primorje-Gorski kotar 8 848 -315 8 533 7.5 103.7 -3.7 100.0 Lika-Senj 1 216 141 1 357 1.2 89.6 10.4 100.0 Virovitica- Podravina 1 983 149 2 132 1.9 93.0 7.0 100.0 Pozega-Slavonia 1 819 99 1 918 1.7 94.8 5.2 100.0 Slavonski Brod- Posavina 3 102 401 3 502 3.1 88.6 11.4 100.0 Zadar 3 485 204 3 690 3.3 94.5 5.5 100.0 Osijek-Baranja 7 183 389 7 572 6.7 94.9 5.1 100.0 Sibenik-Knin 2 255 302 2 557 2.3 88.2 11.8 100.0 Vukovar- Sirmium 3 634 583 4 216 3.7 86.2 13.8 100.0 Split-Dalmatia 10 144 291 10 435 9.2 97.2 2.8 100.0 Istria 6 666 -500 6 166 5.4 108.1 -8.1 100.0 Dubrovnik- Neretva 2 996 44 3 041 2.7 98.5 1.5 100.0 Meimurje 2 601 -3 2 597 2.3 100.1 -0.1 100.0 City of Zagreb 28 566 -3 380 25 186 22.2 113.4 -13.4 100.0 139 Table B3.3. Primary, secondary and total gross disposable income of household sector in Croatia in 2003, by counties in millions of kunas Structure, as % of county disposable Total Structure, income Primary Secondary Counties disposable as % of total income income Total income income Primary Secondary disposable income income income Croatia 125 190 -2 971 122 220 100 102.4 -2.4 100.0 Zagreb 9 611 -832 8 779 7.2 109.5 -9.5 100.0 Krapina-Zagorje 3 563 -18 3 546 2.9 100.5 -0.5 100.0 Sisak-Moslavina 4 459 292 4 751 3.9 93.8 6.2 100.0 Karlovac 3 572 115 3 687 3.0 96.9 3.1 100.0 Varazdin 4 914 -105 4 809 3.9 102.2 -2.2 100.0 Koprivnica- Krizevci 3 524 -40 3 484 2.9 101.1 -1.1 100.0 Bjelovar- Bilogora 3 342 127 3 468 2.8 96.4 3.6 100.0 Primorje-Gorski kotar 9 611 -392 9 219 7.5 104.2 -4.2 100.0 Lika-Senj 1 428 94 1 522 1.2 93.8 6.2 100.0 Virovitica- Podravina 2 072 187 2 258 1.8 91.7 8.3 100.0 Pozega-Slavonia 1 871 115 1 986 1.6 94.2 5.8 100.0 Slavonski Brod- Posavina 3 283 427 3 710 3.0 88.5 11.5 100.0 Zadar 3 848 170 4 018 3.3 95.8 4.2 100.0 Osijek-Baranja 7 753 364 8 117 6.6 95.5 4.5 100.0 Sibenik-Knin 2 469 311 2 779 2.3 88.8 11.2 100.0 Vukovar- Sirmium 3 893 588 4 480 3.7 86.9 13.1 100.0 Split-Dalmatia 11 103 199 11 302 9.2 98.2 1.8 100.0 Istria 7 093 -523 6 570 5.4 108.0 -8.0 100.0 Dubrovnik- Neretva 3 220 1 3 221 2.6 100.0 0.0 100.0 Meimurje 2 829 -19 2 810 2.3 100.7 -0.7 100.0 City of Zagreb 31 733 -4 030 27 703 22.7 114.5 -14.5 100.0 140 Table B3.4. Structure of primary income of household sector in Croatia in 2001, by counties in millions of kunas Structure, as % of county primary income Gross Other Total Mixed Counties wages and primary primary income ** Other salaries* income*** income Mixed Gross W&S primary income income Croatia 83 708 17 053 6 533 107 294 78.0 15.9 6.1 Zagreb 6 304 1 450 302 8 056 78.3 18.0 3.7 Krapina-Zagorje 2 291 723 149 3 164 72.4 22.9 4.7 Sisak-Moslavina 2 997 701 241 3 938 76.1 17.8 6.1 Karlovac 2 366 458 176 2 999 78.9 15.3 5.9 Varazdin 3 209 801 249 4 259 75.3 18.8 5.8 Koprivnica- Krizevci 1 983 916 211 3 110 63.8 29.4 6.8 Bjelovar- Bilogora 1 835 921 167 2 924 62.8 31.5 5.7 Primorje-Gorski kotar 6 818 854 574 8 246 82.7 10.4 7.0 Lika-Senj 784 265 66 1 115 70.3 23.7 5.9 Virovitica- Podravina 1 231 580 116 1 927 63.9 30.1 6.0 Pozega-Slavonia 1 205 414 93 1 712 70.4 24.2 5.4 Slavonski Brod- Posavina 2 049 776 150 2 976 68.9 26.1 5.0 Zadar 2 512 611 181 3 305 76.0 18.5 5.5 Osijek-Baranja 5 243 1 149 398 6 790 77.2 16.9 5.9 Sibenik-Knin 1 626 337 106 2 069 78.6 16.3 5.1 Vukovar- Sirmium 2 291 996 159 3 446 66.5 28.9 4.6 Split-Dalmatia 7 709 1 199 453 9 362 82.3 12.8 4.8 Istria 4 736 963 456 6 155 76.9 15.7 7.4 Dubrovnik- Neretva 2 127 464 160 2 751 77.3 16.9 5.8 Meimurje 1 708 576 159 2 442 69.9 23.6 6.5 City of Zagreb 22 684 1 899 1 968 26 551 85.4 7.2 7.4 *Gross wages and salaries includes social contribution and taxes on personal income ** Mixed income presents income from unincorporated enterprises in the ownership of households (craftsman and agricultural producers) ***Other primary incomes includes property income and imputed dwelling rents 141 Table B3.5. Structure of primary income of household sector in Croatia in 2002, by counties in millions of kunas Structure, as % of county primary income Gross Other Total Mixed Counties wages and primary primary income ** Other salaries* income*** income Mixed Gross W&S primary income income Croatia 90 125 18 076 6 945 115 146 78.3 15.7 6.0 Zagreb 6 784 1 609 372 8 766 77.4 18.4 4.2 Krapina-Zagorje 2 466 739 145 3 350 73.6 22.1 4.3 Sisak-Moslavina 3 230 728 246 4 204 76.8 17.3 5.8 Karlovac 2 609 515 190 3 314 78.7 15.5 5.7 Varazdin 3 505 877 269 4 651 75.4 18.9 5.8 Koprivnica- Krizevci 2 103 966 225 3 294 63.8 29.3 6.8 Bjelovar- Bilogora 1 944 946 180 3 070 63.3 30.8 5.9 Primorje-Gorski kotar 7 330 931 586 8 848 82.9 10.5 6.6 Lika-Senj 851 285 80 1 216 70.0 23.4 6.6 Virovitica- Podravina 1 273 592 118 1 983 64.2 29.8 6.0 Pozega-Slavonia 1 310 415 93 1 819 72.0 22.8 5.1 Slavonski Brod- Posavina 2 149 798 155 3 102 69.3 25.7 5.0 Zadar 2 675 613 197 3 485 76.8 17.6 5.7 Osijek-Baranja 5 550 1 196 437 7 183 77.3 16.7 6.1 Sibenik-Knin 1 774 367 114 2 255 78.7 16.3 5.0 Vukovar- Sirmium 2 454 1 011 169 3 634 67.5 27.8 4.7 Split-Dalmatia 8 394 1 280 471 10 144 82.7 12.6 4.6 Istria 5 161 1 028 476 6 666 77.4 15.4 7.1 Dubrovnik- Neretva 2 340 494 163 2 996 78.1 16.5 5.4 Meimurje 1 867 564 170 2 601 71.8 21.7 6.5 City of Zagreb 24 357 2 121 2 088 28 566 85.3 7.4 7.3 *Gross wages and salaries includes social contribution and taxes on personal income ** Mixed income presents income from unincorporated enterprises in the ownership of households (craftsman and agricultural producers) ***Other primary incomes includes property income and imputed dwelling rents 142 Table B3.6. Structure of primary income of household sector in Croatia in 2003, by counties in millions of kunas Structure, as % of county primary income Gross Other Total Mixed Counties wages and primary primary income ** Other salaries* income*** income Mixed Gross W&S primary income income Croatia 99 433 18 434 7 324 125 190 79.4 14.7 5.8 Zagreb 7 613 1 622 376 9 611 79.2 16.9 3.9 Krapina-Zagorje 2 705 704 155 3 563 75.9 19.7 4.3 Sisak-Moslavina 3 423 791 245 4 459 76.8 17.7 5.5 Karlovac 2 841 552 179 3 572 79.6 15.4 5.0 Varazdin 3 817 824 273 4 914 77.7 16.8 5.6 Koprivnica- Krizevci 2 296 1 000 228 3 524 65.2 28.4 6.5 Bjelovar- Bilogora 2 126 1 031 184 3 342 63.6 30.9 5.5 Primorje-Gorski kotar 8 018 956 636 9 611 83.4 10.0 6.6 Lika-Senj 1 046 288 94 1 428 73.3 20.2 6.6 Virovitica- Podravina 1 322 624 125 2 072 63.8 30.1 6.0 Pozega-Slavonia 1 380 389 102 1 871 73.7 20.8 5.5 Slavonski Brod- Posavina 2 320 803 161 3 283 70.6 24.5 4.9 Zadar 3 032 591 226 3 848 78.8 15.4 5.9 Osijek-Baranja 5 984 1 327 441 7 753 77.2 17.1 5.7 Sibenik-Knin 1 932 406 130 2 469 78.3 16.5 5.3 Vukovar- Sirmium 2 651 1 057 185 3 893 68.1 27.2 4.7 Split-Dalmatia 9 357 1 246 499 11 103 84.3 11.2 4.5 Istria 5 582 1 004 507 7 093 78.7 14.2 7.2 Dubrovnik- Neretva 2 573 472 175 3 220 79.9 14.7 5.4 Meimurje 2 064 593 172 2 829 73.0 20.9 6.1 City of Zagreb 27 349 2 153 2 231 31 733 86.2 6.8 7.0 *Gross wages and salaries includes social contribution and taxes on personal income ** Mixed income presents income from unincorporated enterprises in the ownership of households (craftsman and agricultural producers) ***Other primary incomes includes property income and imputed dwelling rents 143 Table B3.7. Sources of secondary income of household sector in Croatia in 2001, by counties in millions of kunas Total Other social Health Social Unemploy. Child sources of Counties Pensions and various insurance welfare benefits allowances secondary transfers* income Croatia 17 990 1 593 981 731 2 415 12 013 35 723 Zagreb 944 117 41 44 144 767 2 056 Krapina-Zagorje 459 44 25 27 74 359 987 Sisak-Moslavina 753 58 57 35 101 504 1 508 Karlovac 596 46 44 28 66 388 1 168 Varazdin 613 61 43 27 105 473 1 322 Koprivnica- Krizevci 328 38 26 18 72 301 783 Bjelovar- Bilogora 387 35 35 24 83 331 895 Primorje-Gorski kotar 1 567 130 45 51 100 878 2 771 Lika-Senj 255 15 12 7 26 151 466 Virovitica- Podravina 261 24 33 18 64 233 632 Pozega-Slavonia 264 23 23 12 66 218 607 Slavonski Brod- Posavina 537 40 52 31 143 451 1 253 Zadar 639 47 40 27 97 436 1 286 Osijek-Baranja 1 181 100 89 65 226 874 2 535 Sibenik-Knin 500 31 65 17 71 322 1 006 Vukovar- Sirmium 723 44 45 30 144 533 1 520 Split-Dalmatia 1 899 147 104 89 303 1 270 3 812 Istria 900 90 27 32 68 562 1 679 Dubrovnik- Neretva 480 41 21 23 82 331 977 Meimurje 272 33 39 17 72 281 714 City of Zagreb 4 432 430 117 107 307 2 352 7 745 *Includes net transfers from abroad 144 Table B3.8. Sources of secondary income of household sector in Croatia in 2002, by counties in millions of kunas Total Other social Health Social Unemploy. Child sources of Counties Pensions and various insurance welfare benefits allowances secondary transfers* income Croatia 18 858 2 023 1 131 866 1 679 12 682 37 239 Zagreb 991 152 47 54 102 818 2 164 Krapina-Zagorje 482 55 29 29 55 378 1 028 Sisak-Moslavina 791 72 66 40 72 530 1 571 Karlovac 625 59 51 37 44 409 1 224 Varazdin 639 79 49 31 76 497 1 372 Koprivnica- Krizevci 345 47 30 22 54 317 816 Bjelovar- Bilogora 406 44 40 32 65 348 935 Primorje-Gorski kotar 1 647 164 52 52 67 930 2 913 Lika-Senj 270 19 14 10 18 159 490 Virovitica- Podravina 273 29 38 23 49 245 656 Pozega-Slavonia 277 29 26 14 46 229 621 Slavonski Brod- Posavina 563 48 60 40 102 472 1 286 Zadar 669 60 46 36 68 463 1 343 Osijek-Baranja 1 239 125 102 75 156 917 2 614 Sibenik-Knin 523 40 75 21 49 340 1 048 Vukovar- Sirmium 756 55 52 39 108 559 1 568 Split-Dalmatia 1 973 188 120 110 200 1 338 3 929 Istria 954 116 31 36 48 600 1 785 Dubrovnik- Neretva 502 53 24 31 51 349 1 010 Meimurje 290 42 45 20 55 298 750 City of Zagreb 4 642 547 135 113 194 2 487 8 117 *Includes net transfers from abroad 145 Table B3.9. Sources of secondary income of household sector in Croatia in 2003, by counties in millions of kunas Total Other social Health Social Unemploy. Child sources of Counties Pensions and various insurance welfare benefits allowances secondary transfers* income Croatia 19 919 2 039 1 225 836 1 612 13 534 39 163 Zagreb 1 062 156 51 51 95 883 2 298 Krapina-Zagorje 510 55 31 27 52 401 1 075 Sisak-Moslavina 845 70 71 33 70 564 1 652 Karlovac 655 58 55 36 41 432 1 277 Varazdin 680 78 53 31 71 530 1 444 Koprivnica- Krizevci 372 47 32 22 53 339 866 Bjelovar- Bilogora 432 44 43 27 63 370 979 Primorje-Gorski kotar 1 712 165 56 55 62 987 3 036 Lika-Senj 287 21 15 9 17 170 518 Virovitica- Podravina 301 27 41 23 49 263 705 Pozega-Slavonia 296 28 28 14 45 244 656 Slavonski Brod- Posavina 603 47 65 39 101 505 1 360 Zadar 722 62 50 35 65 501 1 434 Osijek-Baranja 1 315 122 111 73 150 977 2 747 Sibenik-Knin 555 40 81 22 47 364 1 108 Vukovar- Sirmium 801 54 56 42 106 596 1 656 Split-Dalmatia 2 080 193 130 102 196 1 432 4 133 Istria 1 001 115 34 37 45 641 1 873 Dubrovnik- Neretva 535 53 26 29 50 373 1 066 Meimurje 306 42 49 23 53 319 791 City of Zagreb 4 851 562 146 107 179 2 643 8 488 *Includes net transfers from abroad 146 Table B3.10. Sources of secondary income of household sector in Croatia in 2001, by counties, as % of total disposable income Total Other social Health Social Unemploy. Child sources of Counties Pensions and various insurance welfare benefits allowances secondary transfers* income Croatia 16.7 1.5 0.9 0.7 2.2 11.1 33.1 Zagreb 12.5 1.5 0.5 0.6 1.9 10.1 27.1 Krapina-Zagorje 14.2 1.3 0.8 0.8 2.3 11.1 30.6 Sisak-Moslavina 17.6 1.3 1.3 0.8 2.4 11.8 35.3 Karlovac 18.6 1.4 1.4 0.9 2.1 12.1 36.4 Varazdin 14.3 1.4 1.0 0.6 2.4 11.0 30.8 Koprivnica- 10.5 1.2 0.8 0.6 2.3 9.7 25.2 Krizevci Bjelovar- 12.5 1.1 1.1 0.8 2.7 10.7 29.0 Bilogora Primorje-Gorski 19.3 1.6 0.6 0.6 1.2 10.8 34.1 kotar Lika-Senj 20.2 1.2 1.0 0.6 2.1 11.9 36.9 Virovitica- 12.5 1.1 1.6 0.8 3.1 11.2 30.3 Podravina Pozega-Slavonia 14.3 1.3 1.2 0.7 3.6 11.8 32.8 Slavonski Brod- 15.7 1.2 1.5 0.9 4.2 13.2 36.7 Posavina Zadar 18.0 1.3 1.1 0.8 2.7 12.3 36.2 Osijek-Baranja 16.2 1.4 1.2 0.9 3.1 12.0 34.8 Sibenik-Knin 20.7 1.3 2.7 0.7 2.9 13.3 41.6 Vukovar- 17.8 1.1 1.1 0.7 3.6 13.1 37.4 Sirmium Split-Dalmatia 19.1 1.5 1.0 0.9 3.0 12.7 38.2 Istria 15.5 1.6 0.5 0.5 1.2 9.7 28.9 Dubrovnik- 16.7 1.4 0.7 0.8 2.8 11.5 33.9 Neretva Meimurje 10.9 1.3 1.6 0.7 2.9 11.3 28.8 City of Zagreb 18.5 1.8 0.5 0.4 1.3 9.8 32.3 *Includes net transfers from abroad 147 Table B3.11. Sources of secondary income of household sector in Croatia in 2002, by counties, as % of total disposable income Total Other social Health Social Unemploy. Child sources of Counties Pensions and various insurance welfare benefits allowances secondary transfers* income Croatia 16.6 1.8 1.0 0.8 1.5 11.2 32.9 Zagreb 12.3 1.9 0.6 0.7 1.3 10.1 26.8 Krapina-Zagorje 14.4 1.6 0.9 0.9 1.6 11.3 30.7 Sisak-Moslavina 17.7 1.6 1.5 0.9 1.6 11.8 35.1 Karlovac 18.1 1.7 1.5 1.1 1.3 11.8 35.4 Varazdin 14.0 1.7 1.1 0.7 1.7 10.9 30.0 Koprivnica- 10.6 1.4 0.9 0.7 1.7 9.7 25.0 Krizevci Bjelovar- 12.7 1.4 1.2 1.0 2.0 10.9 29.1 Bilogora Primorje-Gorski 19.3 1.9 0.6 0.6 0.8 10.9 34.1 kotar Lika-Senj 19.9 1.4 1.0 0.7 1.3 11.7 36.1 Virovitica- 12.8 1.3 1.8 1.1 2.3 11.5 30.8 Podravina Pozega-Slavonia 14.4 1.5 1.4 0.7 2.4 11.9 32.4 Slavonski Brod- 16.1 1.4 1.7 1.1 2.9 13.5 36.7 Posavina Zadar 18.1 1.6 1.2 1.0 1.8 12.6 36.4 Osijek-Baranja 16.4 1.6 1.4 1.0 2.1 12.1 34.5 Sibenik-Knin 20.4 1.6 2.9 0.8 1.9 13.3 41.0 Vukovar- 17.9 1.3 1.2 0.9 2.6 13.2 37.2 Sirmium Split-Dalmatia 18.9 1.8 1.1 1.1 1.9 12.8 37.7 Istria 15.5 1.9 0.5 0.6 0.8 9.7 29.0 Dubrovnik- 16.5 1.7 0.8 1.0 1.7 11.5 33.2 Neretva Meimurje 11.2 1.6 1.7 0.8 2.1 11.5 28.9 City of Zagreb 18.4 2.2 0.5 0.4 0.8 9.9 32.2 *Includes net transfers from abroad 148 Table B3.12. Sources of secondary income of household sector in Croatia in 2003, by counties, as % of total disposable income Total Other social Health Social Unemploy. Child sources of Counties Pensions and various insurance welfare benefits allowances secondary transfers* income Croatia 16.3 1.7 1.0 0.7 1.3 11.1 32.0 Zagreb 12.1 1.8 0.6 0.6 1.1 10.1 26.2 Krapina-Zagorje 14.4 1.6 0.9 0.7 1.5 11.3 30.3 Sisak-Moslavina 17.8 1.5 1.5 0.7 1.5 11.9 34.8 Karlovac 17.8 1.6 1.5 1.0 1.1 11.7 34.6 Varazdin 14.1 1.6 1.1 0.6 1.5 11.0 30.0 Koprivnica- 10.7 1.3 0.9 0.6 1.5 9.7 24.8 Krizevci Bjelovar- 12.5 1.3 1.2 0.8 1.8 10.7 28.2 Bilogora Primorje-Gorski 18.6 1.8 0.6 0.6 0.7 10.7 32.9 kotar Lika-Senj 18.8 1.4 1.0 0.6 1.1 11.2 34.1 Virovitica- 13.3 1.2 1.8 1.0 2.2 11.7 31.2 Podravina Pozega-Slavonia 14.9 1.4 1.4 0.7 2.3 12.3 33.0 Slavonski Brod- 16.3 1.3 1.8 1.0 2.7 13.6 36.6 Posavina Zadar 18.0 1.5 1.2 0.9 1.6 12.5 35.7 Osijek-Baranja 16.2 1.5 1.4 0.9 1.8 12.0 33.8 Sibenik-Knin 20.0 1.4 2.9 0.8 1.7 13.1 39.9 Vukovar- 17.9 1.2 1.3 0.9 2.4 13.3 37.0 Sirmium Split-Dalmatia 18.4 1.7 1.1 0.9 1.7 12.7 36.6 Istria 15.2 1.8 0.5 0.6 0.7 9.8 28.5 Dubrovnik- 16.6 1.6 0.8 0.9 1.5 11.6 33.1 Neretva Meimurje 10.9 1.5 1.7 0.8 1.9 11.4 28.2 City of Zagreb 17.5 2.0 0.5 0.4 0.6 9.5 30.6 *Includes net transfers from abroad 149 BACKGROUND PAPER #5 REGIONAL DISPARITIES IN LABOR MARKET PERFORMANCE IN CROATIA--ROLE OF INDIVIDUAL AND REGIONAL STRUCTURAL CHARACTERISTICS Xubei Luo The labor market performance in Croatia failed to keep pace with the moderately good overall macroeconomic development in the last few years. With a stagnant total employment rate, the large disparities in employment and earning across individual groups and regions have become one of the concerns for the long-term sustainable development of the economy. Using the Labor Force Survey (LFS) data 2002-2004, this paper studies the labor market performance in Croatia at the national and regional levels. The results show that both one's individual characteristics (including age, education and gender) and where he or she works plays a role in his or her employment and earning. The regional difference in employment and earning is reduced to a large extent when the difference in individual characteristics is accounted for. The simulations shed light on the effectiveness of the nation-wide education policy and regional specific labor market policy, and suggest that improving human capital endowment and adjusting labor market structure are both important to rebalance regional development and enhance total welfare. The paper is structured as follows. The first section describes the labor market performance in Croatia, and presents the disparities across individual groups and regions. The second section studies the determination of employment and earning at the national and regional levels, and examines the role of the individual characteristics and the regional structural characteristics. The third section simulates the effects of the nation-wide education policy and regional specific labor market policy on labor market development. The final section concludes. LABOR MARKET PERFORMANCE IN CROATIA The economy of Croatia has been growing quite satisfactorily in the period of 2002-2004 with a GDP growth rate of 4.4 percent.33 However, the labor market performance was stagnant ­ only 55 percent of the working-age population was employed, which lagged behind the employment objective of the Lisbon Agenda. The fruits of economic growth were not shared equally across the society ­ although wage increased slightly in nominal terms, with a long-term unemployment rate of about 60 percent,34 a large percentage of the active population was not able to be fully integrated / re-integrated into the society. The poverty rate stayed around 11 percent.35 33 Data source: Central Bureau of Statistics and Croatian Pension Insurance Fund, cited from Nestic (2006) 34 Long-term unemployment rate is measured as the share of the unemployed population who are unemployed for at least 12 months. 35 The poverty rate is subject to the poverty line of yearly consumption of 22145 kuna per equivalent adult. See Nestic (2006) for further discussion. 151 There exist large disparities across region / counties and individual groups. Others being equal, the individuals living in the Central and Eastern regions suffered disproportionately. In 2003, the GDP per capita of the richest county, City of Zagreb, is about three times that of the poorest county, county of Slavonski Brod-Posavina.36 The employment rate varies amongst the 21 counties, from less than 43 percent (County of Vukovar-Sirmium) to almost 70 percent (County of Krapina- Zagorje). Within many counties, the labor market performance differs at a more disaggregate level ­ for example, the county of Lika-Senj, unemployment is much more serious in the inland area than in the coastal area. The youth, the less educated, and the female face more difficulties in labor market. The unemployment rate of the youth (15-25 years old) was three times that of the other working-age groups. The earning level of the individuals with tertiary education is at least twice that of the individuals with basic education or less. Female face harder time in securing a job and in having a good salary compared with male... Limited Change in Labor Market Performance at the National Level The Croatian labor market performance is lagging behind many EU countries. The employment rate in Croatia37, which is about 55 percent, is 20 percent below the Lisbon target, which is 70 percent (Figure 5.1). The average employment rate of the EU 25 countries in 2002-2004 is about 63 percent, and that of the EU 15 countries is 64.5 percent.38 Figure 5.1: Total Employment Rate of EU Countries (2002-2004) 36 Data source: Lovrincevic and Mikulic (2006) 37 Based on the LFS questionnaires, employment is defined as the working-age population who worked for at least an hour in the reference week or had a job to return to if they did not work in the reference week. 38 Data source: Eurostat statistics. Website: http://epp.eurostat.cec.eu.int/portal/page?_pageid=1996,39140985&_dad=portal&_schema=PORTAL&screen =detailref&language=en&product=sdi_ed&root=sdi_ed/sdi_ed/sdi_ed_emp/sdi_ed1400 152 The labor market performance in Croatia did not change much in the recent years.39 Based on the LFS data, about 65 percent of the working-age population (15-65 years old) participated in the labor force, of which 55 percent were employed. Despite the relatively strong overall macroeconomic performance, the unemployment rate in Croatia, according to the definition of the International Labor Organization (ILO),40 hovered around 14-15 percent (Table 5.1). Table 5.1: Labor Force Participation Rate, Employment Rate, and ILO Unemployment Rate in Croatia (2002-2004) Labor force participation rate Employment rate ILO unemployment rate mean 95% confidence mean 95% confidence mean 95% confidence value interval value interval value interval 2002 64.88% [64.09% - 65.66%] 53.94% [53.15% - 54.72%] 15.04% [14.36% - 15.72%] 2003 64.73% [63.85% - 65.62%] 54.13% [53.25% - 55.00%] 14.51% [13.74% - 15.28%] 2004 65.90% [65.05% - 66.75%] 55.36% [54.55% - 56.17%] 13.98% [13.30% - 14.67%] Note: According to the International Labor Organization, the labor force participation rate is a measure of the proportion of a country's working-age population that engages actively in the labor market, either by working or looking for work; the employment rate (employment-to-population ratio) is defined as the proportion of a country's working-age population (15-65) that is employed. The unemployment rate is defined as the proportion of the labor force (working-age population that is employed or unemployed) that does not have a job and is actively looking for work. For the individuals who are employed, the distribution of their monthly earning (in kuna) also stayed relatively stable in the period of 2002-2004, although the mean level in nominal terms increased slightly (Figure 5.2).41 More than 90 percent of the individuals employed have a monthly 39 The unemployment rate in Croatia was about did not change much since the later 1990s (Rutkowski, 2002). Although the mean level of employment rate increases / unemployment rate decreases slightly in the period of 2002-2004, the changes are not significant in statistical senses. 40 The ILO unemployment rate, based on the LFS carried out by the Croatian Bureau of Statistics, is lower than the administrative unemployment data, which is maintained by the Croatia Employment Service. See appendix 1 for the definitions of employment and unemployment based on the LFS questionnaires. According to the statistics of the Central Bureau of Statistics and Croatian Pension Insurance Fund, the registered unemployment rate was around 20 percent in 2002-2004. The administrative data are affected by the incentive to register as "unemployed". On the one hand, some persons may register as unemployed in order to be eligible to unemployment benefits, pension, health insurance, and social assistance benefits, although they are not genuinely unemployed; on the other hand, the persons who are genuinely unemployed may not register if they are not eligible for the benefits associated wit the registration (Rutkowski, 2002). Overall, the administrative data may overestimate the actual magnitude of unemployment. In this study, we use the LFS data and follow the ILO unemployment definition. 41 Based on the LFS questionnaires, earning is measured as the usual net monthly earnings/salary on the main job in kuna. In Croatia, the average level of monthly earning is around 3013 kuna, 3205 kuna, and 3326 kuna respectively in 2002, 2003, and 2004 in nominal terms. In this study, we study the determination of employment and earning of the individuals with positive earning reported. The analysis may be subject to the three data constraint: 1. For the individuals who are employed with more than one job, their monthly earnings reported might under-represent the real income. 2. We have to neglect the individuals without positive earning although they reported as employed in the survey, which represent about 5 percent of the population employed. 3. No further information on in kind income is available from the survey. In this paper, all labor market indicators are measured for the individuals who are in the working age, and based on the LFS 2002- 2004 data if not otherwise specified. We inter-changeably use the wordings "wage" and "earning". 153 earning level less than 5000 kuna. The relatively high density of the distribution at the lower end may partly be resulted from the high incidence of the low-paid jobs. Figure 5.2: Distribution of Nominal Monthly Earning of the Individuals Who are Employed (2002-2004) .0003 .0002 Kernel density .0001 0 2000 4000 6000 8000 10000 Monthly earning 2002 2003 2004 Note: For practical purposes, we neglect the observations with extreme values of monthly earning (monthly earning > 15000 kuna), which represent less than 0.2 percent of the total sample. Large Difference in Labor Market Opportunities Across Individual Groups In order to increase the number of observations, we pool the six rounds of the LFS data 2002-200442 together, given significant changes of labor market performance in Croatia were absent in 2002- 2004. We use the pooled data as if they were a single sample from a larger survey, fielded over a longer period, given each consecutive round's sample in LFS can be considered as a random, equal- probability sub-sample of a larger sample taken in a single instance for the period as a whole for all practical purposes. 43 We find that, in 2002-2004, the key indicators, such as employment rate / unemployment rate and earning level, differ largely across age groups, education groups, gender, and regions. The youth group has the lowest employment rate/highest unemployed rate among other working-age population ­ the employment rate of the individuals who are 15-25 years old is less than 25 percent, which represents only a third of the employment rate of the individuals who are 25-50 years old; similarly, the unemployment rate of the individuals who are 15-25 years old is three times that of the age group 25-50 (Table 5.2). The youth employment rate in Croatia is about twice of the EU 42 The LFS were implemented in two six-month rounds each calendar year in 2002-2004. 43 See Appendix 2 and Munoz (2006) for the rational and techniques of data pooling. 154 average. 44 The youth Croatian who manages to find a job also suffers from low earnings ­ in average, the youth who are employed earn less than 2500 kuna per month, which is roughly 1000 kuna (or 30 percent in relative terms) less than those who are 25 years or older. With a high unemployment rate and a low earning level, the youth in Croatia faces a particularly difficult situation. Table 5.2: Employment Rate, Unemployment Rate and Monthly Earnings, by Age Groups in Croatia (2002-2004) Age range Employment rate ILO unemployment rate Monthly earning (kuna) (15, 25) 24.87% 35.93% 2449 [25, 50) 72.41% 12.86% 3316 [50, 65) 42.05% 8.47% 3463 As in many other countries, the individuals in Croatia with better education have higher probabilities to get employed, and the earning level monotonically increases with their levels of schooling (Table 5.3). The individuals with tertiary education not only have lower unemployment rate than those with lower education, but also have a much higher earning level if they get employed. For example, the unemployment rate of the individuals with university education and above is about 7.5 percent, and that of the individuals with basic education is 16 percent. The earning level of the individuals with university education and above is in average three times that of the individuals with no school or uncompleted basic education, and about 70 percent higher than the individuals with lower secondary vocational education. 45 Table 5.3: Employment Rate, Unemployment Rate and Monthly Earnings by Education Groups in Croatia (2002-2004) Employment Unemployment Monthly earning rate rate (kuna) No school or uncompleted basic education 30.79% 12.47% 1675 Basic education 37.02% 16.05% 2256 Lower secondary vocational education 61.56% 16.03% 3115 Higher secondary education 31.68% 17.52% 3247 Two-year post secondary education 73.50% 8.69% 4313 University education and above 82.80% 7.46% 5252 Gender is also an important dimension of inequality in Croatia. Female has a lower participation rate, a lower employment rate/ a higher unemployment rate, and a lower monthly earning than male (Table 5.4). 44 According to the Eurostat statistics of "total unemployment rate by age group", in 2002-2004, the average youth unemployment rate of the EU 25 countries is 18.7 percent and the EU 15 countries is 16.6 percent. Website http://epp.eurostat.cec.eu.int/portal/page?_pageid=1996,39140985&_dad=portal&_schema=PORTAL&screen =detailref&language=en&product=sdi_ed&root=sdi_ed/sdi_ed/sdi_ed_emp/sdi_ed1432 45 The low employment rate/high unemployment rate of the individuals with "high secondary education" (grammar school) may be linked to their low participation rate, which is less than 40 percent compared with 70-90 percent of the other individuals with secondary education and above. They represent less than 4 percent of the working-age population. 155 Table 5.4: Employment Rate, Unemployment Rate and Monthly Earnings by Gender in Croatia (2002-04) Labor force Monthly earning participation rate Employment rate Unemployment rate (kuna) Male 71.91% 61.35% 12.96% 3514 Female 58.63% 47.80% 16.36% 2978 Disparities in Labor Market Performance at the Regional and County Levels The variation of employment rate, unemployment rate, and earning level across age groups, education groups, and gender suggests that the opportunities in labor market that an individual faces is associated with his or her individual characteristics. However, the disparities in employment and earning, as those in GDP per capita, is also large across counties / regions in Croatia. Following the regional divides in the 1998 Living Standards Assessment (World Bank, 2000), the 21 counties in Croatia can be grouped into 5 regions for analytical purposes as follows: ˇ Region 1 ­ Central Region: Krapina-Zagorje, Sisak-Moslavina, Karlovac, Varazdin, Koprivnica-Krizevci, Bjelovar-Bilogora and Medimurje. ˇ Region 2 ­ Eastern Region: Virovitica-Podravina, Pozega-Slavonia, Slav. Brod-Posavina, Osijek-Baranja, and Vukovar-Sirmium. ˇ Region 3 ­ Zagreb Region: Zagreb County and Zagreb City. ˇ Region 4 ­ Adriatic North Region: Primorje-Gorski kotar, Lika-Senj, and Istria. ˇ Region 5 ­ Adriatic South Region: Zadar, Sibenik-Knin, Split-Dalmatia and Dubrovnik- Neretva. The human capital endowment, measured by years of school or highest attained level of schooling46, varies across regions. The Figure 5.3 show that the Central Region and Eastern Region have a higher percentage (about 40 percent) of working-age population with "no school or uncompleted basic education" or "basic education" compared with the other three regions (about 25 percent). Zagreb region has a disproportionately high percentage of working-age population with university education and above. Both indicators of education, "level of schooling" and "years of school", indicate that the Central Region and Eastern region are the least well endowed in human capital. 46 In the following sections, we simply call the "highest attained level of schooling" as the "level of schooling". 156 Figure 5.3: Distribution of Human Capital by Region (2002-2004) Figure 5.3a: Level of Schooling 60 55 54 51 % of the working-age population 49 48 50 40 30 31 30 21 20 19 20 12 9 8 8 9 8 10 6 5 6 6 6 7 4 4 4 4 4 4 5 5 0 No school or Basic education Low er Higher Tw o-year post University uncompleted secondary secondary secondary education and basic education vocational education education above education Central Region Eastern Region Zagreb Region Adriatic North Region Adriatic South Region Figure 5.3b: Years of School Adriatic South Region Adriatic North Region Zagreb Region Eastern Region Central Region 9 9.5 10 10.5 11 11.5 12 The Zagreb Region is in a leading position (Table 5.5) in two dimensions, including human capital endowment and earning level. The Eastern region, however, is lagging behind in many aspects, including a less satisfactory human capital endowment, a low employment rate / a high unemployment rate, and a low earning level. Although the participation rate and employment rate are high in the Central Region, the earning level and human capital endowment are low. Adriatic North Region and Adriatic South Region have similar level of earning and human capital, which are both slightly lower than those in the Zagreb Region, but the employment rate is lower in the Adriatic South Region. 157 Table 5.5: Summary Statistics on Labor Market Indicators by region in Croatia (2002-04) Active Monthly population Participation Employment Unemployment earning rate rate rate rate (kuna) Central Region 63.90% 72.30% 60.81% 11.47% 2806 Eastern Region 63.00% 61.11% 47.86% 19.93% 2826 Zagreb Region 66.09% 63.56% 55.91% 11.84% 3735 Adriatic North Region 64.38% 66.06% 58.47% 9.92% 3498 Adriatic South Region 62.28% 62.22% 48.94% 20.23% 3524 Croatia 64.01% 65.17% 54.47% 14.51% 3276 Although regions differ from each other, there is significant heterogeneity between counties. The active population rate, the participation rate, the employment rate, the unemployment rate, the mean monthly earning, and the mean level of years of school vary across counties within each region to a different extent (Table 5.6). ˇ On employment rate ­ three (out of a total of five) counties, including Vukovar-Sirmium, Slav. Brod-Posavina, and Osijek-Baranja, in the Eastern Region are among the four counties with the lowest employment rate (below 48 percent), while the county of Virovitica-Podravina, with an employment rate of 58 percent, has higher employment rate than all counties in Zagreb Region and Adriatic South Region. Five (out of a total of seven) counties, including Krapina-Zagorje, Varazdin, Bjelovar-Bilogora, Koprivnica-Krizevci, and Medimurje, in the Central Region are among the six counties with the highest employment rate (above 59 percent), while the county of Sisak-Moslavina, with an employment rate of 51 percent, ranks below all counties in the Zagreb Region and the Adriatic North Region and some counties of the Eastern Region. ˇ On earning level ­ all 11 counties with the lowest monthly earning level are in the Eastern Region and the Central Region. The average monthly earning of the City of Zagreb is almost 20 percent higher than the national average. ˇ On years of school ­ seven out of eight counties, except the county of Lika-Senj, with the lowest human capital endowment (lower than nine years of school) are in the Eastern Region and the Central Region. All four counties with average education higher than ten years of school are in the Zagreb Region, the Adriatic North Region, and the Adriatic South Region.There is a strong correlation (over 0.8) between years of school and level of earning for the 21 counties. There also exist heterogeneities within many counties (Table 5.7). For example, the distribution of earning for the individuals who are employed are the most unequal in the county of Lika-Senj and many counties in the Central Region and Eastern Region, where the average earning level is low. However, the two counties in the Zagreb Region, where the average earning level is the highest, have the most equal distribution. The graph 4 shows that, for example, in the County of Lika-Senj, the earning of the individuals who are employed clusters at two separate low levels, which is the so-called twin-peak distribution in econometrics; however, in the County of Zagreb, the earning concentrates at a relatively higher level. The large inequality in distribution of earning might aggravate the difficult situation that poor people face in the County of Lika-Senj. 158 Table 5.6: Summary Statistics on Labor Market Indicators by County in Croatia (2002-04) n Active Employ- Monthly population Participa- ment Unemploy- earning Years of County rate tion rate rate ment rate (kuna) school 1 County of Zagreb 65.36% 62.27% 52.91% 14.78% 3319 9.63 2 County of Krapina-Zagorje 63.89% 82.04% 69.62% 4.89% 3012 8.68 3 County of Sisak-Moslavina 62.94% 64.07% 50.70% 18.98% 3084 9.10 4 County of Karlovac 61.30% 70.73% 57.12% 15.53% 2967 9.13 5 County of Varazdin 66.19% 73.31% 64.22% 8.20% 2573 9.57 6 County of Koprivnica-Krizevci 63.89% 74.51% 62.85% 10.15% 2604 8.77 7 County of Bjelovar-Bilogora 63.04% 75.44% 63.86% 10.91% 2603 8.68 8 County of Primorje-Gorski kotar 64.65% 63.89% 56.63% 11.21% 3697 10.56 9 County of Lika-Senj 58.48% 88.10% 69.03% 8.68% 3033 8.51 10 County of Virovitica-Podravina 62.45% 70.59% 58.09% 13.96% 2341 8.59 11 County of Pozega-Slavonia 61.26% 65.36% 53.52% 13.90% 2670 8.49 12 County of Slavonski Brod-Posavina 60.96% 56.91% 46.45% 15.73% 2662 8.75 13 County of Zadar 63.04% 62.47% 49.38% 18.90% 3426 9.77 14 County of Osijek-Baranja 65.68% 62.08% 47.43% 22.89% 2958 9.57 15 County of Sibenik-Knin 56.91% 65.79% 45.91% 28.16% 3262 9.01 16 County of Vukovar-Sirmium 61.39% 56.75% 42.57% 24.07% 3106 8.67 17 County of Split-Dalmatia 63.22% 60.53% 48.14% 19.64% 3617 10.17 18 County of Istria 65.51% 64.15% 58.70% 8.38% 3330 9.91 19 County of Dubrovnik-Neretva 62.67% 65.34% 53.88% 17.42% 3519 10.22 20 County of Meimurje 65.95% 67.60% 58.98% 12.27% 2999 9.60 22 City of Zagreb 66.38% 64.06% 57.08% 10.71% 3895 11.49 Table 5.7: Inequality of Earning at the County Level in Croatia (2002-04) n number County Gini of earning 1 County of Zagreb 0.22 18 County of Istria 0.23 22 City of Zagreb 0.25 2 County of Krapina-Zagorje 0.25 8 County of Primorje-Gorski kotar 0.25 17 County of Split-Dalmatia 0.26 3 County of Sisak-Moslavina 0.26 15 County of Sibenik-Knin 0.27 19 County of Dubrovnik-Neretva 0.27 4 County of Karlovac 0.29 14 County of Osijek-Baranja 0.30 16 County of Vukovar-Sirmium 0.30 13 County of Zadar 0.30 20 County of Meimurje 0.31 11 County of Pozega-Slavonia 0.35 10 County of Virovitica-Podravina 0.36 12 County of Slavonski Brod-Posavina 0.36 7 County of Bjelovar-Bilogora 0.36 5 County of Varazdin 0.38 9 County of Lika-Senj 0.38 6 County of Koprivnica-Krizevci 0.38 Note: The Gini coefficient of earning is calculated for the individuals who are employed with positive earning. 159 Figure 5.4: The Distribution of Monthly Earning in the County of Zagreb and the County of Lika-Senj (2002-04) .0004 .0003 Kernel density .0002 .0001 0 0 2000 4000 6000 8000 10000 Monthly earning County of Zagreb County of Lika-Senj Determination of Employment and Earning The distribution of earning largely varies across regions, age groups, education groups, and gender. Figure 5.5 shows that the distribution of the earning in the Central Region and that in the Eastern Region are very similar; while the distribution of the other three regions clustered to a higher end. Figure 5.6 shows that for the individuals employed, the youth group has the lowest level of earning. Their situations are worse when the low youth employment rate is taken into consideration. Figure 5.7 shows that the level of earning monotonically increases with the level of schooling in a significant way. Figure 5.8 shows that female have lower earning than male. In this section, we will study the determination of employment and earning at the national and regional levels, decompose the roles of individual and regional structural characteristics in earning differentials for the individuals employed, and try to examine their different roles between regions. Figure 5.5: Distribution of Earning by Region in Croatia (2002-2004) .0004 .0003 Kernel density .0002 .0001 0 0 2000 4000 6000 8000 10000 Monthly earning Central Region Eastern Region Zagreb Region Adriatic North Region Adriatic South Region 160 Figure 5.6: Distribution of Earning by Age Group in Croatia (2002-2004) .0006 .0004 Kernel density .0002 0 0 2000 4000 6000 8000 10000 Monthly earning 16-24 years old 25-49 years old 50-64 years old Figure 5.7: Distribution of Earning by Education Group in Croatia (2002-2004) .0006 .0004 Kernel density .0002 0 0 2000 4000 6000 8000 10000 Monthly earning No school or uncompleted basic education Basic education Lower secondary vocational education Higher seccondary education Two-year post secondary education University education and above 161 Figure 5.8: Distribution of Earning by Gender in Croatia (2002-2004) .0004 .0003 Kernel density .0002 .0001 0 0 2000 4000 6000 8000 10000 Monthly earning Male Female Determination of Employment and Earning at the National and Regional Levels An individual chooses whether to join the work force or whether he or she gets a job depends on whether his or her reservation wage is greater than the wage offered by the employers. The individuals who choose to work are different from those who choose not to work. In other words, a random participation-in-the-labor-force assumption is unlikely to be true. In the survey, we can only observe the earnings of the individuals who work for the individuals who would have wages lower than their reservation wages may be less likely to choose to work. 47 The individual characteristics, captured by age, education, and gender, as well as the specific regional structural characteristics affect both the chances for the reservation wages and the offer wages; while some individual characteristics (in the jargon of economics, the identifying variables), such as marital status and being household head or not, which mainly affect an individual's incentives to work but not his or her competitiveness in job markets, may affect only the chances for the reservation wages (the chance of getting a job) but not the offer wages (the earning if he or she gets a job). Assuming that the monthly wage is a function of age, education, and gender, whereas the probability of employment (the likelihood of the wage being observed) is a function of marital status and whether the individual in question is a household head or not, and wage (via the inclusion of age, education, and gender, which we use to determine wage), we will use the Heckman estimation model instead of the Ordinary Least Square regressions to fit a wage model, for the sample of observed wages could be biased upward.48 To capture the non-linear effects of age on 47 It does not necessarily imply that the individuals who do not choose to work would have a lower level of earning should they choose to work. It can be the case that they could have even higher offer wages than those who actually choose to work ­ the former may have higher offer wages, but they choose not to work because they have even higher reservation wages. See Gronau (1974). 48 For further discussion on the methodology of the Heckman estimation, see Heckman (1976; 1979) and Greene (2003). 162 employment and earning, we introduce the variable "age square" into the equations.49 Based on the pooled LFS data 2002-2004 for the entire country, we estimate the equations as follows: Income equation: wage = 0 + 1age + 2 age 2 + 3education + 4 sex + i regional _ dummyi + 1 Selection equation (wage is observed if): 0 + 1age + 2 age 2 + 3education + 4 sex + 5 married + 6 household _ head + i regional _ dummyi + 2 > 0 where i =2,3,4,5, considering Region 1 (Central Region) as reference; and corr ( 1 , 2 ) = . The estimation of regression 1 in Table 5.8 shows that, at the national level, age plays a significantly positive role on employment and earning, but the positive effects decrease as age increases;50 human capital, here measured by years of school, has positive effects on employment in a significant way, and the earning level of an individual monotonically increases with his or her education endowment; female have a harder time to get a job than male, and face a lower level of earning if they manage to get employed nonetheless, others being equal; being married or household head increases an individual's probability to be employed, partly because of his/her stronger incentives to seek employments. The estimations of regression 2 show that all regional dummies are significant. It suggests that the probability of employment and the level of earning significantly differ across regions for individuals with similar characteristics in age, education, and gender. More specifically, on employment, in comparison with the Central Region (where the highest employment rate is the highest), individuals with similar individual characteristics have lower chance to get employed in the other four regions ­ among these four regions, the individuals have the lowest chance to get employed in Adriatic South Region, and the highest chance in Adriatic North Region. On earning, in comparison with the Central region (where the monthly earning level is one of the lowest, similar as the Eastern Region), the individuals who are employed have much higher earning in Zagreb Region, Adriatic North Region, and Adriatic South Region. Female have lower probability to get a job than male in every region ­ they have the hardest condition in the Eastern Region and the best condition in the Zagreb Region; for those who manage to get a job, female workers earn less than male, particularly in the Adriatic North Region and the Adriatic South Region. 49 We had also tried to introduce the "years of school square" into the equations. However, the results show that effects of education on employment and earning are not significantly non-linear. 50 Based on the results of Table 5.8, the positive effects of age on employment reach the maximum level before those on earning reach the maximum level. Instead of arguing the exact golden age for employment/earning, we would like to suggest that one's earning level (if he or she is employed) may continue to increase for a certain period after his or her age has the maximum positive impacts on employment, because experience counts. The studies at the regional level show that, given the difference in job market structure and in labor supply, the magnitude of the effects of experience varies across regions. 163 Table 5.8: Determination of Employment and Earning at the National Level in Croatia (2002-04) using years of school as an indicator of human capital Regression 1 Regression 2 Estimation of income equations (dependant variable: monthly earning (in thousand kuna)) Coef. t-stat. Coef. t-stat. age 0.09 12.57 0.09 12.45 age*age 0.00 -9.32 0.00 -9.24 years of school 0.34 103.1 0.32 94.17 sex (male=0, female=1) -0.64 -34.94 -0.67 -36.66 Regional dummies (Central Region as reference) Eastern Region 0.06 2.44 Zagreb Region 0.68 27.25 Adriatic North Region 0.60 20.41 Adriatic South Region 0.50 18.29 constant -2.68 -16.1 -2.69 -16.45 Estimation of selection equations Coef. t-stat. Coef. t-stat. age 0.23 85.78 0.23 85.48 age*age 0.00 -91.1 0.00 -90.73 years of school 0.05 28.14 0.05 29.83 sex (male=0, female=1) -0.24 -21.44 -0.23 -21.26 Regional dummies (Central Region as reference) Eastern Region -0.21 -15.03 Zagreb Region -0.20 -14.53 Adriatic North Region -0.10 -5.81 Adriatic South Region -0.30 -20.35 married 0.21 17.38 0.21 17.3 household head 0.20 15.54 0.20 15.45 constant -4.58 -91.26 -4.47 -88.11 Log likelihood -121325 -120545 Number of observations 78148 78148 Note: All coefficients are significant in 1%. Given the large variation of the labor market structure and performance in different regions, we relax the implicit assumption on the homogeneity of the roles of the individual characteristics in employment and earning across regions, and estimate the Heckman selection and income equations separately. The results of Table 5.9 show that, in each region, the variables of age, education, gender, marital status, and household head, have their significant expected signs. However, their marginal effects on employment and earning differ. In each region, the positive effects of age on employment reach the maximum level when age is about 40, while those on earning vary. It suggests that the rewards to experience differ across regions depending on the supply and demand in the job market. The return of education also differs to a large extent across regions. Among the five regions, the return of an additional year of school on employment is the highest in the Eastern Region and the lowest in the Central Region. However, for the individuals who are employed, their return of education for each additional year of school on earning is the highest in the Zagreb Region, closely followed by the Central Region, and the lowest in the Eastern Region, others being equal (Table 5.10). 164 Table 5.9: Determination of Employment and Earning at the Regional Levels --using years of school as an indicator of human capital Central Region Eastern Region Zagreb Region Adriatic West Region Adriatic South Region Coefficient t-statistics Coefficient t-statistics Coefficient t-statistics Coefficient t-statistics Coefficient t-statistics Estimation of income equations (dependant variable: monthly earning (in thousand kuna)) age 0.124 9.73 0.094 6.62 0.060 3.4 0.122 4.56 0.044 2.18 age*age -0.001 -8.46 -0.001 -4.98 0.000 -1.63 -0.001 -3.6 0.000 -0.93 years of school 0.333 56.61 0.283 40.01 0.344 49.38 0.302 27.15 0.300 30.71 sex (male=0, female=1) -0.628 -19.26 -0.571 -14 -0.525 -15.17 -0.900 -15.95 -0.810 -16.18 constant -3.287 -11.79 -2.252 -6.9 -1.944 -5.01 -2.418 -4.06 -1.018 -2.2 Estimation of selection equations age 0.198 39.2 0.198 34.33 0.280 46.89 0.263 33.27 0.230 35.93 age*age -0.003 -41.25 -0.002 -35.68 -0.004 -50.54 -0.003 -35.65 -0.003 -37.89 years of school 0.022 6.33 0.079 21.13 0.058 15.51 0.047 8.9 0.072 16.61 sex (male=0, female=1) -0.218 -9.93 -0.412 -16.16 -0.148 -6.69 -0.215 -6.9 -0.196 -7.24 married 0.235 9.97 0.241 8.97 0.239 9.26 0.134 3.73 0.186 6.28 household head 0.104 4.01 0.198 6.79 0.183 7.27 0.188 5.3 0.317 9.99 constant -3.655 -38.27 -4.464 -41.36 -5.563 -49.4 -5.040 -33.31 -5.094 -41.38 Log likelihood -31949.15 -23680.53 -27311.09 -15275.03 -21660.24 Number of observations 19519 16479 18379 9355 14416 165 Table 5.10: Ranking of Impacts of Years of School on Earning and on Employment Across Regions Region Impacts on earning Region Impacts on employment Eastern Region 0.283 Central Region 0.022 Adriatic South Region 0.300 Adriatic North Region 0.047 Adriatic North Region 0.302 Zagreb Region 0.058 Central Region 0.333 Adriatic South Region 0.072 Zagreb Region 0.344 Eastern Region 0.079 The difference in the impacts of education on employment and on earning at the regional level may be explained by the difference in the distribution of human capital and the difference in employment structure. For example, the employment rate in Central Region is high. The large share of the low-paid manufacturing job may lower the comparative attractiveness of the better educated individuals in the labor market when they seek employment; while for those who do manage to secure a job, they may have a relatively high earning given the scarcity of the human capital (as the statistics show, the average years of school in Central Region is relatively low compared with the national average). In Eastern region, the large impact of years of school on employment may reflect the low-demand from the job market, which corresponds to the low employment rate there. Only those relatively well educated have a better chance to get a job; while the tight competition for a job on the supply side lowers the market clearing level of earning for those who are employed. In the Adriatic South Region, where employment rate is relatively low and a large part of jobs are seasonal, the individuals with better education may have better chance to find a job. Trying to study the effects of different levels of education in addition to those of "years of school", we use the "level of schooling" as the indicator of human capital to reexamine the determination of the employment and earning (Tables 11 and 12). 51 The results using "level of schooling" offer support to those using "years of school" as the indicator of human capital. At the national level, the selection equations indicate that individuals with vocational secondary education, two-year post secondary education, and university education and above have significantly higher chance to get a job. Individuals with basic education do not have better chance to get employed compared with those with no education or uncompleted basic education.52 It seems to suggest that, in the Croatian economy, there is a certain threshold for the level of schooling to be effective, which is likely to be secondary education, below which the return is low. The income equations suggest that each additional level of schooling has significantly positive impacts on earning in all regions. For the individuals who are employed, others being equal, those with secondary education have higher earning than those with basic 51 In the analysis, we have six levels of education, including "no school or uncompleted basic education", "basic education", "lower secondary vocational education", "higher secondary education (grammar school)", "two-year post secondary education", and "university education and above". To compare the effects of levels of education on employment and earning, we use "no school or uncompleted basic education" as the reference. 52 The results show that the individuals with grammar school as their highest level of schooling do not have advantage in getting a job compared with the others. We will not examine this in details because of two reasons: 1. very few individuals belong to this category; 2. the individuals who belong to this category have low participation rates. 166 education or less; and those with tertiary education (in particular university and above) have still higher earning than those with secondary education. Table 5.11: Determination of Employment and Earning at the National Level in Croatia (2002-04) - using level of schooling as indicator of human capital Regression 3 Regression 4 Estimation of income equations (dependant variable: monthly earning (in thousand kuna)) Coef. t-statistics Coef. t-statistics age 0.113 15.71 0.109 15.48 age*age -0.001 -13.24 -0.001 -13.08 Level of schooling ("No school or uncompleted basic education" as reference) Basic education 0.649 12.96 0.571 11.57 Lower secondary vocational education 1.645 34.07 1.441 30.03 Higher secondary education 1.986 29.43 1.726 25.83 Two-year post secondary education 2.815 49.13 2.577 45.24 University education and above 3.781 69.98 3.494 64.72 sex (male=0, female=1) -0.660 -36.11 -0.688 -38.25 Regional dummies (Central Region as reference) Eastern Region 0.065 2.56 Zagreb Region 0.729 29.46 Adriatic North Region 0.634 21.62 Adriatic South Region 0.538 19.74 constant -0.661 -4.25 -0.743 -4.89 Estimation of selection equations Coef. t-statistics Coef. t-statistics age 0.220 81.64 0.220 81.28 age*age -0.003 -87.7 -0.003 -87.31 Level of schooling ("No school or Uncompleted basic education" as reference) Basic education -0.041 -1.8 -0.034 -1.48 Lower secondary vocational education 0.278 12.69 0.307 13.86 Higher secondary education -0.183 -6.02 -0.141 -4.59 Two-year post secondary education 0.564 19.11 0.606 20.35 University education and above 0.615 22.74 0.662 24.08 sex (male=0, female=1) -0.213 -19.15 -0.212 -18.96 Regional dummies (Central Region as reference) Eastern Region -0.213 -14.92 Zagreb Region -0.205 -14.5 Adriatic North Region -0.109 -6.38 Adriatic South Region -0.316 -21.11 married 0.201 16.44 0.201 16.39 household head 0.197 15.46 0.197 15.35 constant -4.062 -78.35 -3.927 -74.79 Log likelihood -120705 -119837 Number of observations 78148 78148 Note: All coefficients are significant in 1 percent, except the dummy "basic education", which is not significant at 5 percent in the selection equations. 167 At the regional level, the positive effects of having tertiary education (including two-year post secondary education and university education and above) are the largest in the Eastern Region and the Zagreb Region, and the lowest in the Central Region. We tend to argue that, the reasons why the return on tertiary education is high in the Eastern Region and Zagreb Region are different ­ in Eastern region, the employment rate is low (the unemployment rate is high), only the most qualified has a higher chance to get a job; while in Zagreb, the high demand for jobs that require high skills (for example, the jobs in financial sector) favor the individuals with better education. The individuals with low-education may face difficulties in finding a job because of the mismatch of skills. In Central Region, to the contrary, the high demand for jobs that require lower skill (for example, the low-paid jobs in manufacturing sector) favors the individuals with less education, who may have lower reservation wage. The return on education, in particular on tertiary education, is the highest for those who are employed in Central Region, followed by Zagreb Region. 168 Table 5.12: Determination of Employment and Earning at the Regional and National Levels Using Level of Schooling as Indicator of Human Capital Central Region Eastern Region Zagreb Region Adriatic North Region Adriatic South Region Coefficient t-statistics Coefficient t-statistics Coefficient t-statistics Coefficient t-statistics Coefficient t-statistics Estimation of income equations (dependant variable: monthly earning (in thousand kuna)) age 0.144 11.9 0.119 8.87 0.070 4.24 0.130 5.47 0.051 2.64 age*age -0.002 -11.23 -0.001 -7.81 -0.001 -2.68 -0.001 -4.58 0.000 -1.69 Basic education 0.486 6.98 0.405 4.59 0.612 3.61 0.374 1.75 0.711 4.31 Lower secondary vocational education 1.456 21.34 1.231 14.26 1.325 7.94 1.278 6.12 1.420 9.07 Higher secondary education 1.845 14.15 1.434 11.06 1.768 9.79 1.293 5.07 1.531 7.83 Two-year post secondary education 2.667 28.48 2.560 22.81 2.339 13.22 2.241 9.82 2.568 15 University education and above 3.995 44.48 3.104 29.39 3.473 20.26 3.073 13.85 3.202 18.99 sex (male=0, female=1) -0.651 -20.12 -0.589 -14.81 -0.563 -16.45 -0.909 -16.38 -0.825 -16.53 constant -1.235 -4.8 -0.597 -2.03 0.536 1.37 -0.305 -0.56 0.953 2.19 Estimation of selection equations age 0.193 37.89 0.194 33.38 0.265 43.55 0.252 31.37 0.219 33.68 age*age -0.002 -40.2 -0.002 -35.5 -0.003 -47.38 -0.003 -33.9 -0.003 -36.01 Basic education -0.082 -2.18 -0.056 -1.27 0.309 4.31 0.199 2.36 -0.019 -0.31 Lower secondary vocational education 0.089 2.39 0.367 8.54 0.805 11.72 0.523 6.45 0.351 5.99 Higher secondary education -0.489 -7.71 0.069 1.05 0.253 3.26 -0.035 -0.34 0.074 0.98 Two-year post secondary education 0.356 6.08 0.874 12.81 0.969 12.33 0.713 7.36 0.736 10.51 University education and above 0.374 6.57 1.101 16.75 1.013 13.94 0.710 7.72 0.831 12.2 sex (male=0, female=1) -0.198 -8.93 -0.400 -15.53 -0.106 -4.7 -0.178 -5.64 -0.181 -6.63 married 0.228 9.57 0.247 9.15 0.202 7.75 0.123 3.4 0.185 6.2 household head 0.105 4.01 0.205 7.01 0.176 6.94 0.186 5.22 0.322 10.13 constant -3.342 -34.99 -3.751 -34.1 -5.289 -41.49 -4.728 -28.66 -4.353 -34.03 Note: "No school or uncompleted basic education" as reference. 169 Role of Individual and Regional Structural Characteristics in Wage Differentials If the labor market structures are similar across regions, individuals with similar characteristics will have similar employment and earning no matter where he or she lives.53 In economic terms, regional discrimination in earning can be said to exist whenever the relative earning of one region exceeds the relative wage that would have prevailed if workers in two regions were paid according to the same criteria based on their individual characteristics. 54 In this section, we will study the relative importance of the role of individual characteristics and regional structural characteristics in wage differential between regions by using the Oaxaca decomposition methods (Oaxaca, 1973; Oaxaca and Ransom, 1994). We apply the wage structure of the reference region to simulate the earning level of the region in question, and decompose the effects on regional wage differential.55 The results of Table 5.13 show that, taking the wage structure of the Central Region as reference, individual characteristics, measured by age, education, and gender, account for about two-thirds of the wage difference between the Eastern Region and the Central Region, and the other characteristics account for the rest one-third; however, individual characteristics account for roughly one-third of the wage difference between the Zagreb Region and the Central Region, between the Adriatic North Region and the Central Region, and between the Adriatic South Region and the Central Region; while the other characteristics accounts for two-thirds.56 It suggests the relative similarity of wage structure between the Central Region and Eastern Region and the large difference in wage structure between the other pairs of regions, which corresponds to the statistics on the regional difference in the labor market structure and performance. Table 5.13: Decomposition of the Effects of Individual and Other Characteristics on Regional Earning Differentials Effects of individual characteristics Effects of other Mean (age, education, and gender) on characteristics on mean monthly mean earning differential between earning differential between earning the region in question and Central the region in question and (kuna) Region Central Region Central Region 2806 ... ... Eastern Region 2826 63.99% 36.01% Zagreb Region 3735 34.55% 65.45% Adriatic North Region 3498 28.45% 71.55% Adriatic South Region 3524 30.70% 69.30% 53 In this study, we focus on the employment and earning differentials at the regional / county level, and leave the interesting discussions on the rural-urban divide for future work. 54 Here, we consider age, education, and gender as the major indicators of individual characteristics, and have to neglect other characteristics, such as entrepreneurship, due to information constraints. In other words, we may underestimate the effects of individual characteristics on wage differentials, and overestimate those of the regional structural characteristics because the inclusion of the effects of the unobservable individual characteristics into the regional ones 55 See appendix 3 for detail discussion on the methodologies of the decomposition of the effects of individual characteristics and regional characteristics on wage differentials. 56 The relative importance of the role of individual and regional structural characteristics in wage differential between each pairs of regions is stable even if we change the reference region. In other words, the individual characteristics play a larger role than regional structural characteristics in the wage differential between the Central Region and the Eastern Region; while the regional structural characteristics play a larger role than individual characteristics in the wage differential between the other pairs of regions. Results are available upon request. 170 Effects of Nation-wide Education Policy and Regional-specific Labor Market Policy An individual may earn more if he or she has better education, or if he or she moves to another region where the labor market better rewards his or her talents, others being equal. In this section, we will simulate the hypothetic effects of the nation-wide education policies and the regional- specific labor market policies on regional wage differentials. Effects of Nation-wide Education Policy The estimations in the previous sections show that education plays a positive role in employment and earning, and the return on education varies across regions. If each working-age individual who is employed has one more year of school, what would be the employment rate and earning level in each region? Based on the selection estimations and the income estimations at the regional level, we will simulate the effects of such hypothetic one-additional-year-of-school national-wide education policy on regional labor market. The simulations include two steps: 1. For employment, taking age, age square, years of school, gender, marital status, and household head as determinants, we estimate the determination of employment in each region; then assuming each working-age individual has an additional year of school and applying the coefficients estimated, we simulate the effects of this nation-wide policy on employment. 2. For earning, taking age, age square, years of school, and gender as determinants, we estimate the determination of earning for the individuals who are employed in each region; then assuming each working-age individual has an additional year of school and applying the coefficients estimated, we simulate the effects of this nation-wide policy on earning. The simulation results of Table 5.14 show that, if the working-age population has one more year of education, the employment rate will increase 2.5 ­ 5.5 percent and the earning level will increase 8 ­ 10.5 percent in five regions. The positive effects on employment are the largest in the Eastern Region, which suggests that for the given wage structure in this lagging region, the shortage of skilled labor (measured by years of school) is an important constraint. Although such education policy has large effects on earning for those who are employed in the Central Region, its effects on employment may be limited. It may suggest that a general increase in human capital endowment at this stage may not much increase the job demand in Central region, but it may, to a large extent, enhance the labor productivity of those who are employed, which leads to the increase in wage. If we consider the product of the increase in employment and the increase in earning as a rough measure of the total effects of the education policy, such nation-wide one-year-of-school-increase policy will have larger effects in the Eastern Region than elsewhere, which may contribute to balance the regional development. In addition, if being unemployed (rather than having low salary) is one of the major reasons for being poor, the large effects on employment in the Eastern Region could be further pro-poor. 171 Table 5.14: Simulations of the Effects on Employment and Earning of Nation- wide Education Policy by Region % increase in employment rate due to Estimated Simulated an additional year of employment rate employment rate school Central Region 60.89% 62.48% 2.54 Eastern Region 47.74% 50.50% 5.47 Zagreb Region 55.67% 58.42% 4.71 Adriatic North Region 58.39% 60.51% 3.50 Adriatic South Region 48.81% 51.45% 5.13 % increase in earning Estimated mean Simulated mean due to an additional year monthly earning monthly earning of school Central Region 2785 3111 10.47 Eastern Region 2811 3100 9.32 Zagreb Region 3703 4046 8.48 Adriatic North Region 3495 3795 7.91 Adriatic South Region 3524 3827 7.90 Effects of Regional Specific Labor Market Policy The estimation results in section 2 show that, after controlling for the individual characteristics, including age, education, and gender, the employment and earning level still differ to a large extent across regions. If the working-age individuals who are employed in one region face the wage structure of another region, others being equal, what would be the employment rate and earning level? Using the similar methodologies as those for the simulation of the nation-wide education policy, we will study the effects of such hypothetical regional specific policy on employment and earning, assuming the labor market structure in one region could be duplicated in another.57 Similarly, the simulations include two steps: 1. For employment, taking age, age square, years of school, gender, marital status, and household head as determinants, we estimate the determination of employment in each region; then assuming the labor market structure of the region in question is adjusted to the one of the reference region and applying the coefficients estimated for that reference region, we simulate the effects on employment in the region in question. 2. For earning, taking age, age square, years of school, and gender as determinants, we estimate the determination of earning for the individuals who are employed in each region; then assuming the labor market structure of the region in question is adjusted to the one of the reference region and applying the coefficients estimated for that reference region, we simulate the effects on earning for those who are employed. 57 The feasibility of the duplication of the wage structure of one region might be subject to further discussion. One objective here is to study the role of regional specific characteristics in employment and earning determination. We hope that the results can, to some extent, shed light on the potential effects of inter-regional migration ­ even if the entire labor market structure of one region will not duplicate that of the other, what if some individuals move to another region? 172 The results of Table 5.15 show that, measured by employment, the labor market is the most dynamic in the Central Region, followed by the Adriatic North Region, and is the least dynamic in the Adriatic South Region and the Eastern Region. Measured by earning, the individuals employed are the least well rewarded by the wage structures in the Central Region and the Eastern Region, and are the best rewarded in the Zagreb Region, closely followed by the Adriatic North Region and the Adriatic South Region. Taking both employment and earning into consideration, the labor market structure in Eastern Region is the least satisfactory one. We tend to argue that the low human capital endowment and the unsatisfactory labor market structure both lead to its backwardness in development. The dynamic labor market structure in the Adriatic North Region might be one of the important factors that contribute to its prosperity. The adjustment of the labor market structure of the Eastern Region towards those of the other regions may help to improve the dynamism of its labor market. Some measures that lower the barriers of labor mobility across regions may also help the individuals who migrate to get jobs that better reward their talents, and hence increase the nation-wide welfare level. 173 Table 5.15: Simulations of the Effects on Employment and Earning of regional Specific Labor Market Policies by Region The labor market structure of Central Region as reference % change in Estimated Estimated monthly simulated employment % change in simulated monthly monthly employment rate earning rate employment rate earning earning Central Region 60.89% 2785 60.89% 0.00 2785 0.00 Eastern Region 47.74% 2811 59.95% 20.36 2903 3.16 Zagreb Region 55.67% 3703 62.07% 10.31 3227 -14.75 Adriatic North Region 58.39% 3495 61.63% 5.27 3076 -13.62 Adriatic South Region 48.81% 3524 61.58% 20.75 3180 -10.83 The labor market structure of Eastern Region as reference % change in Estimated Estimated monthly simulated employment % change in simulated monthly monthly employment rate earning rate employment rate earning earning Central Region 60.89% 2785 48.57% -25.37 2701 -3.12 Eastern Region 47.74% 2811 47.74% 0.00 2811 0.00 Zagreb Region 55.67% 3703 51.43% -8.23 3088 -19.91 Adriatic North Region 58.39% 3495 50.56% -15.47 2959 -18.09 Adriatic South Region 48.81% 3524 50.37% 3.11 3056 -15.34 The labor market structure of Zagreb Region as reference % change in Estimated Estimated monthly simulated employment % change in simulated monthly monthly employment rate earning rate employment rate earning earning Central Region 60.89% 2785 52.84% -15.25 3241 14.07 Eastern Region 47.74% 2811 51.74% 7.73 3367 16.50 Zagreb Region 55.67% 3703 55.67% 0.00 3703 0.00 Adriatic North Region 58.39% 3495 54.57% -6.98 3553 1.64 Adriatic South Region 48.81% 3524 54.45% 10.37 3659 3.68 The labor market structure of Adriatic North Region as reference % change in Estimated Estimated monthly simulated employment % change in simulated monthly monthly employment rate earning rate employment rate earning earning Central Region 60.89% 2785 57.21% -6.43 3237 13.97 Eastern Region 47.74% 2811 56.13% 14.94 3363 16.41 Zagreb Region 55.67% 3703 59.12% 5.84 3623 -2.21 Adriatic North Region 58.39% 3495 58.39% 0.00 3495 0.00 Adriatic South Region 48.81% 3524 58.29% 16.27 3607 2.28 The labor market structure of Adriatic South Region as reference % change in Estimated Estimated monthly simulated employment % change in simulated monthly monthly employment rate earning rate employment rate earning earning Central Region 60.89% 2785 47.20% -29.02 3157 11.77 Eastern Region 47.74% 2811 46.42% -2.83 3280 14.30 Zagreb Region 55.67% 3703 50.12% -11.06 3542 -4.55 Adriatic North Region 58.39% 3495 49.21% -18.65 3417 -2.28 Adriatic South Region 48.81% 3524 48.81% 0.00 3524 0.00 174 Conclusions The labor market performance in Croatia in 2002-2004 varies across individual groups and across regions. Both individual characteristics and regional characteristics play important roles in the determination of employment and earning. The youth, the less well-educated, and female face more difficulties in getting a job with a good salary. A large part of the difference in regional labor market performance is associated with the difference in the human capital endowment. The combination of the favorable human capital endowment and the well rewarding labor market structure contributes to the good economic performance of the Zagreb Region, followed by the Adriatic North Region; while the combination of the less satisfactory situation of these two results in the backwardness of the development in the Eastern Region. Central Region has the highest employment rate, but the low human capital endowment and the prevalence of the low-paid manufacturing jobs keeps the earning level to the low end similar to that in the Eastern Region. The individuals have relatively good salaries in Adriatic South Region if they manage to get employed, but the less dynamic labor market and the seasonal nature of many job-demands there lead to the high unemployment. Both nation-wide education policies and regional specific labor market policies are effective in enhancing employment and earning. The entire country will benefit from an improvement in human capital endowment and an appropriate adjustment of labor market structures, and the lagging Eastern Region in particular will benefit more than the other regions. 175 References Green W.H. (2003), Econometric Analysis, 5th ed. Upper Saddle River, NJ: Prince-Hall. Gronau R. (1974), "Wage comparisons: a selectivity bias", Journal of Political Economy92, pp.1119- 1155. Heckman J. (1976), "The common structure of statistical models of truncation, sample selection, and limited dependent variables and a simple estimator for such models", The Annals of Economic and Social Measurement 5, pp.475-492. Lovrincevic Z. and D. Mikulic (2006), "Regional development and social indicators in Croatia", Croatia Living Standards Assessment background paper, mimeo, The World Bank. Munoz J. (2006), "Visit to Zagreb", Croatia Living Standards Assessment background paper, mimeo, The World Bank. Nestic D. (2006), "Recent economic developments in Croatia", Croatia Living Standards Assessment background paper, mimeo, The World Bank. Oaxaca R. L. (1973), "Male-female wage differentials in urban labor markets", International Economic Review, Vol.14, No.3, pp.693-709. Oaxaca R.L. and Ransom M.R. (1994), "On discrimination and the decomposition of wage differentials", Journal of Econometrics 61, pp.5-21. Rutkowski J. (2002), "Toward a dynamic labor market and employment growth", Croatia Country Economic Memorandum background paper, mimeo, The World Bank. 176 APPENDIX 1 DEFINITION OF EMPLOYMENT BASED ON THE LFS QUESTIONNAIRES Definition of Employment (15,65) years old? no Excluded from the analysis Entire population yes Current activity status? (Q22 ) Worked Did not explicitly work (Q22: A1-7) (Q22: A 8-12) Worked at least an Worked at least an hour last week for hour last week? (Q23) payment or family gain? (Q23 ) no yes no yes A job to return? Employ=1 Employ=0 Employ=1 (Q25 ) no yes Employ=0 Employ=1 Definitions of Unemployment (ILO / broad) (15,65) years old? no Excluded from the analysis Entire population Employ = 1? yes Employ=1 no If you were offered a job now, would you be able to Economically no start working in the next 2 weeks? (Q83) inactive Employ=0 & yes Unemploy=0 Would you like to work, provided you could find an adequate job? (Q68 ) Economically inactive no Employ=0 & Unemploy=0 yes Broad or ILO Unemploy=1 Broad unemploy=1 Did you look for work in the last 4 weeks? (Q69 of 2002) no ILO unemploy=0 yes Broad unemploy=1 ILO unemploy=1 Note: Qx stands for the question number x in the LFS questionnaire 2002; Ax the answer of the relevant question. 177 APPENDIX 2 RATIONAL FOR 2002-2004 LFS DATA POOLING58 The samples from six consecutive rounds of the LFS in 2002-2004 can be pooled together with their current design, as if they were a single sample from a larger survey, fielded over a longer period. The reason is that each round's sample can ­ for all practical purposes ­ be considered as a random, equal- probability sub-sample of a larger sample taken in a single instance for the period 2002-04 as a whole.59 Munoz (2006) suggests that the effect of pooling on sampling weights is trivial. Although in theory the weights could require a more elaborate calculation, in practice they are unlikely to differ significantly from the weights currently used for the analysis of individual rounds, scaled by overall constants that account for the larger total size of the pooled samples. In fact, as long as the only parameters estimated from the pooled samples are proportions, averages and ratios (rather than totals), the analysis can be conducted with the pooled databases without any re-weighing at all. The LFS implements two six-month rounds each calendar year, but both annual rounds share the same Primary Sampling Units (PSUs). Because of clustering effects, the precision obtained from pooling the two rounds of the same year is less than if the two rounds had been fielded on different PSUs. Pooling consecutive years of the LFS, however, the precision expected would be roughly that of survey with twice or three times as many PSUs as each annual survey: sampling errors about 30 percent and 40 percent smaller, respectively To illustrate the actual figures involved, we estimated the sampling errors and confidence intervals for three basic indicators from the LFS: 60 ˇ The employment rate ˇ The unemployment rate ˇ The average monthly earnings from the primary job, and At three levels of temporal aggregation: ˇ One round (July to December 2004) ˇ Two pooled rounds (the whole year 2004) and ˇ Six pooled rounds (all three years 2002 to 2004), and four levels of geographic aggregation: ˇ Each of Croatia's 21 counties ˇ Five regions (Central Region, Eastern Region, Zagreb Region, Adriatic North Region and Adriatic South Region) ˇ Three major zones (Central & Eastern Region, Zagreb Region, and Adriatic Region) and ˇ The entire country. 58 See Juan Munoz's background report (2006). 59 Munoz (2006) suggests that the LFS will maintain this attribute in the near future ­ at least for the period 2002-05, and possibly a little beyond that, although this will almost certainly not be the case of the LFS in the long run, since it will eventually incorporate some of the paneling features generally implemented by surveys of its kind elsewhere in the European Union. 60 Calculations were done with the Stata statistical software. The employment and unemployment rates were computed as ratio estimators using the "svyratio" command, whereas the average monthly earnings from the primary job used the "svymean" command. The sampling errors of the Gini coefficient were computed with the "svyregress" command, taking advantage of a transformation that reduces it to a regression coefficient. The transformation is described by Giles, E. A. in Calculating a Standard Error for the Gini Coefficient: Some further results. University of Victoria Department of Economics Working Paper EWP0202, Victoria, B.C. (2002). It is implemented as a Stata add-on command (svygini) that extends Giles' idea to complex survey designs. It can be obtained from the Juan Munoz by simple request to juan.munoz@ariel.cl. 178 Table A1: Croatia Labor Force Survey Employment rate. Estimations, Standard Errors and Coefficients of Variation for ˝ year, 1 year and 3 years, by County and Statistical Zones Employment Rate ˝ year (Jul-Dec 2004) 1 year (2004) 3 years (2002-04) Est SE CV Est SE CV Est SE CV Counties 1 Zagreb County 52.99% 1.54% 2.91 53.07% 1.18% 2.21 52.91% 0.70% 1.33 2 Krapina-Zagorje 70.31% 4.27% 6.07 73.61% 3.03% 4.12 69.62% 1.74% 2.49 3 Sisak-Moslavina 51.92% 3.03% 5.84 53.24% 2.50% 4.70 50.70% 1.24% 2.45 4 Karlovac 61.08% 2.82% 4.61 56.68% 2.56% 4.51 57.12% 1.43% 2.50 5 Varazdin 65.67% 2.12% 3.23 64.15% 1.83% 2.86 64.22% 0.88% 1.38 6 Koprivnica-Krizevci 58.52% 3.22% 5.50 61.65% 2.45% 3.97 62.85% 1.51% 2.40 7 Bjelovar-Bilogora 62.73% 3.61% 5.75 61.05% 2.41% 3.95 63.86% 1.63% 2.55 8 Primorje-Gorski kotar 55.53% 1.78% 3.21 57.51% 1.28% 2.23 56.63% 0.61% 1.08 9 Lika-Senj 66.05% 6.72% 10.18 71.07% 5.57% 7.83 69.03% 3.01% 4.36 10 Virovitica-Podravina 56.76% 4.46% 7.86 59.24% 3.76% 6.34 58.09% 1.69% 2.91 11 Pozega-Slavonia 55.59% 4.83% 8.69 58.53% 2.88% 4.93 53.52% 1.68% 3.15 12 Slavonski Brod-Posavina 46.20% 2.65% 5.74 45.66% 2.20% 4.83 46.45% 1.21% 2.61 13 Zadar 48.14% 3.33% 6.92 49.46% 2.94% 5.94 49.38% 1.43% 2.89 14 Osijek-Baranja 52.15% 1.74% 3.34 51.33% 1.59% 3.10 47.43% 0.88% 1.85 15 Sibenik-Knin 57.27% 3.81% 6.65 53.47% 2.95% 5.52 45.91% 1.72% 3.75 16 Vukovar-Sirmium 41.82% 1.90% 4.55 43.32% 1.56% 3.59 42.57% 1.12% 2.64 17 Split-Dalmatia 51.41% 1.73% 3.37 50.05% 1.13% 2.26 48.14% 0.60% 1.24 18 Istria 54.70% 2.03% 3.72 58.69% 1.38% 2.35 58.70% 0.79% 1.35 19 Dubrovnik-Neretva 60.05% 1.94% 3.22 55.21% 1.71% 3.10 53.88% 0.91% 1.69 20 Meimurje 59.87% 1.82% 3.05 59.19% 1.08% 1.83 58.98% 0.84% 1.43 22 Zagreb City 56.36% 1.20% 2.13 56.60% 0.81% 1.43 57.08% 0.45% 0.79 5 regions Central Region 61.21% 1.17% 1.91 61.17% 0.93% 1.52 60.81% 0.51% 0.84 Eastern Region 49.45% 1.14% 2.31 49.93% 0.95% 1.90 47.86% 0.54% 1.12 Zagreb Region 55.40% 0.96% 1.74 55.61% 0.67% 1.20 55.91% 0.38% 0.68 Adriatic North Region 56.12% 1.37% 2.45 59.06% 1.02% 1.73 58.47% 0.51% 0.88 Adriatic South Region 52.80% 1.26% 2.38 51.13% 0.94% 1.83 48.94% 0.49% 1.00 3 regions Central & Eastern Region 55.77% 0.84% 1.51 55.96% 0.67% 1.20 54.87% 0.38% 0.68 Zagreb Region 55.40% 0.96% 1.74 55.61% 0.67% 1.20 55.91% 0.38% 0.68 Adriatic Region 54.16% 0.93% 1.72 54.34% 0.70% 1.29 52.79% 0.36% 0.69 Croatia 55.17% 0.53% 0.96 55.36% 0.40% 0.73 54.47% 0.22% 0.40 179 Table A2:Croatia Labor Force Survey Unemployment rate (ILO definition). Estimations, Standard Errors and Coefficients of Variation for ˝ year, 1 year and 3 years, by County and Statistical Zones. Unemployment Rate ˝ year (Jul-Dec 2004) 1 year (2004) 3 years (2002-04) Est SE CV Est SE CV Est SE CV Counties 1 Zagreb County 12.80% 1.44% 11.26 13.43% 1.06% 7.90 14.78% 0.67% 4.55 2 Krapina-Zagorje 2.70% 1.07% 39.44 2.79% 0.88% 31.70 4.89% 0.72% 14.78 3 Sisak-Moslavina 16.03% 2.26% 14.12 17.12% 1.89% 11.03 18.98% 1.07% 5.66 4 Karlovac 13.92% 2.43% 17.48 16.93% 2.12% 12.54 15.53% 1.19% 7.69 5 Varazdin 7.72% 1.51% 19.59 8.56% 1.11% 12.92 8.20% 0.50% 6.10 6 Koprivnica-Krizevci 13.60% 2.90% 21.33 11.13% 1.65% 14.78 10.15% 0.84% 8.26 7 Bjelovar-Bilogora 12.35% 2.70% 21.86 10.92% 1.97% 18.04 10.91% 1.10% 10.04 8 Primorje-Gorski kotar 11.47% 1.59% 13.90 10.35% 1.20% 11.59 11.21% 0.61% 5.44 9 Lika-Senj 6.06% 1.92% 31.68 6.16% 2.15% 34.95 8.68% 1.43% 16.46 10 Virovitica-Podravina 15.66% 4.27% 27.30 14.89% 2.86% 19.24 13.96% 1.26% 9.05 11 Pozega-Slavonia 13.75% 3.80% 27.60 11.25% 1.66% 14.72 13.90% 1.41% 10.17 12 Slavonski Brod-Posavina 21.84% 2.57% 11.77 20.69% 2.05% 9.89 15.73% 1.18% 7.50 13 Zadar 17.07% 2.83% 16.59 15.39% 2.13% 13.85 18.90% 1.16% 6.13 14 Osijek-Baranja 20.80% 1.96% 9.41 20.61% 1.54% 7.45 22.89% 0.89% 3.87 15 Sibenik-Knin 18.01% 4.01% 22.29 22.26% 3.19% 14.33 28.16% 2.21% 7.83 16 Vukovar-Sirmium 26.28% 2.81% 10.69 24.76% 1.99% 8.02 24.07% 1.24% 5.13 17 Split-Dalmatia 19.17% 1.64% 8.56 19.05% 1.15% 6.03 19.64% 0.62% 3.14 18 Istria 8.58% 1.85% 21.59 8.28% 1.12% 13.54 8.38% 0.66% 7.92 19 Dubrovnik-Neretva 13.00% 2.50% 19.20 17.57% 2.56% 14.54 17.42% 1.19% 6.82 20 Meimurje 11.80% 2.09% 17.73 12.22% 1.51% 12.39 12.27% 1.00% 8.14 22 Zagreb City 9.70% 0.98% 10.07 10.21% 0.71% 6.92 10.71% 0.39% 3.68 5 regions Central Region 11.03% 0.83% 7.52 11.28% 0.62% 5.50 11.47% 0.35% 3.05 Eastern Region 20.89% 1.26% 6.01 19.91% 0.90% 4.52 19.93% 0.52% 2.63 Zagreb Region 10.56% 0.81% 7.67 11.09% 0.59% 5.31 11.84% 0.34% 2.89 Adriatic North Region 9.91% 1.13% 11.37 9.19% 0.81% 8.76 9.92% 0.44% 4.42 Adriatic South Region 17.67% 1.20% 6.78 18.62% 0.92% 4.96 20.23% 0.51% 2.54 3 regions Central & Eastern Region 15.36% 0.74% 4.83 15.06% 0.54% 3.55 15.06% 0.30% 2.02 Zagreb Region 10.56% 0.81% 7.67 11.09% 0.59% 5.31 11.84% 0.34% 2.89 Adriatic Region 14.55% 0.85% 5.83 14.73% 0.64% 4.33 15.91% 0.35% 2.21 Croatia 13.95% 0.47% 3.34 13.98% 0.34% 2.43 14.51% 0.19% 1.32 180 Table A3: Croatia Labor Force Survey Average earnings from the Primary Job Estimations, Standard Errors and Coefficients of Variation for ˝ year, 1 year and 3 years, by County and Statistical Zones Monthly earning ˝ year (Jul-Dec 2004) 1 year (2004) 3 years (2002-04) Est SE CV Est SE CV Est SE CV Counties 1 Zagreb County 3387 69 2.05 3436 55 1.59 3319 37 1.12 2 Krapina-Zagorje 3230 160 4.96 3191 111 3.47 3012 72 2.39 3 Sisak-Moslavina 3094 131 4.23 3199 106 3.32 3084 64 2.08 4 Karlovac 3195 189 5.92 3110 150 4.82 2967 78 2.62 5 Varazdin 2505 136 5.44 2602 125 4.81 2573 88 3.42 6 Koprivnica-Krizevci 2656 253 9.53 2720 214 7.88 2604 115 4.41 7 Bjelovar-Bilogora 2494 123 4.92 2491 127 5.11 2603 86 3.30 8 Primorje-Gorski kotar 4115 145 3.53 3965 91 2.30 3697 49 1.33 9 Lika-Senj 3102 196 6.31 3245 176 5.42 3033 186 6.12 10 Virovitica-Podravina 2745 266 9.68 2620 223 8.53 2341 108 4.61 11 Pozega-Slavonia 2563 251 9.77 2723 226 8.31 2670 119 4.46 12 Slavonski Brod-Posavina 3040 131 4.32 2872 115 4.01 2662 76 2.87 13 Zadar 3335 190 5.69 3279 155 4.71 3426 86 2.52 14 Osijek-Baranja 3020 97 3.22 3094 80 2.60 2958 52 1.74 15 Sibenik-Knin 3501 190 5.41 3629 203 5.60 3262 95 2.90 16 Vukovar-Sirmium 3368 239 7.09 3318 152 4.59 3106 87 2.79 17 Split-Dalmatia 3863 130 3.36 3781 96 2.53 3617 49 1.35 18 Istria 3633 123 3.39 3529 95 2.69 3330 51 1.53 19 Dubrovnik-Neretva 3818 244 6.39 3678 180 4.89 3519 88 2.51 20 Meimurje 3591 270 7.52 3481 161 4.62 2999 97 3.25 22 Zagreb City 4072 90 2.22 4058 73 1.80 3895 36 0.91 5 regions Central Region 2914 71 2.43 2936 57 1.95 2806 35 1.23 Eastern Region 3007 78 2.59 3003 60 2.01 2826 35 1.25 Zagreb Region 3880 68 1.76 3885 55 1.41 3735 27 0.73 Adriatic North Region 3840 95 2.47 3737 64 1.72 3498 38 1.08 Adriatic South Region 3719 90 2.42 3654 70 1.92 3524 36 1.01 3 regions Central & Eastern Region 2955 53 1.78 2966 42 1.40 2815 25 0.88 Zagreb Region 3880 68 1.76 3885 55 1.41 3735 27 0.73 Adriatic Region 3770 66 1.74 3691 48 1.31 3513 26 0.74 Croatia 3440 35 1.03 3427 27 0.80 3276 15 0.47 181 APPENDIX 3 METHODOLOGY OF DECOMPOSITION OF THE ROLE OF INDIVIDUAL AND REGIONAL CHARACTERISTICS IN WAGE DIFFERENTIALS Following the Oaxaca decomposition methodologies (Oaxaca, 1973), we consider Region 1 (Central Region) as the region of reference (region r ), and compare the earning of each region i ( i =2, 3, 4, 5) with the earning of Region 3. The difference in wage structure, which is the so-called regional discrimination coefficient ( D ) in the jargon of economics, can be expressed as: Wi / Wr - (Wi / Wr ) 0 D= (1) (Wi / Wr ) 0 where (Wi / Wr ) stands for the observed earning ratio between region i and region r ; and (Wi / Wr ) 0 the earning ratio between region i and region r in the absence of regional discrimination. An equivalent expression in natural logarithms is: ln( D + 1) = ln(Wi / Wr ) - ln(Wi / Wr ) 0 (2) If regional discrimination is absent, one of the following two assumptions may hold: (i) the earning structure currently faced by workers in region r would also apply to workers in each region i ; or (ii) the earning structure currently faced by workers in each region i would also apply to workers in region r . If the first (second) assumption holds, the workers in region r ( i ) would on average receive, in the absence of discrimination, the same earnings as they currently receive, but that discrimination takes the form of the workers in region i ( r ) receiving less (more) than a non- discriminating labor market would award them. We estimate the earning equation in semi-logarithmic forms separately for each region:61 ln(W p ) = Z p + p ' p =1,2,...n where W p = the monthly earning rate of the p -th worker; ' Z p = a vector of individual characteristics, including age, years of school, gender; = a vector of coefficients; p = a disturbance term. When the regional earning differential is expressed in natural logarithms, the formulation of the discrimination coefficient in (2) and our alternative assumptions about which earning structure would prevail in the absence of discrimination together imply that the earning differential can be decomposed into the effects of regional discrimination and the effects of difference in individual characteristics. Wi - Wr Let G = (4) Wr Then ln(G + 1) = ln(Wi ) - ln(Wr ) Where the Wi and Wr are the average monthly earning for workers in region i and region r , respectively. From the properties of ordinary least squares estimation, we have 61 See Oaxaca (1973). 182 ' ln(Wi ) = Z i i^ (5) ' ^ ln(Wr ) = Z r r (6) ' ' where Z i and Z r stand for vectors of mean values of the regressors for workers in region i and ^ ^ region r , respectively; i and r the corresponding vectors of estimated coefficients. Upon substitution of (5) and (6) into (4), we have ' ^ ' ^ ln(G + 1) = Z i i - Z r r (7) If we let ' ' Z ' = Zi - Z r (8) ^ ^ = - ^ (9) r i ^ ^ ^ and substitute i = r - in (7), then the regional earning differential can be written as: ^ ln(G + 1) = Z ' r - Z i ' ^ (10) On the basis of equation (2) and the assumption that the current earning structure in region r would apply to both region i and region r in a non-discriminating labor market, it can be that W^ ^ ln( i ) 0 = Z ' r (11) Wr ^ ln( D + 1) = - Z i ^' (12) The expressions (11) and (12) represent the decomposition of the wage differential into the estimated effects of differences in individual characteristics and the estimated effects of discrimination, respectively. 183 BACKGROUND PAPER #6 ASSESSING THE FLEXIBILITY OF THE CROATIAN LABOR MARKET INTRODUCTION High unemployment and low employment with stagnant pool of the unemployed persons have been repeatedly portrayed as one of the mayor weaknesses of the Croatian economy. Diagnosing causes of unemployment and long-term unemployment in any country, even with well-established data sources and stable economic relations, is a demanding task and reported results are subject to a lively international debate. Research of unemployment causes in a transition economy with unstable economic relations and poor data sources is even more challenging task. This report aims to shed more light on policies and labor market features in Croatia that are often found to influence unemployment. Particular attention is paid to comprehensive analysis of the labor market flexibility as it has been often reported to lie at the root of unemployment in Croatia. First part of the report provides overview of the main labor market features. Although unemployment and long-term unemployment do not depart from values recorded in other transition countries, those appear to be in the higher area of the range observed in those countries. Following chapter provides an overview of major labor market policies in Croatia in international perspective including unemployment benefit system, employment protection legislation, active labor market policies and labor taxation. Finally, having those policies, especially employment protection legislation, in mind, a more in depth analysis of the job flows is provided. Unique series of job flows in Croatia is constructed and its properties are described. Also, a series of decompositions of the job flow data are performed in order to distinguish more and less flexible parts of the economy. OVERVIEW OF THE MAJOR TRENDS IN THE LABOR MARKET Demographic Trends The transition process in Croatia started against a background of unfavorable demographic trends. The Croatian society was ageing even before the transition started as a consequence of low fertility rate, which approached the level observed in Western countries. This process continued even further during the last decade. The share of persons younger that 15 years in total population decreased from 19.7 percent to 17.1 percent between the 1991 and 2001, while the share of person older than 65 years increased from 11.8 percent to 15.7 percent. As the average age of the population increased by about 2 years, the share of persons in working age decreased by 1.3 percent and reached 67.2 percent. Therefore, the process of ageing society in Croatia is already shrinking the working age population at the expense of older persons. Unfavorable demographic developments were coupled during the early transition with the extensive use of early retirements as a means of solving the unemployment prospects for many older workers. As a consequence of such measures, the share of retired persons below the age of 65 culminated at 39 percent off all retired persons in the year 2000, which exceeded the number of active persons in the 50-65 age group. Also, the ratio of employed to retired persons dropped to about 1.4, which is less than any of the Western countries is expected to reach within the next 20 years. The awareness of the 185 non-sustainability of such policies led to change of the pension rules in 2001 and considerably slowed the inflow of new persons into the retirement. Adverse demographic trends in Croatia are expected to continue in the future. By the middle of the 21st century, according to the UN demographic projections, Croatian population may reduce by a quarter. At the same time the average age of the population is expected to increase by ten years, and it is likely that the share of persons older than 65 years will grow further and reach a third of total population. These trends will undeniably put a large additional burden on the Croatian pension system and therefore it is necessary to emphasize policies needed to increase the participation rate. Activity and Employment Trends Employment dynamics in Croatia during the transition exhibits a pronounced fall followed by a partial recovery, which was typical for all of the CEE transition countries. However, the relative magnitude of the Croatian fall in employment depends on the way it is measured. If the fall is measured using the administrative data on employment, it appears to be substantial. According to that source, employment decreased by about a quarter between 1990 and 2001, the year employment reached its lowest level. On the other hand, census data, which are more similar to the ILO methodology, reveal a modest fall of the employment rate from 57.7 percent to 52.3 percent, corresponding to the employment fall of roughly half the size of the fall shown by the administrative data over the same period. Even controlling for the fact that employment fall started in the late 1980's and comparing the census 2001 data with census 1981 data, the employment fall does not appear to be much greater. The large difference between those two sources show that administrative statistical sources, geared towards a few large systems during the communism, often failed to adequately encompass the growth of self- employment and small enterprises. Figure 6.1: Dynamics of GDP, Average Net wage and Employment (1990=100) 120 GDP 110 Real net wages Civilian employment 100 90 80 70 60 50 40 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Source: CBS. Close to half of the fall in employment rate between 1991 and 2001, according to the population census, pertained to the decrease in activity, i.e. persons leaving labor force due to early retirement or because they became discouraged and stopped searching for job. The rest of the employment fall was reflected in the increase of the unemployment rate, which rose from 9.7 percent to 14.9 percent, according to the census data. Simultaneous increase of unemployment and decrease of activity occurred in all the CEE transition countries and therefore the general behavior of the Croatian labor market in transition does not diverge from their performance. However, there were several distinctive features of the labor market transition in Croatia. First, due to more decentralized nature of decision- 186 making in enterprises and certain market elements of the system, there was significant unemployment even before the start of the transition. The unemployment rate according to 1981 census was 4.9 percent and it rose to 9.7 percent until 1991. Due to that fact, even a modest decrease in employment, measured by the standard of CEE countries, brought Croatian unemployment rate in the upper part of the range observed in those countries as it reached 14.9 percent in 2001, according to the census. Further on, the employment in Croatia, according to both administrative and survey sources, reached its lowest level during the early 2001 and started to recover thereafter. This means that there was a rather prolonged period of fall in employment, despite the moderate cumulative fall. The slow and prolonged employment decrease was the consequence of delayed restructuring process and many enterprises operating under the soft budget constraint until the late 1990's. Significantly lower activity rate than the rest of the CEE transition countries, even before the transition, is the final important distinguishing feature of the Croatian labor market. Although employment rate did not decrease as much as it did in some other CEE transition countries, the employment rate in 2001 was considerably lower due to lower employment rate at the beginning of the transition. Table 6.1: Decomposition of the Change in Employment (In Percentage Points) Change resulting from: Country Period Change in Unemployment Activity Rate Demography employment rate Croatia 1981-1991 -0.8 3.1 -3.0 0.7 Croatia 1991-2001 -6.9 2.5 3.5 0.9 Bulgaria 1989-1996 -22.2 -9.9 -10.5 -1.8 Czech R. 1989-1996 -9.6 -2.6 -8.0 1.0 Hungary 1989-1996 -22.9 -6.9 -16.7 0.7 Poland 1989-1996 -13.1 -9.4 -4.6 1.0 Romania 1989-1996 -5.6 -6.3 -0.9 1.5 Slovak R. 1989-1996 -11.6 -8.4 -4.7 1.6 Sources: for Croatia 1981, 1991 and 2001 Population censuses; for other countries Ambrus-Lakatos and Schaffer eds. (1998) Industrial and Occupational Change The general process of structural change in successful transition countries involved decrease of the employment share of agriculture and industry while at the same time the share of services increased. Croatia had an initial advantage in this process since it had the highest share of employees in services of all transition countries at the beginning of the transition process (Vujci, 1998). Moreover, most of the change in employment structure in Croatia during the transition was in the "right" direction, if the employment structure of the EU countries is considered as a target of the restructuring process. However, the restructuring process in Croatia took place through job destruction, while there were few expanding activities until very recently. Due to the speed and efficiency of transformation, the shares of persons employed in agriculture and industry in Croatia have now reached levels comparable with EU countries. The share of persons employed in manufacturing in the first half of 2004 was 21.7 percent according to the labor force survey, which is close to the EU average, while the share of persons employed in agriculture was 16.5 percent, which is still within the shares observe in EU member countries. However, the survey agricultural employment includes many unpaid family workers or retired persons complementing their income in that way since administrative data, which include employees in legal entities and individual farmers with pension insurance, show that agriculture makes only 6.6 percent of total employment. That figure is much closer to the EU average than the survey data. The high share of employment in services in Croatia is a result of many workers employed in wholesale and retail trade; repairs; transport, hotels and restaurants in Croatia. The employment shares of those sectors, according to the official data, are at the higher boundary of the EU countries, which reflects the fact that Croatia has fairly developed tourist sector due to natural advantages. 187 Table 6.2: Structure of Employment in Central and Eastern Europe and EMU (in %) Hungary Poland Czech EMU12 EMU EMU Croatia Croatia (2000) (2000) Rep. (weighted average) (min) (max) LFS administrative (2000) (1999) (1999) (1999) (2004) (2004) Agriculture, hunting 5.9 18.8 5.1 5.3 1.9 18.1 16.5 6.6 forestry; fishing Manufacturing 28.1 23.8 30.2 20.4 13.9 23.7 21.7 24.0 (incl.energy) Construction 6.4 7.0 9.3 7.3 6.0 10.5 8.2 8.5 Wholesale and retail trade; 24.5 23.4 25.6 25.0 22.3 28.2 26.0 30.7 repairs; transport, hotels, restaurants Fin. Intermediat.; real est., 7.2 7.7 10.6 13.0 7.0 23.5 27.8 8.6 renting, business activities Other service activities 27.9 19.3 19.2 29.1 22 36.3 16.5 25.6 Sources: CBS and Csajbók and Csermely (2002) At the level of particular manufacturing sectors, successful restructuring process during the transition involved decrease of employment in labor intensive activities, such as food products and beverages, textiles and leather. At the same time employment tended to move towards more capital and knowledge intensive activities, such as electric instruments, radio and communication equipment. Although structure of manufacturing employment in Croatia mostly does not diverge from the patterns observed in EU countries, the employment share of labor intensive industries is still fairly high. Therefore, this process is yet expected to take place in Croatia. Recent advancement in privatization and increased number of jobs generated by small private companies brought further decrease of the share of public sector employment, which was reduced from more than 40 percent in the first half of 2001 to less than 31 percent in the first half 2005. Table 6.3: Shares of Gross Manufacturing Output by Industry Hungary EMU10 Croatia (excl. Belgium , Luxembourg) 2001 2001 1997 change weighted average min max 2004 2000 change Food products and beverages 15.5 21.7 -6.2 15 9.3 28.7 25.0 27.1 -2.1 Tobacco products 0.8 0.7 0.1 1.5 0.1 2.6 2.8 2.6 0.2 Textiles 1.3 2.2 -0.9 3 0.8 7.6 2.0 1.5 0.6 Wearing apparel and fur 1.9 1.5 0.4 2 0.3 6.9 3.3 4.3 -0.9 Leather and leather products; footwear 0.6 0.8 -0.2 1.3 0.2 4.5 1.5 1.4 0.1 Wood and wood products; furniture 1.1 1.6 -0.5 1.5 0.9 5.1 3.3 2.3 1.0 Paper and paper products 1.8 1.9 -0.1 3 1.4 14.2 4.0 4.1 -0.1 Publishing, printing and recorded media 2.6 3.3 -0.7 3.7 2.6 7.7 8.8 6.8 2.0 Coke and refined petroleum products; 5 7.1 -2.1 2.1 1.4 2.9 3.5 4.7 -1.1 nuclear fuel Chemicals and chemical products 6.8 9.2 -2.4 11.8 5.1 17.8 8.9 11.2 -2.3 Rubber and plastic products 3.7 3.6 0.1 1.4 0.7 2.2 3.0 2.3 0.7 Other non-metallic mineral products 2.8 3.3 -0.5 3.7 1.9 6.5 6.7 6.0 0.7 Basic metals 3.9 5.5 -1.6 5.1 0.9 7.8 1.5 1.2 0.2 Fabricated metal prod., excl. machinery 4.1 4.5 -0.4 6.1 2.1 7 5.9 4.9 1.0 Machinery and equipment 4.3 5.5 -1.2 9.6 3.1 12.4 4.0 3.4 0.6 Office machinery and computers 5.2 6 -0.8 1.9 0 17.8 0.5 0.2 0.4 Electr. machinery and apparatus, n.e.c. 11.2 4.3 6.9 4.6 2.2 5.4 3.5 3.3 0.2 Radio, TV, communicat.equipment 10 5 5 3.6 1.1 8.1 2.7 4.1 -1.3 Precision instruments 1 1.4 -0.4 1.9 0.3 3 1.0 1.0 0.0 Road vehicles 14.5 9 5.5 11.9 0.6 14.5 0.5 0.3 0.1 Other transport equipment 0.5 0.4 0.1 2.4 0.7 3.7 4.2 4.7 -0.5 Furniture manufacturing not elsewhere 1.2 1.2 0 2.7 1.4 4.6 2.6 2.5 0.1 classified Recycling 0.1 0.1 0 0.1 0.1 0.1 0.8 0.3 0.5 Sources: CBS and Csajbok and Csemely (2002) 188 RECENT LABOR MARKET INDICATORS IN COMPARATIVE PERSPECTIVE According to the criterion of registered unemployment, which stood at about 19 percent at the end of 2004, Croatia had one of the more serious unemployment situations among the Central and Eastern European transition countries, most of which are 'new' EU members. However, according to the survey unemployment rate, which is a much better basis for comparison because it is calculated according to the uniform methodology proposed by the International Labor Organization, which is not affected by the peculiarities of the national systems62, Croatia did not stand out from the group. According to survey unemployment, Poland and Slovakia had the highest unemployment rates, which amounted to 17-19 percent in 2004. The fact that Poland is the most successful transition country, according to recovery of GDP in comparison to the pre-transition period, suggests that favorable macroeconomic indicators alone are not sufficient to solve unemployment problems. The survey unemployment rate for Croatia in that year was about 14 percent, which is somewhat lower than in the countries mentioned and also represents a slight decline compared to previous years, but it is still about 3 percentage points above the average for the ten selected countries. Bulgaria was in the same group with Croatia, while Lithuania's, Estonia's and Latvia's unemployment rates were slightly lower. Final four of the selected transition countries had single digit survey unemployment rates. Romania owes its low unemployment rate to the relative slowness of the reforms and the importance of agriculture for its economy, which is related to relatively high employment rates. The Czech Republic has since the beginning of transition had the best unemployment indicators, but the advancement of the reforms led to unemployment growth in that country. Hungary and Slovenia represented the two countries that had the least problem with unemployment in 2004, and they were even slightly better than the EU average. However, Hungary is also an example of a country where a large decline in employment during the 1990s did not even remotely affect unemployment, pointing to some deficiencies of focusing on unemployment rates63. Figure 6.2: Administrative (registered) and Survey Unemployment Rates, 2004 20 Administrative, end of period ILO survey, period average % 15 10 5 0 Slovak R. Hungary Latvia Slovenia Bulgaria Croatia Estonia Poland Romania Lithuania Czech R. Sources: CBS (2005); Eurostat (2005); WIIW (2005). 62 The peculiarities of the national systems primarily pertain to incentives to register such as privileges arising from registering and conditions these persons have to meet in order to be able to register. 63 For example, the employment rate declined by almost 23 percentage points in Hungary between 1989 and 1996 (from almost 80 percent to about 55 percent, or to a level comparable to Croatia; according to Boeri, Burda and Kölló, 1998), which is the largest reduction among the transition countries of Central and Eastern Europe. However, about three-quarters of the employment decline pertained to exits from the labor force (often supported by early retirements, while discouraged worker effect was important as well) so the unemployment rate remained single-digit despite recorded employment fall. 189 Along with an unemployment rate that hovered above the average of the selected transition countries, Croatia also had higher long-term unemployment. High levels of total and long-term unemployment in Croatia fit well within the general features of the unemployment in Central and Eastern European countries, where high unemployment has been tied to high long-term unemployment and exclusion from the labor market. Fairly high long-term unemployment in Croatia has been a consequence of weaker labor market dynamics, or fewer newly-employed workers, but also of fewer separations. The share of employees with tenures of less than one year in total employment (accession rate) in Croatia is amongst the lowest for the selected countries, regardless of the source used. Moreover, average tenure64 in Croatia amounts to 12 years, which puts Croatia, together with Slovenia, at the top of the range observed in selected countries and confirms that the Croatian labor market is one of the least dynamic. Figure 6.3: Overall and Long-term Unemployment, 2004 70 Share of long-term unemploymentl (%) Slovakia 65 60 Bulgaria Romania 55 Poland Slovenia Estonia Lithuania 50 Czech R. Croatia 45 Hungary Latvia 40 5 7 9 11 13 15 17 19 21 Unemployment rate (%) Sources: Eurostat (2004); CBS (2004). Survey unemployment rate has since the introduction of the Labor force survey in 1996 constantly been significantly lower than the administrative unemployment rate. Moreover, although both of them have been rising, when the survey unemployment rate reached its peak value of 17.0 percent in the second half of 2000, they have been diverging steadily because administrative unemployment rate was growing faster than the survey unemployment rate. The divergence process lasted until the second half of 2002, when the gap started to close due to faster reduction of the administrative unemployment rate. Therefore, it may seem that the inferences drawn depend upon the information source chosen. However, some conclusions can be drawn with confidence regardless of the source used. First of all, although long-term unemployment has been on a decrease recently, about half of all unemployed, according to the survey conducted in 2004, have been looking for job over a year. This is not unusual since most transition countries experience problems with high unemployment rates and insufficient job creation, which translates into the long-term unemployment (to an even greater extent than it does in Croatia). Secondly, unemployment is mostly confined to persons with little education and especially persons with vocational education, which experienced a strong reduction in jobs available during the transition. These are the groups that remain stuck with the unemployment office for the longest periods and often drop out of the labor force. Finally, young persons have been the most seriously affected by unemployment in Croatia. They are the group with immensely high unemployment, as their survey unemployment rates during the last couple of years hovered above 30 percent, despite some recent reductions. Even if the national average for unemployment rates is taken into the account, Croatia still faces the highest relative unemployment rate of young persons amongst all of the CEE transition countries. Moreover, even such a high unemployment rate of young persons does not reveal the full picture because of their low participation rate. While the employment rate for prime-age men 64 Job tenure pertains to the length of continuous employment with a single employer. 190 in Croatia has been only slightly lower than the EU average and the employment rate for prime-age women is close to the EU average, there is major gap in the young age. That gap accounts for most of the employment gap between the Croatia and the EU. Therefore, it can be concluded that entry into the employment remains a significant problem, even though most of the new employment takes the form of the flexible fixed-term contracts. Persons with permanent jobs in Croatia do not fare bad in comparison to their colleagues from other CEE transition countries. First of all, they face a low risk of loosing a job. Although Croatian unemployment rate is in the upper part of the range observed in CEE transition countries, the inflow of persons with permanent contracts into the unemployment is not so large. Further on, average real wage in Croatia is the second highest amongst the transition countries of Central and Eastern Europe (next to Slovenia). It remains high even after it is accounted for the high level of prices in Croatia. Finally, there is a rather low degree of wage inequality, reflecting a rigid wage structure. On the basis of presented roughly sketched lifecycle of a typical worker, it can be concluded that there is still a low degree of dynamism and too few new jobs available in the labor market to mop up more of the unemployed persons. The overall level of unemployment does not have to be directly related to the dynamics of job creation and job destruction, but lower job creation directly increases the duration of unemployment or the level of long-term unemployment. Moreover, as worker flows are mostly confined to the group of younger, more dynamic employees, it adds even more to the problem of high long-term unemployment in Croatia. However, there are also some recent improvements that need to be mentioned on the positive side. It has to be noted that government introduced a responsible wage policy before a couple of years, which managed to stop and even reverse the trend of rising unit labor costs. This policy was additionally supported with the repeated reductions of non-wage labor costs so Croatian employers now face the smallest burden of taxes and contributions amongst the CEE transition countries. With reaching advanced stage of the enterprise restructuring process recently, generation of new jobs in the sector of small enterprises and trades and crafts managed to reverse a decade long trend of decreasing employment. All these development were reflected in the steadily rising trend of vacancies. Finally, a new Labor code has been introduced in the summer 2003. The new Labor code is more aligned with the EU guidelines and practice of EU member countries. It also provides more maneuver space for the enterprises in enhancing their flexibility and. Finally, it reduces the overarching impact of legislation and creates an area for implementation of collective bargaining, additionally reinforced by the creation of institutional support for social partnership. Therefore, it can be expected that a new set of policies will bring some more dynamism into the labor market. LABOR MARKET INSTITUTIONS AND UNEMPLOYMENT Common discussion in theoretical and empirical literature on labor market performance is on the impact of various policies, especially so-called labor market institutions, upon joblessness and long- term unemployment. The different positions in the discussion can be placed into two main groups. One of the views is skeptical about the possible influences of labor market institutions and rather focuses on the insufficiency of demand for labor arising from various causes (i.e. inadequate macroeconomic policies). Some recent studies, such as Ederveen and Thissen (2004) and Arandarenko (2004), although they differ in respect to some of the research issues and observed countries, clearly belong to this group as they are particularly skeptical on the possible influence of labor market institutions. Ederveen and Thissen (2004) argue that transition countries in general have more flexible labor market arrangements than Western European countries, which do not translate into superior labor market outcomes. Therefore, labor market institutions, according to their view, are not an important cause of poor labor market performance in those countries. Similarly to cited study, Arandarenko (2004) shows that labor markets of some South-East European countries with more flexible institutions did not in general perform better, reiterating Freeman's (2000) point that each county should seek whichever institutions suit them the best. 191 An alternative view holds the position that some state interventions, such as demand management policies, might affect labor market performance only in the short-run, while in the longer run unemployment converges towards its equilibrium level, which is affected by other variables belonging to a wide range of so-called `labor market institutions. Recent contributions to the debate holding this position are discussions by Nickell, Nunziata and Ochel (2002) and European Commission (2004). After description of the labor market institutions that might affect labor market outcomes, this report will consider in more detail labor market policies employed in Croatia. There is no single agreed view on the relevant set of variables and policies with the potential to affect labor market outcomes. In principle, it should include all variables that may affect firms' hiring decisions (or labor demand), the size and the structure of labor supply and interaction between those two (bargaining framework). The European Commission (2004) distinguishes three different groups of variables that could influence labor market performance. First is the group comprising of policies (or variables) that are under direct control of policy makers. The second group contains variables describing socio-economic structure of the labor market, or centralization and coordination of the collective bargaining. Finally, there are also some external factors, such as openness and productivity growth, which affect employment. This report will focus on the first group of factors that fall within the domain of policymakers. Moreover, this report will not question relative roles of each particular policy, as findings with respect to those often differ (the Commissions report in particular summarizes different findings), but it will rather focus on the benchmarking of those policies in Croatia against the `old' and selected `new' EU member states. In the recent 'Employment in Europe' report, European Commission (2004) considered the roles of: labor taxes and contributions (that are found to weakly negatively influence employment), generosity of the unemployment benefits (found to reduce employment) and active labor market policies (compensating for the negative impact of unemployment benefits) as the relevant set of policies. The range of policies discussed by Nickell et. al. (2005) is somewhat broader and it includes the unemployment benefit system, employment protection laws, active labor market policies and labor taxes. Also, Nickell, et. al. (2005) focus on unemployment rather than on employment. They also include some additional external factors so their full list consists of the real interest rate, union structures and the extent of co-ordination in wage bargaining, terms of trade changes and shifts in trend productivity growth. According to their findings, changes in labor market institutions explain around 55 per cent of the rise in European unemployment from the 1960s to the first half of the 1990s. The most prominent among those were changes in the benefit system and labor taxes, while shifts in the union variables and movements in employment protection law were a bit less important. This report will look at the later of the two presented and more extensive set of policies so it will seek to describe the differences in the unemployment benefit systems, employment protection laws, active labor market policies and labor taxes in the following chapters. EMPLOYMENT PROTECTION LEGISLATION Employment protection legislation (EPL) in Croatia was until recently ranked among the strictest in Europe, according to the composite index of the EPL strictness developed by OECD. Reform of the EPL, taking place between 2001 and mid-2003, resulted in significant reductions of those restrictions. The reform enjoyed strong support from various international institutions so it was embodied as a structural benchmark in both the agreement with the World Banks on Structural Adjustment Loan and the Stand-by arrangement with the International Monetary Fund. However, both the awareness of the necessity of the reform as well as the final result of the negotiations were outcomes of the dialogue within the domestic tripartite institutions. The case for the changes of the Labor code was reinforcement by the impression that it was considerably more restrictive than similar legislation in developed countries as well as in the neighboring transition countries thus making the creation of new jobs more expensive (and the problem and cost of firing employees) to the entrepreneurs, driving them into the sphere of the informal economy and discouraging foreign investors. This process was also in accordance with existing tendencies in many other European economies (Garibaldi and Mauro, 2002) 192 so Croatia followed tendencies present in advanced European economies, albeit with a delay in the implementation positioning Croatia as a laggard. Figure 6.4: EPL Indices - International Comparison 6 5 4 3 2 1 0 Sweden Belgium New Zealand Switzerland Hungary USA Canada Slovakia Slovenia Finland Germany Norway Ireland Spain Netherlands HR proposal Portugal UK Australia Austria France Czech R. Denmark Poland Japan Estonia Croatia 03 Italy Croatia 01 "Old" EU members average OECD average "New" EU members average Sources: OECD (1999); Riboud, Sánchez-Páramo and Silva-Jauregui (2002); Micevska (2003); Matkovi and Biondi (2003). The implemented reform positioned Croatia close to the average strictness of labor market regulation in the old EU member states, according to the level of the EPL index, which is similar to strategies employed by most of the Central and Eastern European countries (Riboud, Sánchez-Páramo and Silva- Jauregui, 2002). However, their strategy with respect to relaxation of legislation governing the labor market was somewhat specific. While the level of EPL subcomponent describing regulation of permanent contracts was in selected Central and Eastern European countries mostly above the average in `old' EU member states (2.5 on average in selected Central and Eastern European countries against 2.4 in `old' EU member states), regulation of temporary employment was considerably less restrictive (1.8 on average in selected Central and Eastern European countries against 2.2 in old EU members). This observation was particularly relevant in cases of Czech R., Hungary, Poland and Slovakia, as well as Estonia, which also introduced a similar configuration of labor market regulation. Figure 6.5: EPL Configuration - Subindices for Temporary and Permanent Employment 5 EPL sub index (temporary employment) Greece France Italy 4 Bulgaria Finland Spain 3 Romania Croatia Slovenia Portugal Belgium 2 Germany; Austria Sw eden Slovakia Estonia Denmark Poland 1 Netherlands Hungary UK Ireland Czech R. 0 0 1 2 3 4 5 EPL sub index (permanent employment) Sources: OECD (1999); Riboud, Sánchez-Páramo and Silva-Jauregui (2002); Micevska (2003); Matkovi and Biondi (2003). 193 In recent times most 'old' EU member states consistently used deregulation of temporary employment in order increase labor market flexibility at the margin. In accordance with this strategy, growth of temporary employment was not only a generator of the overall employment growth65, but it also absorbed some of the employees previously engaged on permanent contracts (Garibaldi and Mauro, 2002). Employers in countries with stringent regulation of permanent contracts were particularly keen on using temporary contracts so it is not surprising to find larger shares of temporary employment in those countries. Central and Eastern European countries were especially successful in employing such a strategy, resulting in even more flexible regulation of temporary employment. However, results of deregulation of temporary contracts in those countries are not yet wholly visible since shares of temporary employment in those countries are mostly below shares in those 'old' EU members with similar configuration of labor market regulation. Moreover, there does not appear to be any relationship between the regulation of permanent contracts and the extent of temporary employment. Therefore, it seems that temporary employment did not nearly develop in transition countries as much as it did in the 'old' EU members, despite somewhat more restrictive regulation of temporary contracts and more flexible regulation of temporary employment in transition countries. This might also imply that labor markets of transition countries are actually somewhat less flexible, contrary to the aggregate level of EPL index suggests. However, it is also possible that EPL indices do not capture such elements as enforcement, which might be weaker in transition countries (Boeri and Jimeno, 2003). Croatia stands out among other Central and Eastern European countries with fairly high share of temporary employment, only slightly lower than share of temporary employment in Slovenia and Poland. Such a position is probably to some extent a heritage of restrictive regulation governing permanent contracts during the 1990's, that was significantly reduced by the reform. Hence, since reform of regulation governing permanent contracts Croatian employers face restrictions similar to the average of other transition countries, while the use of temporary contracts has evolved further. That could indicate that the actual flexibility of the labor market, which is under significant influence of regulation, after reform became somewhat more favorable, especially in comparison to other Central and Eastern European countries66. Figure 6.6: EPL Sub-index for Permanent Contracts and Temporary Employment (as a share of total, 2003) Share of temporary employment (%) 35 Spain 30 25 Portugal 20 Finland France Netherlands 15 Greece Poland Sweden Slovenia Denmark Croatia 10 Germany Belgium Hungary Italy UK Bulgaria Austria Czech R. 5 Ireland Slovakia Estonia Romania 0 0 1 2 3 4 5 EPL sub index (permanent contracts) Sources: OECD (1999); Riboud, Sánchez-Páramo and Silva-Jauregui (2002); Micevska (2003); Matkovi and Biondi (2003); Eurostat (2004) and CBS (2004). 65 The exemptions to this general tendency were countries like the Ireland with flexible regulation governing permanent contracts. 66 Such a general conclusion does not have to reflect actual position of all different groups of workers so detailed analysis would indicate that older workers (that are also the most protected) are not very flexible, while younger worker happen to be greatly flexible. 194 PASSIVE LABOR MARKET POLICIES Passive labor market policies include many different measures that aim to increase incomes of unemployed persons. Unemployment benefits are regularly prominent amongst those so it is the reason why the level of wages replaced by the unemployment benefits and the duration of the entitlements are usually considered as main determinants of the unemployment benefit system. However, the strictness with which the benefit system is operated and the coverage of the unemployed persons can be even more important than the level of wages replaced by the benefits and the duration of the entitlements. Table 6.1 compares spending on passive policies amongst the 'old' EU members, selected 'new' EU members and Croatia. Spending on passive labor market policies in all Central and Eastern European `new' member states, for which such data is available, was far below such spending in all `old' EU members during the second half of the 1990's. In terms of GDP it was about 60 percent lower and after normalization by unemployment rates (that are somewhat higher in the `new' member states), the difference slightly increases - spending in new member states was only about a third of average spending in the old member states (Greece is the only country among the old members with comparable level of spending). The level of spending in Croatia, even normalized with the unemployment rate, was somewhat below those countries' average67, although the difference was not significant so it fits fairly well with data on passive policies in selected `new' member states. As can be seen from the following table, unemployment benefit system in Croatia, with respect to the average benefit level68 (it was about a quarter of the average wage, or roughly the same as the average of selected 'new' member countries) and benefit duration (it was one year, also close to the average of selected countries) does not differ from unemployment insurance systems that are in place in most new member states. Somewhat lower expenditure levels in Croatia than in new EU member states resulted from lower coverage of unemployed persons with benefits than it is the case in any of the selected countries. How is it possible to reconcile observed similarities in the design of unemployment insurance system with lower somewhat spending and coverage? First, lower spending and coverage were consequences of higher youth unemployment rates in Croatia, who as a rule have less work experience, and therefore even if they are eligible to unemployment insurance, those are quickly exhausted. Moreover, Croatia also has fairly high long-term unemployment and many of the long-term unemployed have exhausted their benefits. Recent increases in relative spending on passive policies in Croatia in part resulted from introduction of the reqirement for war veterans to register, but also from increasing worker flows that brought more dynamics to the labor market and reduction of ILO unemployment rate. It is interesting to look at the effects of the reform introducing activation procedures (including more rigorous job availability conditions and requiring demonstration of job search effort) and closer monitoring of the registered persons into the public employment service at the beginning of 2002. The reform was followed by the drop of the registered unemployment (by about 16 percent in the following year), but it was not the key source of the reduction as the drop in the number of veterans claiming benefits accounted for the majority of the decline. Therefore, it seems that the Croatian case only to a limited extent complies with the view that the strictness with which the benefit system is operated also determines the unemployment benefit system. This might suggest that benefits were well targeted prior to the tightening of the benefit system or that changes to the procedures were not that important as there was no significant reduction in the number of recipients in the aftermath of the reform. 67 Data on passive policies should be supplemented with other forms of social assistance that are available to the unemployed. In Croatia there are as many users of the social assistance amongst the unemployed as there are unemployment benefit recipients, although the amount of social assistance is less than half of the unemployment benefit. 68 There should be some variation in the level of unemployment benefits, based on previous earnings, but in practice it operates as a flat-rate system. 195 Table 6.4: Spending on Passive Labor Market Policies in Selected EU Countries and Croatia Spending on passive labor market policies (in Spending per percentage point of LFS % of GDP) unemployment (*100) 1997 1998 1999 2000 2001 2002 1997- 1997 1998 1999 2000 2001 2002 1997- 2002 2002 Austria 1.28 1.27 1.19 1.05 1.06 1.24 1.18 28.9 28.3 30.1 28.7 29.2 29.8 29.2 Belgium 2.64 2.46 2.32 2.18 2.25 2.40 2.37 28.6 26.3 27.0 31.8 33.7 32.7 30.0 Denmark 3.83 3.41 3.15 3.04 2.95 3.28 72.9 69.9 65.3 68.6 67.9 68.9 Finland 3.10 2.53 2.34 2.13 2.01 2.06 2.36 24.4 22.3 23.0 21.9 22.0 22.6 22.7 France 1.84 1.80 1.75 1.64 1.63 1.81 1.75 15.9 16.2 16.8 18.1 19.4 20.3 17.8 Germany 2.52 2.28 2.12 1.90 1.94 2.13 2.15 26.1 25.1 25.2 24.5 24.7 24.6 25.0 Greece 0.47 0.46 0.44 0.40 0.38 0.43 4.9 4.2 3.7 3.6 3.7 4.0 Ireland 1.91 1.51 1.10 0.79 0.70 1.20 19.4 20.1 19.6 18.4 18.1 19.1 Italy 0.86 0.76 0.68 0.62 0.61 0.63 0.69 7.4 6.5 6.1 6.0 6.5 7.0 6.6 Luxembourg 0.60 0.49 0.44 0.43 0.52 0.50 22.2 18.2 18.3 18.4 24.9 20.4 Netherlands 3.03 2.52 2.12 1.81 1.65 1.72 2.14 61.4 65.9 65.9 62.1 66.0 63.2 64.1 Portugal 0.83 0.80 0.81 0.90 0.91 0.85 12.2 15.6 17.7 22.2 22.9 18.1 Spain 1.78 1.54 1.40 1.34 1.32 1.55 1.49 10.5 10.1 10.9 11.8 12.4 13.7 11.6 Sweden 2.03 1.86 1.62 1.37 1.07 1.05 1.50 20.5 22.7 24.0 24.4 21.9 21.3 22.5 United 0.56 0.43 0.39 0.31 0.27 0.39 Kingdom 81 7.0 6.5 5.7 5.5 6.6 Czech 0.20 0.23 0.30 0.29 0.24 0.27 0.25 Republic 4.2 3.6 3.5 3.3 3.0 3.7 3.6 Hungary 0.63 0.62 0.57 0.47 0.38 0.37 0.51 7.0 7.3 8.2 7.4 6.8 6.6 7.2 Poland 1.06 0.53 0.62 0.81 0.97 1.14 0.85 9.7 5.2 4.6 4.9 5.2 5.8 5.9 Slovak 0.71 0.85 1.05 0.84 0.54 0.48 0.75 Republic 6.0 6.8 6.2 4.5 2.8 2.6 4.8 Croatia 0.56 0.44 0.44 0.53 0.59 0.70 0.54 5.6 3.8 3.3 3.3 3.7 4.7 4.1 EU-15 1.82 1.61 1.46 1.33 1.29 1.62 1.52 20.7 19.8 19.7 20.1 20.8 24.8 21.0 EU-19 1.57 1.39 1.28 1.17 1.13 1.30 1.31 17.7 16.6 15.6 15.0 14.8 16.5 16.0 EU- 4 0.65 0.56 0.63 0.60 0.53 0.57 0.59 7.1 5.9 5.5 4.8 4.2 4.4 5.3 Sources: OECD (2004); OECD (2002); Croatian employment service (1998); Croatian employment service (1999); Croatian employment service (2000); Croatian employment service (2001); Croatian employment service (2002); Croatian employment service (2003). Table 6.5: Comparative Indicators of Unemployment Allowance (Late 1990's) Average unemployment Duration Coverage insurance (in months) (administrative (% of average wage) unemployment) Czech R. 20.0 6 49.2 Estonia 7.2 3-6 55.1 Hungary 23.0 12 73.6 Poland 40.0 12-24 22.9 Slovakia 32.8 6-12 28.0 Slovenia 43.9 3-24 32.6 Average 28 (29 witout EE and SLO) Croatia 26* 3-12** 18.5 * based on the amount of 900 HRK ** prior to the 2004 reform; with the exception of older workers Source: Riboud, Sánchez-Páramo and Silva-Jauregui (2002) 196 Figure 6.7: Tradeoff between EPL and Passive Labor Market Policies 4.0 Croatia (2001) Portugal 3.5 Greece Italy Spain 3.0 France Sweden Belgium EPL index (1999) 2.5 Germany Austria Netherlands Finland 2.0 Denmark 1.5 Ireland 1.0 UK 0.5 0.0 0 5 10 15 20 25 30 35 Spending on passive policies, per unemployed (1999, PPP USD thousands) Sources: Matkovi and Biondi (2003); OECD (1999); Young (2003); Croatian employment service (1998); Croatian employment service (1999); Croatian employment service (2000); Croatian employment service (2001); Croatian employment service (2002); Croatian employment service (2003). Another serious change of the unemployment benefit system was made at the beginning of 2004. Unemployment benefit system became more generous as a compensation for the liberalization of EPL due to reform of the Labor Code (particularly for the reductions of notice period and severance pay). It included an increase of the unemployment benefit ceiling (by about 11 percent, from 900 to 1000 kunas) and extension of benefit duration (from 12 to 15 months). EPL and unemployment benefits are substitutes as insurance from the loss of income. EU and international institutions, particularly the ILO, sometimes advocate substitution of excessive EPL regulation by unemployment benefits as it imposes fewer restrictions on labor market flexibility (Young, 2003). ILO especially highlights an example of Denmark, a country that managed to use passive and active labor market policies along with flexible EPL in order to substantially reduce the risk of income loss for the workers. This recommendation seemed suitable in the Croatian case, which until recently combined fairly low spending on active and passive labor market policies with stringent EPL regulation. Given the low level of expenditures on unemployment benefits, increases in level and duration of benefits seemed justified. However, it seems that the configuration of Croatian labor market institutions was more similar to Greece and Italy, countries that on the aggregate were providing more security to the employed, than to other transition countries. Further on, it was not the amount of average unemployment benefit that distinguished Croatia from other transition countries, but the coverage of the unemployed, which was the lowest in the group. Therefore it seems more justified to consider other ways to increase the coverage of unemployed persons. One should keep in mind that the process of introducing more flexibility into the labor marked shall lead to a more dynamic labor market and easier entry into employment. This might decrease the share of youth among the unemployed and shorten the average duration of the period of unemployment, which can be expected to increase the coverage and expenditure per unemployment percentage. Extending the duration of unemployment benefits should be particularly carefully introduced as some studies indicate that the duration of unemployment benefits increases the average duration of unemployment more than the benefit level (Cazes and Nesporova, 2003). Finally, greater reliance on unemployment benefits as an insurance form the loss of income should come in pair with measures to activate the unemployed that are in part implemented through the reform of the employment service. 197 ACTIVE LABOR MARKET POLICIES Given the diversity of measures falling in the group of active policy measures in the labor market (APLM), it is very difficult to give them a common description. However, they have three main characteristics in common (Dar and Tzannatos, 1999): ˇ Mobilization of the labor supply, most often through employment schemes and employment subsidies. ˇ Development of capacities and skills that stimulate and facilitate employment through measures such as training. ˇ Promotion of labor market efficiency through measures such as counseling and job search assistance services. From the above mentioned, it is clear that active policy measures, unlike the passive policy measures which are used to attempt to financially insure the unemployed, are most often used to reduce imperfections which exist in the labor market and promote more efficient outcomes as well as to modify those market outcomes towards more socially acceptable ones. In other words, those measures are usually used to increase employment and wages of particular categories in the labor market which are thought to be particularly exposed to the risk of unemployment and poverty. The level of spending on active labor market policies in Croatia during the late 1990's did not significantly differ from the spending levels in selected Central and Eastern European countries. The only exception was Hungary with somewhat higher spending. However, average spending in all those countries was far below average spending in 'old' EU members, which typically exceeded 1 percent of GDP. Considerably lower level of expenditures probably reflects financial constraints, but also limited experience with the design and implementation of those policies. Table 6.6: Spending on Active Labor Market Policies total spending (in % of GDP) Spending per percentage point of unemployment,*100 1997- 1997- 1997 1998 1999 2000 2001 2002 1997 1998 1999 2000 2001 2002 2002 2002 Austria 0.45 0.44 0.51 0.50 0.53 0.53 0.49 10.16 9.80 12.90 13.69 14.60 12.76 12.32 Belgium 1.23 1.41 1.31 1.33 1.34 1.25 1.31 13.33 15.10 15.24 19.37 20.08 17.05 16.70 Denmark 1.66 1.69 1.79 1.59 1.68 31.61 34.67 37.09 35.94 34.83 Finland 1.54 1.40 1.23 1.00 0.95 1.00 1.19 12.13 12.33 12.10 10.26 10.42 10.99 11.37 France 1.35 1.31 1.38 1.32 1.29 1.25 1.32 11.70 11.83 13.20 14.51 15.36 14.05 13.44 Germany 1.23 1.26 1.32 1.22 1.19 1.17 1.23 12.74 13.84 15.70 15.74 15.17 13.50 14.45 Greece 0.34 0.46 0.40 3.55 4.18 3.87 Ireland 1.13 1.13 29.13 29.13 Italy 0.42 0.58 0.58 0.60 0.63 0.57 0.56 3.63 4.96 5.15 5.77 6.67 6.34 5.42 Netherlands 1.47 1.59 1.72 1.69 1.74 1.84 1.68 29.77 41.57 53.44 58.02 69.59 67.63 53.34 Portugal 0.78 0.77 0.82 0.61 0.75 11.52 14.94 18.03 15.06 14.89 Spain 0.49 0.65 0.96 0.92 0.83 0.85 0.78 2.89 4.27 7.49 8.13 7.81 7.52 6.35 Sweden 2.02 1.96 1.78 1.37 1.39 1.39 1.65 20.39 23.91 26.45 24.44 28.50 28.15 25.31 UK 0.38 0.33 0.36 0.36 0.35 0.37 0.36 5.53 5.35 6.08 6.68 6.97 7.25 6.31 Czech R. 0.12 0.14 0.18 0.22 0.22 0.18 0.18 2.48 2.21 2.10 2.55 2.76 2.47 2.43 Hungary 0.44 0.39 0.41 0.39 0.48 0.52 0.44 4.89 4.63 5.97 6.20 8.60 9.32 6.60 Poland 0.48 0.46 0.35 0.23 0.13 0.11 0.29 4.42 4.52 2.61 1.40 0.70 0.56 2.37 Slovakia 0.22 0.32 0.37 0.45 0.34 1.31 1.71 1.91 2.40 1.83 Croatia 0.17 0.23 0.47 0.34 0.23 0.13 0.26 1.68 1.99 3.47 2.13 1.47 0.91 1.94 EU-15 1.03 1.07 1.15 1.04 1.03 1.02 1.04 13.00 15.14 18.57 18.97 20.39 18.52 17.69 EU-19 0.90 0.93 0.93 0.85 0.84 0.82 0.88 11.30 13.01 14.68 14.97 15.88 14.28 14.50 EU- 4 0.35 0.33 0.29 0.29 0.30 0.32 0.31 3.93 3.79 3.00 2.96 3.49 3.69 3.31 Sources: OECD (2004); OECD (2002); Croatian employment service (1998); Croatian employment service (1999); Croatian employment service (2000); Croatian employment service (2001); Croatian employment service (2002); Croatian employment service (2003) 198 Although overall spending on labor market policies in selected Central and Eastern European countries was considerably below that in most 'old' EU members, there are no important differences in terms of the balance between active and passive policies. The ratio of spending on active to spending on passive policies was roughly 0.7 to 1 in 'old' EU members, but it was significantly lower in selected Central and Eastern European countries. Spending on passive labor market policies surpassed spending on active policies in all of the selected countries but Sweden. However, ratio of spending on active to passive policies increased during the observed period as OECD (1994) in the first half of the 1990's called for a progressive shift of resources from passive income support to active measures, but also on their enhancement, as those policies improve access to the labor market and help "push" the unemployed persons into the employment (Nickell, 2005). Spending on passive labor market policies was in Croatia on average twice as large as spending on active policies, reflecting once again similarities to Central and Eastern European countries. Table 6.7: Structure of Spending on Labor Market Policies Spending on labor market policies Spending per percentage point of unemployment (*100) Passive Active Total Passive Active Total Austria 1.18 0.49 1.68 29.2 12.3 41.5 Belgium 2.37 1.31 3.69 30.0 16.7 46.7 Denmark 3.28 1.68 4.96 68.9 34.8 103.8 Finland 2.36 1.19 3.55 22.7 11.4 34.1 France 1.75 1.32 3.06 17.8 13.4 31.2 Germany 2.15 1.23 3.38 25.0 14.4 39.5 Greece 0.43 0.40 0.83 4.0 3.9 7.9 Ireland 1.20 1.13 2.33 19.1 29.1 48.3 Italy 0.69 0.56 1.26 6.6 5.4 12.0 Netherlands 2.14 1.68 3.82 64.1 53.3 117.4 Portugal 0.85 0.75 1.60 18.1 14.9 33.0 Spain 1.49 0.78 2.27 11.6 6.4 17.9 Sweden 1.50 1.65 3.15 22.5 25.3 47.8 UK 0.39 0.36 0.75 6.6 6.3 12.9 Czech R. 0.25 0.18 0.43 3.6 2.4 6.0 Hungary 0.51 0.44 0.94 7.2 6.6 13.8 Poland 0.85 0.29 1.15 5.9 2.4 8.3 Slovakia 0.75 0.34 1.09 4.8 1.8 6.6 Croatia 0.54 0.26 0.81 4.1 1.9 6.0 EU-15 1.52 1.04 2.56 21.0 17.7 38.7 EU-19 1.31 0.88 2.18 16.0 14.5 30.5 EU- 4 0.59 0.31 0.90 5.3 3.3 8.6 Sources: OECD (2004); OECD (2002); Croatian employment service (1998); Croatian employment service (1999); Croatian employment service (2000); Croatian employment service (2001); Croatian employment service (2002); Croatian employment service (2003). Although macroeconomic evidence indicates that active labor market policies contribute to the reduction of unemployment and particularly long-term unemployment (European Commission, 2004; Cazes and Nesporova, 2003), it does not necessarily mean that every such program was successful Therefore, actual impacts of particular policies are usually carefully evaluated or such programs are first implemented on experimental basis. There is hardly any information on the impacts on active labor market policies in Croatia and the scarce evidence that exists is not encouraging. Dorenbos and van Winden (2001), the only rigorous study known to the author, report that program of public works performed in 1999 did not have a significant impact on employability of participants. However, this program was performed only for a short period of time and employed only a limited amount of resources. It is possible to gain somewhat better insight into the effectiveness of active labor market policies by inspection of their structure. 199 If the structure of expenditures on active policy measures in the `old' EU members is observed, an absolute domination of various training programs and subsidized employment is visible, to which more than half of the total expenditures was allocated, while other forms of active labor market policies were less important. The expenditures structure in the group of advanced transition countries somewhat differed from that average, as they relied more on subsidized employment and less on training programs. Although the expenditure levels on active policy measures in Croatia did not significantly differ from the group of advanced transition countries, it even more deviated from the structure observed in `old' EU members. Moreover, most of the wage subsidies in Croatia went to regular employment in the private sector rather than direct job creation, which is more common in other countries. Wage subsidies and measures stimulating self-employment and small enterprise were of special importance in Croatia, according to the structure of funds spent by the CES, while other countries employed less wage subsidies and more public sector jobs. Those wage subsidies were general, meaning that most of the unemployed were eligible for some form of wage subsidy. The goal of such general wage subsidies was not formaly announced, although it might be assumed that they aimed to increase total employment and reduce unemployment. However, some of the studies report stronger impact of the active labor market policies on long-term unemployment than on total unemployment (Cazes and Nesporova, 2003). Moreover, rigorous empirical evaluations indicate that net job creation, after deadwight, substitution and displacement effects are excluded, rarely exceeds a third of subsidized persons and far often remains below that (Dar and Tzannatos, 1999). Therefore, the efficiency of implemented active labor market policies is questionable. Table 6.8: Structure of Spending on Active Labor Market Policies in % of GDP in % of total active labor market policies Sub- Sub- Measures Public Youth Measures Publ. Train- sidized Train- Youth Sidized empl. mea- for the Total empl. for the ing employ- ing measures employ- services sures disabled services disabled ment ment Austria 0.14 0.18 0.03 0.09 0.06 0.49 27.4 36.1 6.8 18.6 11.1 Belgium 0.20 0.26 0.01 0.73 0.12 1.31 15.0 19.8 0.4 55.5 9.3 Denmark 0.12 0.94 0.10 0.24 0.28 1.68 7.1 55.7 5.9 14.4 16.8 Finland 0.12 0.38 0.19 0.39 0.11 1.19 9.8 32.2 16.2 32.6 9.3 France 0.17 0.28 0.38 0.40 0.09 1.32 13.0 21.3 28.5 30.5 6.7 Germany 0.23 0.34 0.08 0.32 0.28 1.23 18.3 27.6 6.2 25.6 22.3 Greece 0.09 0.14 0.10 0.07 0.01 0.40 22.5 33.8 23.8 17.5 2.5 Ireland 0.24 0.15 0.18 0.53 0.03 1.13 21.2 13.3 15.9 46.9 2.7 Italy 0.07 0.21 0.28 0.00 0.56 12.4 37.6 50.0 0.0 Netherlands 0.28 0.45 0.05 0.34 0.55 1.68 16.9 26.9 3.0 20.2 33.0 Portugal 0.11 0.25 0.26 0.09 0.04 0.75 14.8 33.6 34.2 12.4 5.0 Spain 0.09 0.22 0.06 0.39 0.03 0.78 11.3 27.4 7.7 50.2 3.4 Sweden 0.4 0.02 0.41 0.54 1.65 22.3 1.4 24.8 32.5 UK 0.0 0.13 0.02 0.02 0.36 12.1 37.2 4.7 5.6 Czech R. 0.07 0.02 0.02 0.05 0.02 0.18 39.6 12.3 9.4 28.3 10.4 Hungary 0.12 0.07 0.00 0.25 0.00 0.44 27.4 16.0 0.0 56.7 0.0 Poland 0.02 0.08 0.10 0.10 0.29 5.1 28.4 33.0 33.5 Slovakia 0.15 0.02 0.00 0.14 0.03 0.34 44.9 5.9 0.7 40.4 8.1 Croatia 0.11 0.01 0.04 0.13 0.00 0.26 40.4 2.4 15.3 47.8 0.0 EU-15 0.16 0.29 0.13 0.31 0.15 1.04 15.6 27.9 12.3 29.6 14.8 EU-19 0.15 0.23 0.11 0.27 0.13 0.88 17.3 26.5 12.0 30.6 14.5 EU- 4 0.11 0.03 0.03 0.13 0.04 0.31 36.6 10.1 8.2 42.7 11.5 Sources: OECD (2004); OECD (2002); Croatian employment service (1998); Croatian employment service (1999); Croatian employment service (2000); Croatian employment service (2001); Croatian employment service (2002); Croatian employment service (2003) The main priority in the area of active labor market policies should be to develop `evaluation culture' and routinely test for their effects. This would help avoid campaign approach to active labor market policies, spending major funds and then closing the programs, which characterized implementation of active labor market policies until now. Programs should be initiated with modest funds and decide on 200 their expansion or closing and redirecting funds into other programs only after evaluations. Programs should also be better focused to long-term unemployed and those at danger of falling into the long- term unemployment rather than to everyone, which proved to be inefficient in many countries. TAXES AND LABOR COSTS Labor taxes and social contributions (direct taxation) are usually one of the most abundant sources of tax revenues in most countries. It is, therefore, not surprising that direct taxes make about half of tax revenues in Croatia. Some studies attribute significant part of unemployment growth to increased taxation of labor (Nickell, Nunziata i Ochel, 2005), which discouraged job creation by increasing labor costs and participation in the labor force by reducing take-home wage. These studies therefore call attention to the goal of many European countries to reduce labor taxes and contributions in order to encourage work and employment. Despite recent efforts to reduce direct taxation, average tax wedge on labor is still high in most European countries - it averaged about 40 percent in EU members in 2003 (which is 1 percentage point less than in the previous year), while in four advanced transition countries that are also OECD member states it averaged 44 percent. With such a perspective in mind, average tax wedge that amounted to 39 percent of total labor costs in Croatia does not seem particularly bad. Tax wedge in Croatia was about 1 percentage point below EU average and as much as 5 percentage points below average of selected advanced transition countries69. Table 6.9: Average Tax Wedge (as percent of average production worker's total labor cost, 2003) Income Social Security Contributions Total Social Security Total Tax Tax Employee Employer Contributions Wedge Austria 8 14 23 37 45 Belgium 20 11 23 34 54 Denmark 32 11 1 12 44 Finland 20 5 19 24 44 France 9 10 29 39 48 Germany 0 12 22 34 34 Greece 10 5 10 15 25 Ireland 14 7 25 32 46 Italy 8 12 12 24 32 Luxembourg 7 22 14 36 43 Netherlands 17 17 17 34 51 Portugal 5 9 19 28 33 Spain 9 5 23 28 37 Sweden 18 5 25 30 48 United Kingdom 14 8 9 17 31 'old' EU average 13 10 17 27 40 Czech R. 9 9 26 35 44 Hungary 10 9 27 36 46 Poland 5 21 17 38 43 Slovak R. 5 9 28 37 42 Average Central Europe 7 12 25 37 44 Croatia 7 17 15 32 39 OECD average 12 8 15 23 35 Sources: OECD and authors' calculation. 69 International comparisons of average tax wedge should be interpreted with caution. Individual tax systems are usually quite complicated with many different rates, deductions and exceptions, so tax wedges calculated on the basis of an average production worker can be misguided. It may particularly be the case if tax systems differ with respect to their progressiveness, if there are differences in income distributions or if there are differences in public goods and services workers receive from the state. However, European Commission (2004) reckons that average tax wedge is a fair proxy as there is high correlation between tax wedges for different socio-economic groups. 201 Burden of labor taxes and contributions in transition was especially heavy if their income levels are taken into the account (compared to poorer 'old' EU members, such as Greece and Portugal). Such a heavy burden was a consequence of widespread use of early retirement schemes and rise in social expenditures that accompanied growing and high unemployment, which helped create a social consensus on the necessity of reforms. On the other hand, taxation level discouraged job creation and shifted part of formal employment into the sphere of informality. Comparatively low taxation burden on labor in Croatia did not occur as an accident, but as a result of persistent policies intended to reduce tax wedge. The total tax wedge on labor in 1995 amounted to 48 percent of total labor costs, which was exceptionally high, even by the merits of 'old' EU member states and transition countries. Until 2001 the tax wedge was reduced to 41 percent and its decreasing tendency was continued afterwards (Rutkowski, 2003). Having all this in mind, it seems worth asking whether it is necessary to pursue even further reductions of the labor taxation, as it was recently discussed in the public, and how to do it. There are several reasons why labor tax wedge should be further decreased, irrespective of recent efforts in many European countries to do so (European Commission, 2004). First of all, labor tax wedge in Croatia is still higher than it is in poorer 'old' EU members (Greece and Portugal) that are more comparable to Croatia than the EU average. Further on, total labor costs in Croatia are higher than labor costs in transition countries with similar productivity levels, such as Poland and Hungary, so unit labor costs in Croatia are the highest among the Central and Eastern European transition countries. Reduction of labor taxes could probably temper growth of labor costs to a greater degree than reduction of other taxes and therefore improve cost competitiveness. Finally, Croatian unemployment rate, as previously discussed, is despite recent decline still high, and easing the pressure on labor costs by reductions of labor taxes could help the recovery of employment. While reduction of labor taxes and contributions can be a mid-term goal, in the short run reducing taxation of individuals with poor qualifications that earn low incomes70 should be s priority because of their high unemployment. Reduction of labor taxation has been implemented in many different ways during the previous decade - by increasing existing tax deductions and introducing new exemptions, reductions of tax rates and contribution rates. There is wide range of instruments now as well, but each of those instruments bear specific implications for different groups of workers. Since workers in Croatia on average pay about five times as much contributions as taxes, which is specific for all transition countries due to unfavorable demographics and widespread use of early retirements, even a modest reduction of contributions could influence income more than a significant reduction of income tax. Such an effect is especially visible in case of low-income workers. For example, taxes account for about 0.5 percent of total labor costs of worker earning minimum wage (in 2005), while total tax wedge, comprised mostly of social security contributions, amount to 32 percent of total labor costs. Therefore, manipulation of taxes (as applied in 2005) bears almost no effect on incomes and labor costs of low-income workers. An alternative route, if one assumes that no negative taxes will be introduced, would be to reduce social security contributions. However, one has to approach such an issue with a caution. Pension reform was an important step toward establishing a link between pension contributions and pensions and reduction of pension contributions could aggravate that link. Reduction of health contributions would not have such an effect as health contributions are not linked to the actual health benefits and therefore resemble taxes. Moreover, health insurance, while a universal benefit, is largely financed through contributions paid by the employees. Labor costs are not considered as labor market policies/institutions, but rather an outcome of an interaction between the supply and demand within the particular set of institution, such as type and coverage of collective bargaining agreements. However, their dynamics affects dynamics of job creation and decisions to participate in the labor force. Although it seems that the level and tendency of labor taxation did not recently exert upward pressure on total labor costs in Croatia, these costs were significantly above labor costs in other transition economies. 70 There are different ways to implement it in practice, such as negative taxes, which is, however, hard to recommend in Croatia because it would strengthen incentives to underreport incomes. 202 Table 6.10: Labor Costs Monthly Total Labor Costs (EUR) Unit Labor Costs (EU-15 = 100) 2001 2002 2003 2001 2002 2003 EU-15 2,511 2,565 2,610 100.0 100.0 100.0 Hungary 416 488 505 58.5 65.8 67.7 Slovakia 394 437 478 64.7 66.8 70.4 Slovenia 1,146 1,207 1,332 91.3 91.2 96.5 Bulgaria 162 174 187 64.0 64.9 68.6 Romania 223 236 243 82.7 75.1 69.6 Croatia 790 848 871 105.4 107.1 106.3 Note: Total Labor Costs includes gross wages as well as employers' contributions. Sources: OECD, Central Bureau of Statistics, National statistical offices. FLEXIBILITY AND LABOR MARKET FLOWS According to Monastiriotis (2003), labor market flexibility refers to the extent to which labor market forces determine labor market outcomes, or the absence of any factors entering the labor market other than supply and demand. This approach looks at flexibility as an outcome, which can be measured, rather than some unrealized potential. However, labor market flexibility is neither uniform nor homogeneous. In line with this notion as well as the discussion above of the studies assessing the flexibility of the Croatian labor market, labor market flexibility can be decomposed in different ways. One logical ways is along two axes: numerical versus functional and internal versus external flexibility. This decomposition gives four different types of flexibility: internal numerical flexibility (adjustability of labor inputs already employed by the firm ­ working hours, working time, leave and holidays), external numerical flexibility (adjustability of labor intake from the external labor market), internal functional flexibility (ability to improve efficiency by reorganizing the methods of production and labor content) and finally external functional flexibility (ability to externalize or diversify parts of production through sub-contracting). These categories overlap to some extent with the approach employed by Lowther (2003) as well as with the indicators Rutkowski (2003) used. Monastiriotis prefers to take a somewhat wider perspective encompassing three different broad domains of flexibility: production function flexibility, labor costs flexibility and supply side flexibility, which further collapse into smaller sub-domains. Table 6.11: Types of Labor Market Flexibility Labor market flexibility Production-function Labor-Costs Supply - side Wage costs (pay) Flex. in Flex. in Determination of Determination of Flex. in Labor Flex. in labor work reservation average wages non- mobility skills input content wages wage acquisition costs Source: Monastiriotis (2003). There are many possible impediments to flexibility, defined in this manner. EPL and other regulations are not the only forces shaping labor market flexibility, but they are prominent in practice. Also, regulations are more likely to affect particular dimensions of flexibility, such as numerical flexibility or flexibility in labor input and labor mobility. Therefore, it is possible for the labor market to retain a certain level of flexibility regardless of increased regulation due to compensating movements in other areas and conversely, more flexible regulation may not to result in the expected increases in overall flexibility. For example, Abraham and Houseman (1993) find that adjustment of the employment level to a fall in demand is much slower in Belgium and Germany than in the United States, but adjustment in the hours of work is similar, which means that internal flexibility almost fully compensates for the lack of external flexibility. However, it is also possible that different restrictions sometimes reinforce 203 each other. For instance, stricter firing regulation is likely to increase insider power of employees and hence reduce wage flexibility (Rutkowski, 2003). Two reports backing the reform of the Croatian Labor Code argue decisively that reform is necessary in order to enhance labor market flexibility. However, as can be seen from the discussion above, there is no agreement on the proper way to define labor market flexibility, let alone to measure it. Moreover, although some of the effects arising from relaxation of EPL have been touched upon in both reports, the discussion from the literature on the overall balance has not been adequately covered. There are some potential benefits of EPL that have to be taken into the account when considering its impact on labor market. The first is the reduction of uninsurable risk, which is the standard reason for the introduction and support of the EPL. In addition to this, EPL may encourage human capital investment by making the relationship between employer and employee more secure. This should have beneficial effects on productivity. EPL also internalizes part of the social costs arising from worker dismissals (OECD, 1999). On the other side, as was properly addressed in the report by Rutkowski (2003), rigid labor market regulation is often blamed for increasing unemployment and decreasing employment in transition countries as well as in Western European countries that have enacted such regulation. However, some authors have taken position that increasing unemployment is not the most costly feature of EPL. Aghion and Howitt (1994) suggest a relationship between job creation, job destruction, and productivity growth. They model a process of creative destruction in which job turnover leads to labor productivity increases. The logic of their theoretical model suggests that firms engaging in restructuring destroy low productivity jobs and create high productivity ones. Therefore, a positive correlation between productivity growth and job turnover, facilitated by low level of EPL, might be expected. To the extent that productivity growth results from the entry of new firms and the experimentation they introduce, the most harmful effect of EPL may not be on job turnover between existing firms, but rather on entry and exit of firms. Scarpetta, Hemmings, Tressel and Jaejoon (2002) report that shifts in market shares of operating firms influence productivity only modestly, while entry and exit of firms can account for 20 percent to 40 percent of total productivity growth. Their empirical analysis confirms that EPL has a strong effect on market access of small and medium-sized firms, which indirectly affects productivity growth. Although there are no similar studies performed in transition countries, intuition would suggest that both of these effects would be particularly harmful in a transition economy needing rapid reallocation of jobs and workers away from the old, inefficient sectors. A high degree of job reallocation may also have some negative effects, at least in the short run, in terms of worker displacement and earnings losses, but the aggregate and long-run benefits are likely to compensate the individual costs (Faggio and Konings, 2003). Having all this in mind, there are several good reasons to focus on labor market flows or external numerical flexibility in order to assess the impact of EPL on labor market flexibility71. First of all, given the theoretical background, the assumption that job and worker flows may capture the impact of labor market regulations such as EPL on labor market adjustment seems acceptable. Theoretical models are mostly concerned with the impact of EPL on job flows, but it is clear that job turnover also affects worker turnover. Although there are supply-side reasons, such as job-shopping, human capital acquisition, career progression and events that affect preferences regarding work (like children), it is a consensus now that there is a major role for demand-side disturbances in explaining worker mobility (Davis and Haltiwanger, 1999). Therefore, it is reasonable to assume that all regulations affecting firms' decisions to employ or dismiss workers have a decisive impact on both job and worker flows. Furthermore, job and worker flows are already widely used to evaluate the functioning of the labor market, which generates some comparative empirical evidence. Faggio and Konings (1998) emphasize 71 Thorough definitions of labor market flows are provided in Appendix A4. 204 that job flows reflect the processes of reallocation and restructuring underlying the reform in transition countries, at least for the initial restructuring that followed the imposition of hard budget constraint and start of labor shedding. Haltiwanger and Vodopivec (2002) consider a labor market to be flexible when it is characterized by the simultaneous presence of contracting and expanding firms. Faggio and Konnings (2003) adopt this definition and supplement it with a notion that in a flexible labor market, workers can move and jobs can be easily created/ destroyed in order to meet new economic conditions. Davis and Haltiwanger (1999) stress that the extent to which labor reallocation and matching process operates smoothly measures the difference between successful and unsuccessful macroeconomic performance. Bertola, Boeri and Cazes (1999) as well as Cazes and Nesporova (2003) emphasize the magnitude of hirings and firings, or worker flows as a means of capturing the impact of regulation on labor market adjustment. Unlike those studies, Monastiriotis (2003) considers the extent of alternative types of employment, such as work at home, part-time and temporary work, to capture external numerical flexibility, which is related to EPL. However, while confining worker mobility and tenure indicators to the supply-side flexibility sub-index, he acknowledges that there is significant overlap between different types of flexibility and certain indicators do not correspond directly and exclusively to one group only. Although it may be difficult to distinguish between worker and job mobility driven by labor supply and labor demand and therefore hard to judge how to represent each of the flexibility dimensions, this paper supports the more conventional view that labor demand is guiding the process of worker and job reallocation and therefore employs those indicators to assess the impact of labor market regulation. There are also a few disadvantages associated with the use of worker and job flows to approximate labor market flexibility. First of all, these indicators are not readily available and they are usually the products of comprehensive research, such as that of Rutkowski (2003). Furthermore, research methodologies conducted for different countries usually differ, as one study uses firm level data, while others also capture within-firm flows that arise between different establishments. There are also some studies that seek to capture flows within establishments, which are the hardest to measure, but give the most accurate measure of flows (Davis and Haltiwanger, 1999). Even if there is the same basic unit of observation, definitions of firms and establishments do not necessarily coincide between different countries and periods. It is also hard to adequately capture organizational changes and change of ownership over time. Finally, definitions of jobs, sampling intervals and coverage of economic activities differ as well. All these problems combine to make data on labor market flows hard to compare. LABOR MARKET ADJUSTMENT IN INTERNATIONAL PERSPECTIVE The most comprehensive work undertaken so far on the impact of EPL on labor market adjustment are the OECD (1997) and OECD (1999) studies conducted for members of that organization. The first study finds job turnover rates to be high across all member states, well in excess of net job creation rates. Also, the flows appear to be remarkably stable for most OECD member countries during the late 1970's ­ early 1990's period, even for those that significantly liberalized their labor market regulation at that time, although there is some variation along the business cycle. The most interesting finding appears to be on the relationship between EPL and job turnover, which is statistically insignificant, while the link between the EPL governing temporary workers and job turnover is somewhat stronger. If there is any relationship between the EPL and turnover, it looks as if a high level of EPL dampens only the cyclical fluctuations of job turnover rates. Also, restrictive EPL may transfer some of the turnover from permanent to temporary workers and from large to small companies, which are less likely to be affected by the regulation. The subsequent study (OECD, 1999) included different measures of labor market dynamics and found some evidence of stricter EPL having somewhat stronger effects on worker turnover than job turnover as it decreases churning of workers between different jobs. This effect is also visible in higher mean job tenure in countries with stricter EPL. As explainedabove, the average EPL level in Central and Eastern European transition countries does not differ much from the EU average. There are large differences between these countries as well, 205 reflecting the fact that many of them shaped their labor market regulation according to legislation in neighboring EU countries (for example, EPL features in Estonia are similar to Sweden, while Slovenian EPL reflects Italian influences; Riboud et al., 2001). This means that transition countries on average have a fairly restrictive labor market regulation and there is also ample variation amongst them, allowing the effects of those differences to be explored. Table 6.12: EPL Index in Transition Countries EPL index EPL components Regular Temporary Collective employment Employment dismissals Bulgaria 2.5 1.9 3.4 1.8 Croatia (before 2003 reform) 3.6 2.8 3.9 5.0 Croatia (after 2003 reform) 2.7 2.6 2.6 3.5 Croatia ­ gov't proposal 2.3 2.4 1.6 3.5 Czech Republic 2.1 2.8 0.5 4.3 Estonia 2.6 3.1 1.4 4.1 Hungary 1.7 2.1 0.6 3.4 Poland 2.0 2.2 1.0 3.9 Romania 2.8 1.7 3.0 4.8 Slovak Republic 2.4 2.6 1.4 4.4 Slovenia 3.5 3.4 2.4 4.8 Average (Croatia excl.) 2.5 2.5 1.7 3.9 EU average 2.4 2.4 2.1 3.2 Sources: Rutkowski (2003), Biondi and Matkovi (2003) and Micevska (2003). Comparative studies of labor market flows in transition countries are particularly scarce, while there are a few more utilizing case-study approaches. Important comparative studies on labor market adjustment in transition countries include recent papers by Faggio and Konings (2003) as well as Cazes and Nesporova (2003). Faggio and Konings (2003) examine job flows in five transition countries. Their study concludes that in countries with more rapid reform, job creation catches-up faster with job destruction. Moreover, the study finds Estonia to be the most dynamic among the transition countries, as job turnover rates in Estonia resemble the behavior of more dynamic market economies, such as the US or UK. Other transition countries included in the study, Bulgaria, Romania, Poland and Slovenia, are more similar to regulated labor markets of Western European countries. What is particularly striking about these job flows is that the volumes involved are not exceptional, as one might expect from the transition process requiring a massive reallocation of resources. However, this finding might be biased because flow data are extracted from the AMADEUS enterprise database comprising data for medium and large enterprises, while small enterprises, which usually account for a substantial part of the overall job turnover, are not included in the database. Table 6.13: Job Flows in Transition Countries (Faggio-Konings dataset) Job creation Job turnover Excess job reallocation Bulgaria 1994-1997 5.7 8.1 4.8 Estonia 1994-1997 8.1 16 13.5 Poland 1994-1997 5.4 8.5 6.3 Romania 1994-1997 9.0 13.0 8.0 Slovenia 1995-1997 5.2 9.6 8.6 Source: Faggio and Konings (2003). If EPL indices for transition countries and data on job flows reported by Faggio and Konings (2003) are observed together, there seems to be no apparent relationship, which is similar to previous findings for OECD member countries. Since EPL data cover late 1990's period, there is a possibility that reform of the labor market regulation in some of those countries changed the EPL level somewhat compared to the mid 1990's period, to which jobs flow data refer. But it is unlikely that their relative rankings changed much. Also, it remains an open question to what extent enterprises in the sample 206 used for the calculation of job flows are representative of the entire enterprise population in those countries. Figure 6.8: EPL Index and Job Flows (Faggio-Konings dataset) 14 Estonia 13 job turnover 12 turnover = 1,47 EPL indeks + 4,30 11 2 R = 0,06 10 9 Slovenia 8 Romania 7 Poland 6 5 Bulgaria 4 1.5 2 2.5 3 3.5 4 EPL index Sources: Same as for tables 2 and 3. Cazes and Nesporova (2003), using a bivariate approach, find some connection between EPL index, especially the difficulty of dismissal sub-index, and worker turnover in six transition countries (Bulgaria, Czech Republic, Estonia, Poland, Russia and Slovenia). However, the link is very weak, as it appears that the impact of other factors, such as macroeconomic and structural reforms, passive and active labor market policies, remains important. Also, their results seem to heavily depend on Slovenia, which has a considerably higher EPL level than other countries in the sample and somewhat lower worker turnover than the average. Papers employing case-study approach provided some similar conclusions to these two studies. Haltiwanger and Vodopivec (2000) also deem Estonia to be a success story due to aggressive pursuit of decentralized wage setting and labor market reform. They contrast their findings with Slovenia, which pursued a more gradual labor market reform, provided more income support and implemented a higher level of EPL. Policies implemented in Slovenia increased the costs of separation for Slovenian firms, which resulted in a lower level of job flows, with the highest job creation rate in Slovenia during the early 1990's barely surpassing half of the observed maximum rate in Estonia. Jurajda and Terrel (2001) contrast Czech and Estonian transition experiences, where Estonia represents a benchmark for rapid transition with high job turnover rates, while the Czech labor market exhibits lower job turnover rates during the early transition, which is symptomatic of a Czech gradualist approach to transition. Therefore, it seems that most comparative insights into the dynamics of job flows in transition countries rest on a few case studies finding contrasts between Estonia and other transition countries with less flexible labor markets, such as Slovenia, Poland or Czech Republic. Table 6.14: Job Flows in Transition Countries (Rutkowski Dataset) Job creation Job turnover Excess job reallocation Bulgaria 2000 6.8 17.6 13.5 Croatia 2001 3.5 8.4 7.0 Lithuania 1998-1999 9.7 20.4 19.4 Poland 1998-1999 5.3 15.4 10.5 Sources: Rutkowski (2003). 207 LABOR MARKET ADJUSTMENT IN CROATIA In his study of the Croatian labor market dynamics, Rutkowski (2003) especially emphasizes data on job flows and tenure structure. With a high EPL index and low job turnover rates reported by Rutkowski, the Croatian labor market most closely resembled the Slovenian case as an archetypal low- flexibility economy. This part of the study will replicate job flow indicators and will also look at worker flow data as well as their particular characteristics, since the motivation for using these indicators in studies of labor market adjustment has already been extensively elaborated. Job turnover rates in Croatia during the 1994-2004 period calculated using enterprise annual reports fluctuated around 15,6 percent, which is in the middle of the 10 percent-20 percent range of job turnover observed in most market economies (OECD, 1999). This job turnover rate is approximately double Rutkowski's (2003) calculations for the value of indicator 2001 (the actual level of job turnover indicator calculated in this study for 2001 was 16.3%). Such a large difference between those two sets of job flow indicators could arise from two possible sources. The first is the sampling procedure used by Rutkowski (2003). This study, in contrast, encompasses the whole population of reporting enterprises72. As job turnover rates exhibited by different enterprises differ wildly, any bias towards government owned and large enterprises would severely reduce job turnover. The second reason for possible differences is the data cleaning procedure, which eliminated some job turnover arising from errors in the data73. Details on data sources and cleaning procedure used in this study are reported in Appendix B4. Figure 6.9: GDP Change, Employment Change and Job Flows in Croatia 10% 5% 0% -5% -10% GDP change Employment change Job creation Job destruction Net job creation -15% 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Source: Croatian Bureau of Statistics; authors calculations based on FINA database. The stagnation of the job turnover rate in Croatia during the 1994-2004 period resulted from falling job destruction rates and increasing job creation, which finally caught up with job destruction in 2001 and subsequently got ahead of it, as shown in Figure 4.2. Such a pattern is typical for a transition economy, but there are several features of this process specific for Croatia. First of all, the variation of job creation and job destruction rates over the observed period is small, having in mind the magnitude of the underlying economic change. Further on, job destruction surpassed job creation for a long time, 72 Rutkowski drew a random sample of 12 thousands firms out of the full population, representing about a quarter of population. The sample was constructed to ensure representation of firms by ownership and region. However, it seems that sample enterprises differ from the population with respect to some of their properties. For instance, while enterprises with up to 50 employees account for less than 14 percent of total employment in Rutkowski's sample, they comprise one third of total employment in entire population of enterprises. Similarly, enterprises with more than 500 employees make up almost half of the total employment in the sample, while their employment share in the whole population of enterprises is approximately one third. 73 Sample bias alone does not explain different result because job turnover measures within certain size classes according to the whole population are somewhat higher than the sample-based turnover measures. 208 until 2001, indicating that the adjustment process has been subdued. The magnitude of the job flows does not seem particularly low in international comparison, as describe above. However, job turnover also does not appear to be as high as might be expected due to the magnitude of the underlying economic change, supporting the view of the protracted adjustment in the labor market. Finally, job turnover exhibits weak overall pro-cyclical properties, which grows somewhat stronger when job turnover is regressed on lagged GDP growth. It was mostly due to pro-cyclical behavior of job creation, while job destruction does not seem to react to swings in economic activity. Such a behavior is untypical for market economies, where more variable counter-cyclical job destruction dominates over pro-cyclical job creation (Davis and Haltiwanger, 1999). This means that employment is quite resistant to short-term economic fluctuations and adjusts sluggishly. The tendency of slowly increasing job creation seems to arise from the growing portion of the more dynamic, newly established private enterprises, in total employment and contraction of the state owned enterprises. Job destruction falls as this new sector, also accounting for major art of job destruction, matures and employment patterns in surviving companies become more stable. Low correlation between the job flows and GDP proves the disconnection of the employment dynamics from GDP dynamics during the observed period. The adjustment of the employment level to the transition shock in Croatia was, therefore, a prolonged process that lasted until 2001, regardless of the non-negligible level of overall job turnover and significant excess job reallocation. Comparison of the reported net job to the change in the overall employment confirms quality of the data. The correlation coefficient between the two is 0.73 and both of them turn positive in 2001, indicating their co-movement. This means that employment in start-ups and outside the enterprise sector that are not captured by the job flow indicators also play some role in explaining overall employment dynamics, but do not cancel out the observed job flows in reporting continuing enterprises. Table 6.15: Correlation Coefficients of Job Flows with GDP and Employment Changes (1994-2004) GDP change GDP change Employment Employment change (t-1) (t) change (t-1) (t) Job creation 43.7% -0.5% 27.7% 44.4% Job destruction 2.2% -9.0% 73.5% 79.8% Net job creat. 20.7% -6.3% 61.7% 73.3% Job turnover 32.6% 10.8% -68.9% -63.2% Source: Croatian Bureau of Statistics; authors calculations based on FINA database. Resistance of job destruction rates to fluctuations of economic activity are typical for heavily regulated labor markets where costs of employment adjustment outweight any possible gains that arise from optimization of employment. In such an economy any easing of restrictions would facilitate immediate adjustment of the employment level through increasing job destruction. However, job destruction rates in Croatia continued to fall following the reform of the EPL in 2003. Even looking at particular categories where employment adjustment was more likely to occur, such as enterprises falling under the increased margin for small enterprise according to new regulations (enterprises with 10-20 employees) only a minor increase of job destruction was visible in 2004. This means that either EPL was not relaxed enough to spur further adjustments or there still remain other types of administrative barriers and state interventions in the economy, limiting labor market flexibility. Decomposed data on job flows (Appendix C4) raise a number of important points. First of all, there is a stark difference between job creation rates in de novo74 private enterprises and all other enterprises, including mixed ownership, majority private as well as fully privatized enterprises. Job creation rates in the former exceed rates in latter by a high multiple. At the same time, job destruction rates did not differ that much between different types of enterprises, although job destruction rates in de novo private enterprises were somewhat higher than the average throughout the entire period. Those enterprises on average accounted for more than 80 percent of the total job creation during the period 74 "De novo" enterprises indicate newly founded enterprises that are privately owned since the establishment. 209 under observation, while their share in total job destruction was a bit higher than one-third. Such dynamics of job flows increased their share in total employment by almost four times during the period under observation as they reached almost half of employment in reporting enterprises by 2004. The endogenous growth of these enterprises provided the only net additions to the total employment. It is possible that EPL affected job dynamics more as a barrier to the establishment of these new firms than as a barrier to growth of the existing companies. Newly established enterprises that generated employment were mostly small. Enterprises with less than 50 employees more or less constantly added new jobs, while those with more than 50 employees until recently persistently shed labor. However, it is decline in the number of jobs destroyed and increase in the number of jobs created in large enterprises that recently ticked the balance in favor of job creation. Although employment share of large enterprises considerably dropped from the initial nine-tenths, it still account for about two-thirds of the reporting enterprises and it appears to be stabilized. Consolidation of employment in large enterprises also coincided with entry of major international retail corporations into the Croatian market. Dynamic enterprises tend to be located in Zagreb and especially in the nearby region, such as the Zagreb County, while in all other regions job destruction was similar to job creation or even prevailed during the observed period. The only exception is Vukovarsko-Srijemska County where return of the refugees in the aftermath of the conflict gave basis for employment growth. With respect to breakdown according to economic activities, more dynamic enterprises tended to work in trade, construction and other business activities. Net job creation rates were fairly small and excess job reallocation remained important during the whole period. Deeper insight into the forces driving excess job reallocation can be inferred from data on job flows and excess job reallocation, decomposed according to various enterprise characteristics, such as size, ownership, sector of economic activity and regional affiliation. Excess job reallocation can be decomposed in two components: the portion of such flows arising within a certain group of enterprises (such as enterprises operating within a certain economic activity) and flows arising between groups of enterprises. The first component is measured at the level of individual economic activities as the sum across all sectors of excess job reallocation in each sector. The second component is measured by summing across sectors the deviation of the absolute value of the growth rate for that sector from the absolute value of the growth rate at the level of individual activities (Faggio and Konings, 1998). Almost one third of excess job reallocation throughout the entire period took place between enterprises belonging to different ownership types, mostly between de novo private enterprises and all other enterprises. Therefore, difference in ownership was the most important factor driving reallocation of jobs during the transition process in Croatia. However, attrition of jobs in the government sector was nevertheless slow during the whole observed period, although the process kept the momentum until the end of the observed period. The decline of total employment was spread over a long period of time due to slow employment adjustment in state owned and privatized enterprises and the time needed for the new private sector to absorb their employees. Other enterprise attributes are not as important for explaining the dynamics of job reallocation. Most excess job reallocation occurred within economic activities (defined at NACE-2 level), with less than 20 percent of total excess reallocation going on between different sectors. Faggio and Konings (1998) report similar results for Romania, while inter-sectoral reallocation accounted for only 10 percent of total excess job reallocation in Bulgaria and as much as half of excess reallocation in Estonia. Although regional imbalances in Croatia are sometimes alleged to be particularly large, according to the regional decomposition of excess job reallocation, this kind of mobility appears low, with an average of about 3 percent of jobs reallocating between the regions. However, this indicator has to be looked at cautiously since employment is registered by enterprise headquarters and not by actual location where the work takes place, which may bias the indicator either way. Also, even modest migration of jobs between the regions can build significant imbalances, if their direction is persistent. 210 Finally, net job creation rates in small enterprises during the early years seem dramatic, but as the share of firms with less than 20 employees approached a quarter of total employment by the end of the period, their growth stalled and the reallocation of jobs between firms of different sizes stopped being important. The thresholds for firm size classes were chosen somewhat arbitrarily, but the choice of six relevant groups would probably capture any more significant movements of jobs between enterprises of different size. Table 6.16: Decomposition of Excess Job Reallocation Arising from Shifts within and between Different Regions, Ownership Types, Size Classes and Economic Activities 1994 1995 1996 1997 1998 1999 2000 2001 2003 2003 2004 Average Regions Within 98.5 99.7 95.7 98.5 94.0 99.8 92.6 94.0 99.3 97.5 97.1 97.0 Between 1.5 0.3 4.3 1.5 6.0 0.2 7.4 6.0 0.7 2.5 2.9 3.0 CV (turnover) 0.22 0.23 0.20 0.25 0.21 0.21 0.20 0.17 0.23 0.17 0.17 0.2 Ownership type Within 51.5 62.1 61.3 63.1 68.8 85.7 80.9 71.1 73.8 78.8 84.6 61.5 Between 48.5 37.9 38.7 36.9 31.2 14.3 19.1 28.9 26.2 21.2 15.4 38.5 CV (turnover) 0.74 0.65 0.58 0.43 0.49 0.47 0.55 0.50 0.47 0.49 0.56 0.50 Size class Within 79.5 82.8 83.4 83.6 82.8 92.8 94.0 92.3 91.6 97.5 100.0 80.8 Between 20.5 17.2 16.6 16.4 17.2 7.2 6.0 7.7 8.4 2.5 - 19.2 CV (turnover) 0.61 0.55 0.51 0.41 0.48 0.50 0.45 0.39 0.42 0.37 0.44 0.47 Economic activities Within sectors 82.6 79.3 79.2 86.6 77.8 81.8 78.7 79.7 89.0 87.7 Between sectors 17.4 20.7 20.8 13.4 22.2 18.2 21.3 20.3 11.0 12.3 CV (turnover) 0.64 1.32 0.63 1.16 1.44 0.76 2.04 0.65 0.93 1.0 Note: CV ­ coefficient of variation of disaggregated job turnover. Source: authors calculations based on FINA database. The aggregate magnitude of job flows that can be considered normal for most countries masks a high degree of segmentation in the Croatian labor market. While state owned enterprises as well as privatized enterprises, still accounting for a major portion of overall employment, remain stagnant throughout the period, the new private sector was exceptionally dynamic. Part of the difference could probably be attributed to the fact that small enterprises (with less than 10 employees) were treated favorably by the Employment Law, with less cumbersome legal firing procedures, which facilitated job reallocation into de novo enterprises. De novo enterprises also did not have to bear disproportionate costs of separation for workers with long tenures that burdened state owned and privatized enterprises. However, due to the sheer volume of job flows, EPL does not seem to seriously affect employment adjustment in de novo enterprises. Stringent regulation perhaps even additionally strained the adjustment in de novo enterprises as other enterprises did not seem to perform substantial adjustment. The results reported confirming a considerable degree of overall flexibility as well as exceptional dynamics among some enterprises simultaneous with a high EPL level are not unique to Croatia. There are other studies of transition countries that are equally inconclusive on the empirical validity of the theoretical relationship between EPL and job flows in transition countries. Davis and Haltiwanger (1999) conclude that there is really not that much variation among countries as roughly one job in ten is created and destroyed every year in most advanced economies and transition countries, regardless of labor market regulation. These authors base their observations in part on the weak comparability of job flow data across countries due to sample and firm differences. According to Davis and Haltiwanger (1999), careful disaggregated analysis, performed to some extent in this study of the Croatian labor market, is essential in order to identify the effects of labor market institutions and policies on labor market flows. 211 Similar to the finding of a fairly high level of job creation in Croatia reported in this paper, Haltiwanger and Vodopivec (2003), using matched employer-employee data, find evidence for Slovenia, another country with a high EPL level, of job turnover rates approaching 20 percent during the 1997-1999 period. These turnover rates, which are even a bit higher than the rates found in Croatia but in line with figures observed in most dynamic transition countries, challenge their older work. There are several possible explanations for empirical findings contrasting theoretical predictions on EPL effects, going beyond problems in measurement of job flows. The first reason is that the EPL index, as the most comprehensive indicator available, is far from perfect in capturing all the regulatory restrictions on hiring and firing. Even Estonia, considered to be the prime example of flexible labor market, has a higher value of its EPL index than Czech Republic or Poland, whose labor markets are regarded as more rigid. Macculloch and Di Tella (2002) for that reason in their study of labor market rigidities rely on an alternative source--survey based data on hiring and firing restrictions from the World Competitiveness Report. Monastiriotis (2003) and Bertola, Boeri and Cazes (2000) also emphasize insufficiencies in the measurement of labor market regulation. Bertola, Boeri and Cazes (2000) take the position that EPL index, as well as other available protection indicators, are based on unsatisfactory information and capture neither the increasing complexity of legal provisions nor their interactions, such as the rise of temporary work. The validity of the EPL index became especially problematic after the reform process gathered pace during the 1980's and 1990's, since this diminished correlations between different components of EPL. The authors conclude that there is no simple means of calculating the relevant indicators and rankings and that there is a need to pursuit further research on that subject. Also, it seems that enforcement of regulation matters very much since available evidence points to differences in the frequency of labor disputes and national practices in interpretation and resolving cases. In light of the ambiguous empirical evidence and the measurement issues, the authors believe that policy recommendations should be formulated with caution. Furthermore, there are many labor market institutions at work beyond the EPL. Cazes and Nesporova (2003) point out that labor market adjustment in transition countries was significantly affected by macroeconomic and structural reforms. Also the role of other labor market institutions, such as passive and active labor market policies, the power of trade unions or the tax burden on labor have to be considered. Some of these factors were already discussed and the failure to adequately account for their role may give rise to flawed conjectures on the impact of EPL. Haltiwanger and Vodopivec (2003) offer an alternative view on the relevant labor market institutions influencing job flows. Their study does not assess dynamics of the Slovenian labor market in a comparative perspective, but detailed analysis of the flows gives some clues as to the forces behind them. Findings concerning the determinants of flows at the firm level show that enterprises with higher wage dispersion exhibit lower job turnover rates. Therefore, although the study does not extend the conclusion too far, high turnover rates may be the consequence of compressed wage structures throughout the economy. This possibility is in line with the theoretical model of Bertola and Rogerson (1996) that emphasizes the possible role of wage setting institutions on job flows as higher job flows offset for lack of wage flexibility. The route taken in this paper suggests that EPL may affect some workers and enterprises, but that their impact on aggregate flexibility becomes less important as employers learn to cope with the burden of regulation. However, the impact of EPL is not negligible, as it moves turnover from enterprises of one type to enterprises of other type. Therefore, it affects the distribution of turnover across enterprises more than turnover volume. 212 Table 6.17: Distribution of Tenures in Transition Countries Under 1 1 and 2 and 5 and 10 and 20 Average Median year under 2 under 5 under under years tenure tenure years years 10 20 or over (years) (years) years years Nesporova and Cazes (2001) - data for 1999 Czech R. 14.6 18.4* 15.3** 26.2 12.3 13.2 8.2 Estonia 18.4 6.7 31.1 23.9 10.8 9.1 6.9 Hungary 12.6 11.3 20.0 25.3 17.9 13.0 8.8 Lithuania 12.8 9.2 29.0 24.8 14.5 9.6 7.6 Poland 10.5 10.4 14.0 20.8 22.3 22.0 11.9 Slovenia 12.0 5.1 18.2 16.5 23.6 24.6 12.0 Rutkowski (2003) Bulgaria (2001) 14.0 9.5 25.2 20.8 19.8 10.8 8.1 5.5 Czech R. (1995) 19.2 36.6 12.0 14.8 17.4 9.0 2.0 Lithuania (2001) 15.4 8.9 21.6 25.4 16.8 11.9 8.3 5.0 Poland (1999) 14.5 11.7 19.0 17.7 20.3 16.7 9.6 6.2 Croatia (2001) 9.7 5.1 17.2 21.3 20.7 26.0 12.2 8.0 own calculation Croatia (2002) 13.4 6.3 16.7 21.1 19.2 23.3 11.8 7.9 Note: * refers to 1-3 years; ** refers to 3-5 years. Source: Rutkowski (2003), Nesporova and Cazes (2001) and own calculations based on Labor Force Survey, 1st half of 2002. The second line of reasoning Rutkowski (2003) used in order to confirm low level of flexibility in Croatian labor market was to compare data on tenures. As was previously showed, some studies using worker flows and tenure data were more successful in finding the impact of EPL on labor market dynamics. Data on average tenures and worker flows may also give more precise information on types of workers that managed to keep their jobs due to EPL and workers that do not seem so well protected by EPL. Indeed, average and median tenures in Croatia, according to the labor force survey data, are together with Slovenian and Polish figures (later based on Cazes and Nesporova, 2003) among the highest in the group of transition countries. Reliance on tenure and worker flow data rather than job flows also has some deficiencies. First of all, there are different sources of worker flow data yielding different results, such as enterprise surveys and labor force surveys, conducted with different frequencies. Second, participation rates of different age groups may affect tenure structure, with low participation rate of young persons increasing the share of persons with longer tenures and hence raising the average tenure75. Furthermore, tenure structures do not account for the dynamics of overall employment, which may also interact with age-specific participation rates. Employment reductions usually affect workers with shorter tenures as the seniority principle often prevails in case of redundancies, which increases average tenure throughout the economy. At the same time, larger number of newly employed persons, facilitated by employment growth, reduces average tenure. A final and more substantial problem with this approach is that EPL may have some effect on job stability only for employees with tenures exceeding a certain threshold, while average tenure may disregard churning that takes place amongst the most dynamic group of employees with the shortest tenures. Notwithstanding the problems elaborated so far, how do worker flows in Croatia compare to other countries? The hiring rate in Croatia, calculated on an annual basis from the LFS database, was 13.4 percent in 2002, while the worker turnover rate was 29.7 percent, which falls in the middle of the range observed in selected transition countries, close to the Estonian and Slovenian figures76. Higher rates compared to results Rutkowski (2003) reported probably stem from the employment growth 75 Low participation rate of young persons may itself be a consequence of labor market rigidities. 76 The hiring rate calculated on an annual basis falls short of the separation rate despite employment growth because it is underestimated to a greater extent as some workers entering employment churn between jobs within the same year (all worker flows including multiple switches within a year under observation are not captured in this framework). The bias is significantly reduced in the quarterly figures. 213 taking place in 200277, on the one hand, and from the growing proportion of temporary workers among the newly employed, on the other hand78. This illustrates the importance of different effects other than EPL on worker flows, such as cyclical dynamics of employment and employers learning about new ways to facilitate flexible arrangements. Moreover, if worker flows are constructed on a quarterly basis, the resulting figures are almost twice as high, revealing the short average duration of many recent temporary contracts. Temporary workers accounted for 12.5 percent of total employment in 2002 and only 15 percent of those had contract duration above 1 year (Crnkovi-Pozai, 2004)79. With significant churning taking place on the low end of the tenure structure, the comparatively high average tenure in Croatia resulted from the gap in the middle of the tenure structure and many workers with long tenures. Table 6.18: Hiring Rates, Separation Rates and Worker Turnover in Transition Countries Hiring rates Separation rates Worker turnover rate Bulgaria (1999 -ES) 27.6 39.9 67.5 Estonia (1998-LFS) 16.0 19.0 35.0 Czech Republic (1998-LFS) 10.5 11.8 22.3 Poland (1998-ES) 24.6 22.8 47.4 Poland (1998-LFS) 21.2 17.0 38.2 Slovenia (2001-ES) 15.6 14.5 30.1 Croatia (2002-LFS-annual) 13.4 16.3 29.7 Croatia (2002-LFS-quarterly) 26.3 26.2 52.5 Source: Cazes and Nesporova (2001) and own calculations based on Labor Force Survey 2002. Figure 6.10: EPL Index and Worker Turnover (Cazes-Nesporova dataset + own calculations) 80 worker turnover 70 Bulgaria 60 50 40 Poland Estonia Slovenia 30 Croatia Czech Republic 20 turnover = -0.98 EPL indeks + 39.58 10 R2 = 0.0017 0 1.5 2 2.5 3 3.5 4 EPL index Source: tables 2 and 8. 77 Employment rate of persons older than 15 years increased from 41.8 percent in 2001 to 43.3 percent in 2002. 78 According to the data from Croatian employment service, in 2002 between 80 percent and 90 percent of the newly employed had temporary contracts, up from 50 percent to 60 percent in 1995. 79 Contact expiration does not necessarily lead to separation, but temporary contracts can be renewed only a limited number of times as total cumulative duration of fixed-term contracts (accounting for almost four-fifths of all temporary contracts) could not exceed 2 years (this period was increased to 3 years after the 2003 changes of the Labor Code), while there were similar limitations to seasonal and other temporary contracts as well. 214 Disaggregated worker flow data according to ownership again reveal high turnover in the private sector, both on the hiring and separation sides, while turnover among the government employees and employees in state owned enterprises is less than half of that figure. Average tenures show the opposite picture, with average tenure in government and state owned sector of about 14 years, which is almost double as much as in the private sector. Unfortunately, the structure of the LFS does not allow us to distinguish between privatized enterprises and de novo private enterprises, which exhibit much higher job turnover rates. Therefore, the average for the largest employment category conceals some of the differences that job flow data indicate. It is also interesting to note that self-employment is the major employment category with the highest average tenure. This is obviously a heterogeneous category, but the average tenure of persons belonging to the self-employed category would suggest that they, on average, do not comprise a flexible group compensating for the lack of flexibility among employees in other sectors, as is sometimes thought about self-employed. It to some extent also reflects the impact of individuals engaged in agriculture, who often do not retire after the mandatory age. If the self- employed are excluded from the calculation, the average tenure for all other categories of employment (comprising almost 83 percent of total employment) is 8.9 years, which is similar to transition countries with more flexible labor markets. Table 6.19: Hirings and Separation Rates, According to Sectors of Ownership (2002) Government Unpaid Self- and state owned Private In family employed sector sector privatization workers On contract Hiring rate 6.9 6.5 22.1 6.4 9.1 39.5 Separation rate 10.4 11.7 21.3 19.4 23.2 29.5 of which: job to job 4.8 2.9 9.3 2.2 3.2 7.4 Total worker 17.3 18.1 43.4 25.8 32.3 68.9 turnover rate Employment share 17.2 35.1 37.9 2.5 5.0 2.4 Average tenure 14.9 14.3 7.8 16.3 16.8 4.2 Source: own calculations based on Labor Force Survey 2002. Worker turnover rates steadily decrease with age, reflecting both decreasing hiring and separation rates. Only the oldest category represents an exception with respect to the separation side, which shows an impact of retirement and persons leaving the labor force. It is particularly interesting to note that although young persons represent a category with exceptionally high unemployment, their turnover rates are tremendous. Although it seems that young persons find a job with ease, it is very hard for them to keep the job or transfer directly into another job as they face considerable job insecurity. Young persons, therefore, find it difficult to cross over the one year tenure threshold, which leads to the gap in the middle of the tenure structure. Older persons participating in the employment face a substantially lower possibility of separation than young persons, but chances of finding a new job decrease even more with age. Persons in each reported age-group on average have about five years longer tenure, except for the last category, reflecting again the impact of self-employed agriculture workers. Table 6.20: Hirings and Separation Rates, According to Age Groups 15-24 25-34 35-44 45-54 55- Hirings 44.9 18.9 9.0 5.0 2.2 Total separations 28.6 17.8 11.4 10.4 22.8 of which: job-to-job 9.7 9.4 5.2 2.9 0.7 Total worker turnover rate 73.5 36.7 20.4 15.3 25.0 Employment share 9.0 23.2 29.9 26.8 11.0 Average tenure 1.8 5.2 10.8 16.7 26.2 Source: own calculations based on Labor Force Survey 2002. Since one possible reaction of employers to high level of EPL (or hiring and firing restrictions) and high non-wage labor costs is to seek informality, it is interesting to ask to what extent employers use informal employment in order to facilitate flexibility. It is notoriously hard to estimate informal employment, although some attempts have been made. Crnkovi-Pozai (1997), using the Labor force 215 survey, estimates informal employment to be about a quarter of total employment in Croatia in 1995. The largest part of informal employment, close to 40 percent, according to actual status in employment of those persons, consisted of unpaid family workers. Persons holding second (or multiple) jobs constitute another important form of informal employment, accounting for about a quarter of the total. As noted above, there are different sources of employment in the informal sector and different degrees of informality as well. The "core" of informal employment is to a large extent fueled by persons whose formal status differs from employment, such as students, unemployed persons or retired persons engaged in economic activities. During the early years of transition, many people officially left employment and went into inactivity, mostly by means of early retirement. Crnkovi-Pozai (1998) estimates that 369 thousand people left activity over the 1990-1996 period80. Many of those persons informally reentered employment as a socially insured group in an attempt to top-up their income. This is not so remarkable since early retirees on average aged 55. Retired persons constituted 6.5 percent of total employment according to ILO criteria and were one of the largest groups in the informal economy. Persons registered as unemployed constitute the second important source of informal employment, but their number has been falling recently. Share of these persons in total employment decreased from close to 4 percent in mid 1997 to a bit over 2 percent by the end of 2000. Several possible factors can account for the declining tendency of informal employment, despite worsening labor market situation until 2001. The first is that government institutions increased their efficiency and law enforcement as the immediate threat of war disappeared after 1995. Ott (2002) in a survey of unofficial economy in Croatia reports a number of different policy areas where significant progress was accomplished. The most important were tax reform, including the introduction of VAT in 1998, and drastic reduction of state arrears in 2000 that changed the image of the government with respect to payments. These changes were paralleled with the fall of attitudes towards opportunistic behavior also documented in Ott (2002) as the number of persons who think that tax evasion and bribe can never be justified has doubled between 1995 and 1999, indicating a rise in tax morality. Simultaneous change in the structure of spending to consumer durables (cars, flats) mainly bought with loans, stabilization of large retail systems and the introduction of foreign firms into the Croatian market also implied fewer opportunities for the work in unofficial economy and reinforced the effect of reforms and change in the attitudes. In addition to these factors, labor tax wedge was reduced from about 48 percent of total labor costs in 1995 to about 41 percent in 2001 (Rutkowski, 2003), which reduced the potential benefits of concealing economic activities. This reduction may not seem substantial, but it means that the amount of taxes and contributions paid on net wage were reduced by about a quarter - from 92 percent to 69 percent of net wage. To the extent that firing restrictions acted as a barrier for formalization of employee status, the growing share of temporary workers with short-term contract demonstrates how employers learned to exploit opportunities to enhance flexibility without violating the law. Finally and more recently, in early 2002 Croatian employment service introduced a number of so called "activation" measures designed to ensure that unemployed people are looking actively for work and to help them to do so. These measures facilitated identification and deletion from the unemployment register of those persons who did not fulfill new obligations due to engagement in informal activities. Comparison of hiring rates in the formal and informal sector is one simple way to test whether informal employment was used to a large extent to promote flexibility. According to the "narrow" definition of informal employment, which excludes second job holders, but includes unpaid family workers as well as the formally unemployed and inactive (this category further collapses into retired persons, pupils and students and housewives) informal employment in 2002 declined to 12 percent of total employment (from over 19 percent of the total in 1995, according to Crnkovi-Pozai, 1997). However, the share of hiring in informal sector in total hiring was 16.3 percent, which was not substantially higher than share of the informal sector in employment. Certain categories amongst 80 This figure is based on a Labor force survey conducted in 1996 and is therefore subject to a recall error. 216 informally employed exhibit higher job turnover, such as the unemployed, inactive and especially pupils and students, but hiring rates of those groups are again not substantially above those observed for workers engaged on formal job-specific contracts. Table 6.21: Employed Persons According to their Formal Status Informal employment Self- Unpaid Pupils Emplo- Self- On Unem- In- House- employed family and Retired yees employed contract ployed active wives (agricul.) workers students Employ. 73.8 8.6 4.7 0.8 1.6 0.5 4.2 2.8 1.5 1.4 share Average 11.3 8.4 23.1 4.5 17.0 2.4 22.2 4.0 2.7 25.1 tenure Hirings 13.4 10.1 2.0 38.5 5.9 61.7 3.4 37.1 42.7 3.8 rate Source: own calculations based on Labor Force Survey 2002 Although informal employment in Croatia comprises of diverse groups and the narrow definition employed above focuses only on one portion of informal employment, notwithstanding individuals omitted by the labor force survey, above results suggest that formal and informal employment do not differ significantly along the dimension of flexibility. Therefore, the description of the informal employment as an adjustment buffer substituting for the lack of flexibility in the formal sector, according to the "dual labor market" hypothesis, is not an appropriate analytical tool to assess the relationship between formal and informal employment in Croatia. The dynamics of informal employment is rather more similar to flows in and out of the formal employment and therefore more likely to support the modeling approach of Boeri and Garibaldi (2002), according to which formal and informal employments enter the matching function in a complementary fashion. Moreover, Boeri and Garibaldi (2002) view informal employment as a joint decision by worker and firm. Evidence on job turnover in informal sector confirms that it is not only an inclination of employers to foster flexibility but also a supply side of the labor market that is relevant for the decision to seek informality. An attempt to top-up income arising from certain formal status, such as pensions or unemployment benefits, which would be lost in case of formal employment, with income from informal activities, acts as an important motivation for many persons engaged in informal activities. This means that if policymakers are determined to tackle the problem of informal employment in Croatia, they should also seek to monitor different groups of benefit recipients for participation in informal activities. The relationship between job and worker flows is another point of interest. The link between these two is of special interest because job turnover in market economies is an important determinant of the overall pace of worker reallocation, usually accounting for roughly 30 to 50 percent of worker turnover (OECD, 1997). In transition countries, an even larger portion of hiring and separation is driven by job destruction and job creation. This means that a major factor underlying worker mobility in transition countries is the changing allocation of jobs across businesses, as opposed to workers reallocating themselves for a given allocation of jobs across businesses (Haltiwanger, Lehmann and Terrell, 2003). Haltiwanger and Vodopivec (2000) report that in Estonia by 1993 job turnover comprised more than two-thirds of worker turnover (worker turnover rate exceeded 35 percent at that period of time). Haltiwanger and Vodopivec (2003) find remarkably similar results for Slovenia during the late 1990's as job turnover accounted for about two-thirds of worker turnover. According to reported findings, job flows in Croatia comprised about a third of worker turnover (calculated on a quarterly basis, better approximating true extent of worker turnover), which is in line with results from advanced market economies, but there are comparatively more workers churning between jobs in Croatia than in other transition countries. With respect to this finding, Croatia is more similar to a mature market economy, as there appears to be significant churning of workers between jobs. However, there is a possibility that regulation provides an important foundation for such a pattern of job and worker flows. The limited maximum duration of temporary contracts drastically 217 increases churning at the short end of the tenure structure, notably among young workers entering employment in the new private sector. However, it has to be kept in mind that job and worker turnover data were constructed using different datasets that can bias the results81. CONCLUSIONS Labor market policies carried out in Croatia did not markedly differ from those pursued in other advanced transition countries, although it would probably be possible to fine-tune those policies as the experience with their implementation increases. The most important reform implemented during the past few years concerns making the EPL less restrictive and increasing the reliance on passive labor market policies, which is in line with ILO and EU recommendations. The duration of benefits was extended, which seemed justified due to the low coverage of unemployed persons. Less restrictive regulation should, in principle, increase labor market flexibility and increase coverage of the unemployed persons. Both job and worker flows, which approximate the extent of external flexibility in labor input that is most likely to be affected by restrictive dismissal regulation, confirm a fair degree of dynamics in the Croatian labor market. This is not unusual since some other countries considered to have low degree of labor market flexibility (an example of Slovenia was emphasized) demonstrate as well that the link between regulation and flexibility is sometimes very weak. Data on average tenure in Croatia, on the other hand, indeed indicates somewhat higher degree of job stability that is amongst the highest in transition countries due to large share of persons with exceptionally long tenures. However, if self- employed persons (many of those are individual farmers with long experience, which are less present in other transition countries) are excluded from the aggregate, average tenure seems substantially lower, in line with most flexible transition economies. Therefore, it seems that data on job flows and worker flows do not single out Croatia from other transition countries, which goes against some previous evidence (e.g. Rutkowski, 2003). Although overall degree of flexibility in Croatia does not seem especially low in international perspective, job flows are not responsive to fluctuations in economic activity, which is typical for heavily regulated economies. Moreover, labor market flexibility is confined to employees in new private sector, mostly in small and medium sized enterprises, with high chance for fluctuation between employment and unemployment. While employment adjustment in state owned and privatized enterprises on average takes a long time, it seems that new private sector performs most of the adjustment. Young workers that recently entered employment are particularly prone to enter the cycle of excessively volatile employment. With still significant employment in state owned enterprises and differing impact of EPL across sectors, it seems that labor market reform did not immediately increase overall flexibility of the labor market. The EPL reform seems most likely to affect the number new establishments, which is still to early to judge on the basis of job flow data. Also, reform in the other areas, such as removal of administrative barriers to investments and facilitating restructuring of state owned enterprises, may prove to be more important for flexibility in the short run. Finally, there are several reasons to remain cautious with respect to reported findings and expected outcomes of reform. First, since reform is the most pronounced in the area of regulations governing temporary contracts, reform may stress the duality of the labor market even more. As the whole area of research is still sketchy and the available datasets still hinder cross-country comparisons, labor economists weren't very successful up to now in measuring the impact of regulation on labor market flexibility in transition countries. This is not so odd because similar problems are still present even in studies of advanced economies. As different studies reveal contradictory results, it seems that at the moment only a few vague policy conclusions can be drawn with confidence from the recent research of the labor market flexibility in transition countries. 81 There is also a possibility that significant bias is introduced into the job flow data by drawing the data only from continuing firms between the subsequent reporting periods. 218 References Aghion, P. and Howitt, P. (1994) "Growth and Unemployment", Review of Economic Studies, Vol. 61, pp. 477-494. Arandarenko M. (2004) "International Advice and Labor Market Institutions in South-East Europe", Global Social Policy, Vol. 4; Br. 1, str. 27 - 53. Boeri, T., Burda M. and J. Kölló (1998) "Mediating the Transition: Labor Markets in Central and Eastern Europe", Economic Policy Initiative Forum Report No. 4, CEPR. Boeri, T. and J. F. Jimeno (2003) "The Effects of Employment Protection: Learning from Variable Enforcement," CEPR Discussion Papers 3926, C.E.P.R. Discussion Papers. Cazes S. and A. Nesporova (2001) "Towards excessive job insecurity in transition economies?", ILO Employment Paper 23, International Labor Organization, Geneva Cazes, S. and A. Nesporova, (2003), Labor Markets in Transition: Balancing flexibility and security in Central and Eastern Europe, Geneva: International Labor Office, ILO Csajbók, A and Á. Csermely, eds. (2002): "Adopting the euro in Hungary: expected costs, benefits and timing", MNB Occasional Papers, no 24. Davis, S. and J. Haltiwanger (1999) "Gross job flows", in: Ashenfelter, O. and Card, D. (Eds.) "Handbook of Labor Economics", Vol. 3, Princeton University Press, Princeton, pp. 2711­2805. Dar A. and Z. Tzannatos (1999) "Active Labor Market Programs: A Review of the Evidence from Evaluations", January 1999 Dorenbos R. and P. van Winden (2001) "Evaluation of program of public works in Croatia: Final report", mimeo DZS (2004) "Aktivno stanovnistvo u Republici Hrvatskoj", Statisticko izvjese br. 1244, Drzavni zavod za statistiku, Zagreb Ederveen S. and L. Thissen (2004) "Can labor market institutions explain unemployment rates in new EU member states?", CPB Netherlands Bureau for Economic Policy Analysis, CPB Document No 59, April 2004 Eurostat (2004) "European Labor Force Survey: Principal Results 2003", Statistics in focus, Population and Social Conditions, 14/2004 Faggio, G. and Konings J. (2003) "Job creation, job destruction and employment growth in transition countries in the 90s", Economic Systems, vol. 27, no. 2, pp. 129-154 (June 2003) Freeman, R.B. (2000) "Single Peaked vs.Diversified Capitalism:The Relation between Economic Institutions and Outcome",NBER Working Paper No.7556, National Bureau of Economic Research. Garibaldi P. i P. Mauro (2002) "Anatomy of Employment Growth", Economic Policy No. 34, Spring 2002 Haltiwanger, J. C. and M. Vodopivec (2000) "Gross Worker and Job Flows in a Transition Economy: An Analysis of Estonia", available online: http://www.bsos.umd.edu/econ/haltiwanger/estonia2.pdf Haltiwanger, J., Lehmann, H. and K. Terrell (2003) "Symposium on Job Creation and Job Destruction in Transition Countries: Symposium Introduction", The Economics of Transition, Vol. 11, No. 2, pp. 205­219 Haltiwanger, J. C. and M. Vodopivec (2003) "Worker flows, job flows and firm wage p: An analysis of Slovenia", The Economics of Transition, Vol. 11, No. 2, pp. 253­290 219 Jurajda, S. and Terrell K. (2001) "What Drives the Speed of Job Reallocation during Episodes of Massive Adjustment?", CERGE-EI Working Paper No. 170 and William Davidson Institute Working Paper No. 432. Elmeskov J. (1998) "The unemployment problem in Europe: Lessons from implementing the OECD Jobs Strategy", Cahiers Papers, Vol. 3, br. 1. Jackman, R. (2002) "Determinants of unemployment in Western Europe and possible policy responses", in United Nations Economic Commission for Europe (ed.) Economic Survey of Europe, No. 2. Geneva: United Nations. Lowther, J. (2003) "Fleksibilnost radne snage i uloga hrvatskih socijalnih partnera u njezinu poveanju", Financijska teorija i praksa, Vol. 27, br. 4, str. 457-479. Nickell, S., Nunziata L. i W. Ochel (2005) "Unemployment in the OECD Since the 1960s. What Do We Know?", The Economic Journal, Vol. 115, br. 1., 1-17. Nesporova, A. (2002) "Unemployment in the Transition Economies", in United Nations Economic Commission for Europe (ed.) Economic Survey of Europe, No. 2. Geneva: United Nations. Matkovi, T. i I. Biondi, 2003. "Reforma zakona o radu i promjena indeksa zakonske zastite zaposlenja", Financijska teorija i praksa, Vol. 27, br. 4, pp. 515-528. Riboud M., Sánchez-Páramo C. i C. Silva-Jauregui (2002) "Does Eurosclerosis Mater? Institutional Reform and Labor Market Performance in Central and Eastern European Countries", in Funck B. and L. Pizzati eds.: Labor, Employment, and Social Policies in the EU Enlargement Process: Changing Perspectives and Policy Options, World Bank, Washington DC Rutkowski, J. (2003) "Analiza i prijedlozi poboljsanja trzista rada u Hrvatskoj", Financijska teorija i praksa, Vol. 27, br. 4, str. 495-513. OECD (1994) "The OECD Jobs Study: Facts, Analysis, Strategies", Paris: OECD. OECD (1999) "Employment Protection and Labor Market Performance", Employment Outlook, str. 49.-132. OECD (2004) "Taxing Wages 2002/2003 (Selected Tables and Graphs)", availalbe online: http://www.oecd.org/dataoecd/34/25/30398959.pdf Peters M., Dorenbos R., van der Ende M., Versantvoort M. and Arents M. (2004) "Benefit Systems and their Interaction with Active Labor Market Policies", Final report, European Commission, Directorate-General for Employment and Social Affairs, Unit EMPL/A/2, Manuscript completed February 2004 Scarpetta, S., Hemmings, Ph., Tressel, Th. and J. Woo (2002) "The Role of Policy and Institutions for Productivity and Firm Dynamics: Evidence from Micro and Industry Data", OECD Working Paper 329. Pizzati, L. i B. Funck (2002) "Labor, employment, and social policies in the EU enlargement process: Changing perspectives and policy options", World Bank, Washington DC Vujci, B. (1997) "Structural Changes in Employment: Where we are, How we got there and where are we going to?", Croatian National Bank, CNB Working paper No. 4, September 1998. WIIW (2004) "Countries in transition 2004", WIIW handboook of statistics, Vienna Young, D. (2003) "Employment protection legislation: its economic impact and the case for reform", EU Economic Papers 186. 220 APPENFICES APPENDIX A6: DEFINITIONS Job Flows Job creation equals employment gains summed over all business units that expanded during the year under observation. Job destruction is equal to employment losses summed over all business units that contracted during the year under observation. Job flows are usually expressed as a proportion of average employment at the beginning and at the end of the period under observation. Capturing job flows taking place within firms or establishments would be highly desirable, but very few studies manage to do this and it is virtually impossible to directly measure these flows using firm-level data. In principle, capturing job flows occurring in firms starting and closing during the period under observation would also be highly desirable, but since it is hard to distinguish between these events and non-reporting, such flows are usually omitted, as was done is this study. Omitting both within-firm flows and flows taking place in starting/closing firms biases job flow data downward. Net employment change equals the difference between job creation and job destruction. Job turnover equals the sum of the absolute value of all business units' employment gains and losses, that is the sum of job creation and job destruction. Excess job reallocation is equal to the difference between job turnover and the absolute value of net employment change. It represents the part of job turnover that is above the amount required to accommodate net employment change. Worker Flows The hiring rate is calculated as a sum of aggregate flows from unemployment to employment, from inactivity to employment and from one employment to another during the year under observation, divided by total employment at the beginning of the period. The separation rate is the sum of aggregate flows from employment to unemployment, from employment to inactivity and from one employment to another, divided by total employment at the beginning of the period (in both cases the average of the period to which Labor force survey refers was used because it is the most reliable estimate). Worker turnover is the sum of hiring and separation rates. Since rough data from the Labor force survey, which records only the last transition from one labor market state to another, was used to calculate worker flows, it is possible that the calculated rates understate the true turnover due to missed multiple transitions. While worker turnover refers to the movement of persons, job turnover encompasses only those movements that involve movement of jobs. Therefore, worker turnover encompasses labor turnover, but these two overlap only if separation is not followed by hiring, or vice versa. Also the coverage of the LFS is much wider than that of the FINA database, as LFS estimates employment to be about twice as much as enterprises report. The difference is due to sectors that do not report to FINA, such as government (public services, defense, the health and education sector) and unincorporated business (trades and crafts). Sources of definitions: Davis and Haltiwanger (1999), Rutkowski (2003) and Cazes and Nesporova (2003) 221 APPENDIX B6: DATA DESCRIPTION The data on job flows are extracted from the FINA (financial agency) database of enterprises' annual reports. This database includes 85,995 enterprises that reported their financial statement at least once in two subsequent surveys during the 1993-2004 period. Since submission of an annual report is a legal obligation for every enterprise operating in Croatia, FINA believes that the reporting enterprises account for the vast majority of operating enterprises. Only a negligible portion of enterprises decides not to report. The population for the calculation of job flow indicators in each single year includes between about 30 and 60 thousand enterprises reporting to subsequent surveys, depending on the actual year the survey was performed. Table B6.1: The Number of Continuing Reporting Enterprises 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 27.959 39.679 48.845 51.734 51.619 50.792 50.048 48.778 51.473 58.684 61.632 Source: authors' calculations. In order to provide consistency of data and clean most of the errors and omissions from the database, a visual inspection of enterprises exhibiting the largest employment fluctuations during any of the years under observation was performed. Elimination of unusually large employment swings among those enterprises on average reduced total job turnover by about 2.5 percentage points or by about 15 percent of corrected job flows. The first Labor force survey in Croatia was conducted in 1996, while regular semi-annual surveys, which are conducted to this day, commenced at the beginning of 1998. The survey conducted in the first half of 2002, which is one of the last to become available, covered 8,095 households with a total of 22,592 persons who agreed to participate. The sample was constructed using a dual-stage stratified random sample procedure. The primary sampling units are segments that consist of one or more census districts formed for the needs of the last Population Census conducted in 2001, and it was from these segments that the final secondary units, inhabited homes, were chosen (Drzavni zavod za statistiku, 2003). 222 APPENDIX C6: DATA TABLES Job Flows by Regions (Counties) ­ in % Job Creation 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Average Zagrebacka 7,0 8,8 10,9 11,4 8,9 9,7 10,4 13,4 12,9 13,1 11,8 10,8 Krapinsko-Zagorska 5,6 3,5 4,3 8,8 6,0 5,9 6,4 9,9 6,3 7,7 6,6 6,5 Sisacko-Moslavacka 4,7 4,0 7,6 8,3 9,9 9,2 9,7 6,4 5,5 7,1 7,9 7,3 Karlovacka 7,0 6,7 5,7 5,5 6,9 6,6 8,1 11,1 14,2 7,9 6,8 7,9 Varazdinska 3,5 2,1 8,6 7,4 7,3 8,2 8,5 8,4 8,2 8,3 6,6 7,0 Koprivnicko-Krizevacka 4,6 4,7 5,3 3,9 5,4 7,0 4,4 4,7 7,0 6,1 5,3 5,3 Bjelovarsko-Bilogorska 4,6 5,0 4,8 5,5 9,7 4,3 7,5 7,4 9,9 10,4 6,9 6,9 Primorsko-Goranska 6,2 4,4 6,9 6,8 7,9 8,1 9,1 9,8 9,2 8,0 8,6 7,7 Licko-Senjska 4,0 4,3 4,3 6,7 6,9 9,7 6,6 6,3 10,2 9,0 6,5 6,8 Viroviticko-Podravska 3,8 2,9 4,9 4,8 4,5 4,4 4,9 4,7 7,3 5,7 7,1 5,0 Pozesko-Slavonska 4,7 6,5 7,8 6,3 6,7 5,7 6,3 6,8 6,9 5,9 4,7 6,2 Brodsko-Posavska 7,3 8,5 8,8 8,4 8,0 6,3 6,8 7,8 12,1 13,2 9,1 8,8 Zadarska 8,3 8,1 8,3 9,5 9,0 6,7 10,7 8,7 11,6 10,0 8,7 9,1 Osjecko-Baranjska 5,5 5,9 6,4 9,7 9,2 7,5 7,7 9,3 7,8 8,5 7,7 7,8 Sibensko-Kninska 4,3 7,6 7,5 8,4 7,9 5,7 7,2 12,2 7,5 7,7 7,2 7,6 Vukovarsko-Srijemska 9,4 8,1 9,5 13,6 17,2 10,0 9,7 11,2 9,3 10,5 9,7 10,7 Splitsko-Dalmatinska 6,9 6,3 8,4 8,4 8,2 7,3 8,2 11,0 10,5 10,0 9,3 8,6 Istarska 7,2 7,3 9,0 6,8 8,8 7,7 10,3 8,9 8,1 7,6 7,6 8,1 Dubrovacko-Neretvanska 4,2 5,6 6,5 9,2 9,3 6,1 8,1 10,1 8,9 8,5 9,8 7,8 Meimurska 7,4 5,8 5,6 7,9 7,0 5,5 7,7 10,6 9,2 9,0 8,0 7,6 City of Zagreb 5,9 6,1 7,9 6,9 6,1 6,0 5,7 7,6 7,5 8,2 7,4 6,8 Average 5,9 5,9 7,5 7,5 7,4 6,8 7,3 8,7 8,6 8,7 7,9 Job Destruction 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Average Zagrebacka -9,6 -10,7 -10,2 -11,3 -9,4 -10,5 -12,8 -6,8 -8,1 -5,3 -5,5 -9,1 Krapinsko-Zagorska -6,5 -7,6 -11,2 -13,7 -8,4 -7,6 -8,8 -6,6 -7,0 -6,9 -5,6 -8,2 Sisacko-Moslavacka -8,2 -8,4 -7,8 -23,0 -11,0 -8,7 -6,4 -15,6 -6,4 -6,0 -5,0 -9,7 Karlovacka -5,9 -7,4 -12,1 -11,0 -10,6 -13,5 -12,3 -9,0 -9,6 -7,0 -7,5 -9,6 Varazdinska -10,2 -10,8 -12,2 -7,5 -6,1 -10,6 -7,1 -6,7 -4,2 -4,7 -6,1 -7,8 Koprivnicko-Krizevacka -4,5 -18,6 -6,6 -9,7 -9,4 -7,5 -7,7 -7,5 -3,0 -5,1 -6,8 -7,9 Bjelovarsko-Bilogorska -10,0 -9,5 -16,4 -12,2 -7,6 -11,5 -8,6 -10,1 -6,1 -4,2 -6,7 -9,4 Primorsko-Goranska -9,6 -11,9 -12,5 -9,1 -10,6 -10,4 -6,9 -6,6 -5,1 -6,0 -7,4 -8,8 Licko-Senjska -9,7 -9,2 -9,3 -8,7 -13,5 -15,8 -6,9 -5,5 -10,6 -7,5 -6,9 -9,4 Viroviticko-Podravska -12,6 -7,8 -7,6 -6,7 -7,7 -7,8 -17,5 -9,2 -5,8 -9,6 -6,7 -9,0 Pozesko-Slavonska -4,4 -12,7 -6,7 -9,0 -7,8 -13,2 -7,3 -5,6 -3,3 -5,6 -6,7 -7,5 Brodsko-Posavska -10,2 -10,2 -11,8 -10,0 -7,2 -11,9 -12,3 -8,0 -8,6 -6,6 -7,3 -9,4 Zadarska -13,5 -8,3 -10,5 -12,7 -6,9 -9,6 -6,8 -10,8 -7,8 -10,2 -7,6 -9,5 Osjecko-Baranjska -9,6 -11,0 -20,1 -11,8 -7,3 -9,9 -11,3 -9,2 -8,0 -7,7 -7,6 -10,3 Sibensko-Kninska -7,5 -6,3 -10,7 -14,0 -11,6 -9,1 -6,9 -7,3 -7,2 -13,0 -14,0 -9,8 Vukovarsko-Srijemska -8,7 -17,0 -13,7 -10,5 -9,9 -16,4 -9,8 -6,9 -6,1 -4,9 -6,5 -10,0 Splitsko-Dalmatinska -9,3 -11,9 -12,6 -12,8 -10,2 -9,3 -7,5 -7,7 -5,8 -6,1 -6,4 -9,1 Istarska -13,8 -11,8 -11,0 -8,9 -8,0 -8,9 -10,1 -7,7 -7,2 -7,3 -7,2 -9,3 Dubrovacko-Neretvanska -8,9 -12,3 -12,7 -12,3 -12,2 -11,9 -11,7 -10,4 -7,1 -8,5 -9,7 -10,7 Meimurska -5,7 -11,3 -12,8 -5,8 -6,2 -7,8 -6,3 -4,9 -6,8 -10,4 -5,5 -7,6 City of Zagreb -8,1 -6,2 -7,1 -8,3 -7,2 -7,8 -5,6 -7,0 -6,4 -5,7 -4,1 -6,7 Average -8,8 -9,2 -10,2 -9,9 -8,3 -9,3 -7,7 -7,6 -6,4 -6,3 -5,8 223 Job Flows ­ By Ownership Type (in %) Job Creation 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Average Government owned Communal etc. 1,3 1,7 2,9 2,9 1,7 1,9 1,6 1,2 1,2 2,7 1,7 1,9 In privatization 3,3 1,3 2,5 4,0 3,5 5,0 3,0 3,5 2,5 4,0 3,4 3,3 Privatization not started 3,0 1,7 4,0 3,6 5,7 3,7 1,8 3,1 3,5 3,2 1,8 3,2 Mixed - majority st. 1,4 1,3 2,7 2,0 3,2 2,3 2,4 5,4 2,8 2,7 3,2 2,7 Privately owned De novo 31,4 23,5 23,8 20,0 17,8 14,9 14,3 16,7 15,8 14,8 13,0 18,7 Privatized 2,8 2,3 2,8 2,4 2,7 2,6 4,1 4,3 4,2 4,2 3,3 3,2 Cooperative 2,1 1,5 1,8 3,8 7,9 3,5 4,1 3,9 7,3 7,4 7,0 4,6 Mixed ­ majority pr. 2,5 2,5 2,5 3,9 3,4 2,6 5,3 3,9 3,3 3,6 4,0 3,4 Average 5,9 5,9 7,5 7,5 7,4 6,8 7,3 8,7 8,6 8,7 7,9 Job Destruction 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Average Government owned Communal etc. 4,7 1,6 1,3 3,9 2,9 2,7 1,6 2,7 6,3 2,4 1,2 In privatization 9,5 10,3 12,8 9,5 12,1 10,7 21,4 6,7 7,0 6,3 4,5 Privatization not started 6,4 13,5 5,3 8,3 7,6 8,7 7,3 3,6 3,6 1,2 2,3 Mixed - majority st. 10,4 12,4 13,6 16,6 9,6 8,0 6,6 10,3 8,4 7,3 7,7 Privately owned De novo -9,2 -11,5 -12,2 -9,7 -10,3 -12,1 -10,4 -8,4 -7,0 -7,0 -7,1 -9,5 Privatized -9,8 -10,1 -9,8 -10,8 -8,8 -9,9 -8,6 -8,0 -5,4 -7,5 -6,2 -8,6 Cooperative -9,5 -8,8 -15,4 -13,0 -15,9 -12,7 -13,0 -10,9 -9,6 -7,1 -7,2 -11,2 Mixed ­ majority pr. -9,2 -8,7 -14,0 -10,1 -8,1 -9,1 -6,0 -9,6 -5,2 -7,6 -5,7 -8,5 Average -8,8 -9,2 -10,2 -9,9 -8,3 -9,3 -7,7 -7,6 -6,4 -6,3 -5,8 -8,8 Employment structure 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Government owned Communal etc. 17,0 17,6 17,6 17,6 18,0 14,5 17,2 15,4 14,7 15,5 15,2 In privatization 8,4 4,1 2,9 2,7 2,2 1,7 1,6 1,4 1,4 1,2 1,1 Privatization not started 3,2 2,5 1,8 1,4 1,6 1,9 1,5 3,3 1,5 2,3 2,1 Mixed - majority st. 21,6 19,7 17,5 14,3 12,1 11,7 11,0 7,3 6,7 5,8 5,2 Privately owned De novo 13,0 18,3 22,8 26,9 30,9 35,0 36,2 39,6 43,6 46,1 49,1 Privatized 20,8 20,9 20,7 19,5 19,0 19,1 18,1 17,5 16,6 15,3 14,1 Cooperative 1,6 0,9 0,7 0,6 0,6 0,5 0,5 0,4 0,4 0,4 0,4 Mixed ­ majority pr. 14,4 16,0 16,0 17,0 15,6 15,6 14,0 15,0 15,2 13,4 12,8 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 224 Job Flows ­ By Firm Size (in %) Job Creation 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Average 1-10 36,8 28,4 26,6 20,7 19,0 17,6 13,5 15,2 15,7 14,9 14,1 20,2 11-20 25,5 19,8 19,9 18,5 17,1 13,5 14,1 15,9 14,3 14,3 13,1 16,9 21-50 15,0 12,2 14,8 15,0 15,3 10,0 12,1 14,0 13,3 12,9 10,5 13,2 51-200 5,4 4,0 5,4 6,7 6,1 6,3 7,9 9,7 8,4 8,9 8,0 7,0 201-500 3,3 2,3 2,4 2,8 4,3 2,5 4,9 6,0 5,8 5,8 5,9 4,2 501- 1,0 1,3 2,5 2,4 1,7 1,8 2,7 3,3 3,7 4,0 3,4 2,5 Average 5,9 5,9 7,5 7,5 7,4 6,8 7,3 8,7 8,6 8,7 7,9 Job 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Average Destruction 1-10 -9,9 -11,7 -12,6 -11,5 -11,6 -13,2 -11,7 -10,8 -11,0 -10,8 -11,5 -11,5 11-20 -12,4 -11,9 -13,3 -11,2 -10,6 -14,0 -12,0 -8,9 -8,7 -7,1 -7,5 -10,7 21-50 -12,8 -14,0 -15,8 -10,7 -11,9 -15,3 -11,2 -11,0 -5,8 -5,4 -6,1 -10,9 51-200 -12,0 -12,1 -12,2 -10,7 -9,2 -10,3 -9,8 -8,4 -5,2 -6,2 -4,9 -9,2 201-500 -9,0 -10,5 -9,4 -10,9 -8,6 -7,8 -5,5 -6,0 -4,4 -6,9 -4,9 -7,6 501- -6,8 -6,2 -7,9 -8,3 -5,1 -4,9 -3,9 -5,2 -5,5 -4,0 -3,3 -5,6 Average -8,8 -9,2 -10,2 -9,9 -8,3 -9,3 -7,7 -7,6 -6,4 -6,3 -5,8 Employment 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Average Structure 1-10 6,9 9,9 12,6 14,0 14,8 16,1 16,1 16,1 16,6 16,7 16,6 6,9 11-20 2,4 3,2 4,1 4,8 5,3 6,0 6,2 6,7 7,3 7,6 7,8 2,4 21-50 4,5 5,5 6,4 7,3 8,3 8,9 8,8 9,3 9,6 10,1 10,7 4,5 51-200 17,0 17,2 17,3 18,3 18,9 20,3 19,3 19,1 19,1 18,6 18,4 17,0 201-500 20,7 19,9 18,5 17,0 16,5 16,6 14,7 15,0 14,7 14,0 13,9 20,7 501- 48,5 44,4 41,1 38,6 36,2 32,1 34,9 33,9 32,8 33,1 32,6 48,5 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 225 Job Flows ­ By Economic Activities (in %) Job Creation 1996 1997 1998 1999 2000 2001 2002 2003 2004 01 - Agriculture, hunting and related service activities 4.3 3.8 5.6 6.2 3.7 7.4 5.2 5.4 6.0 02 - Forestry, logging and related service activities 0.3 0.3 0.4 2.5 8.7 0.1 0.4 4.8 0.3 05 - Fishing, aquaculture and service activities incidental to fishing 12.4 6.6 6.1 9.5 9.4 13.5 11.7 8.7 8.8 10 - Mining of coal and lignite; extraction of peat - - - - - 11 - Extraction of crude petroleum and natural gas; service activities incidental to oil and gas extraction 15.4 0.0 0.0 0.1 3.6 3.7 7.1 3.2 12 - Mining of uranium and thorium ores - 13 - Mining of metal ores - 27.1 - - - - - - 14 - Other mining and quarrying 10.9 9.3 6.2 12.1 6.6 10.5 6.8 16.2 10.2 15 - Manufacture of food products and beverages 3.1 4.5 5.3 3.2 4.3 5.2 5.1 6.4 4.4 16 - Manufacture of tobacco products 1.5 1.6 - - 3.8 2.7 2.0 1.7 1.6 17 - Manufacture of textiles 2.7 6.2 4.6 1.9 4.5 6.6 4.2 6.8 6.7 18 - Manufacture of wearing apparel; dressing and dyeing of fur 3.7 3.9 3.4 3.9 6.2 6.5 3.6 4.6 4.3 19 - Tanning and dressing of leather; manufacture of luggage, handbags, saddlery, harness and footwear 3.9 5.7 2.5 7.7 4.8 6.1 2.5 4.3 8.3 20 - Manufacture of wood and of products of wood and cork, except furniture 7.8 7.6 10.0 8.0 8.7 7.5 8.2 11.0 10.5 21 - Manufacture of paper and paper products 1.8 1.7 2.6 6.5 2.3 2.8 7.3 10.9 5.7 22 - Publishing, printing and reproduction of recorded media 9.4 5.7 6.4 5.7 4.2 8.6 8.6 8.2 8.4 23 - Manufacture of coke, refined petroleum products and nuclear fuel - - 0.0 3.1 1.3 - 0.0 0.2 0.4 24 - Manufacture of chemicals and chemical products 2.2 0.6 0.8 1.2 1.1 1.2 1.8 1.6 1.0 25 - Manufacture of rubber and plastics products 6.0 7.4 11.8 7.3 7.0 12.4 11.0 9.7 7.7 26 - Manufacture of other non-metallic mineral products 3.1 3.9 3.2 4.5 4.4 5.3 6.8 9.6 7.5 27 - Manufacture of basic metals 1.9 2.2 9.4 1.3 4.0 5.1 3.8 1.8 4.8 28 - Manufacture of fabricated metal products, except machinery and equipment 8.5 9.4 8.7 5.7 9.2 11.3 11.0 10.5 9.5 29 - Manufacture of machinery and equipment n.e.c. 5.6 4.4 3.4 4.7 5.0 6.3 7.3 6.9 7.2 30 - Manufacture of office, accounting and computing machinery 23.0 22.0 26.4 31.7 34.7 7.0 12.9 10.8 10.9 31 - Manufacture of electrical machinery and apparatus n.e.c. 2.3 4.9 2.9 3.2 6.2 6.7 3.1 7.0 8.9 32 - Manufacture of radio, television and communication equipment and apparatus 3.3 2.7 3.6 4.5 3.2 12.6 4.2 4.3 17.1 33 - Manufacture of medical, precision and optical instruments, watches and clocks 5.7 7.7 4.3 5.2 4.8 7.3 9.1 6.0 8.5 34 - Manufacture of motor vehicles, trailers and semi- trailers 3.9 2.3 7.5 0.8 1.5 28.7 7.8 4.5 2.1 35 - Manufacture of other transport equipment 2.7 2.8 5.0 7.2 6.0 9.9 6.3 2.6 5.2 36 - Manufacture of furniture; manufacturing n.e.c. 4.2 3.8 3.9 5.4 11.1 5.1 5.7 6.1 6.0 37 - Recycling 5.7 3.9 11.1 10.6 4.2 11.0 10.5 11.1 11.3 40 - Electricity, gas, steam and hot water supply 3.5 2.6 1.3 0.7 0.9 0.4 3.0 1.2 0.7 41 - Collection, purification and distribution of water 6.1 4.4 3.8 2.6 1.6 2.1 1.8 2.8 2.7 45 - Construction 13.9 10.7 10.1 7.4 7.0 10.7 14.6 12.7 10.0 50 - Sale, maintenance and repair of motor vehicles and motorcycles; retail sale of automotive fuel 12.5 14.3 14.7 12.8 12.2 13.2 13.4 14.9 11.6 51 - Wholesale trade and commission trade, except of motor vehicles and motorcycles 21.6 19.0 18.0 15.6 15.5 17.7 16.2 14.9 12.9 226 52 - Retail trade, except of motor vehicles and motorcycles; repair of personal and household goods 9.6 10.2 9.2 10.0 12.4 14.9 12.2 12.3 11.3 55 - Hotels and restaurants 6.3 6.0 6.9 3.8 9.7 8.5 6.9 6.1 8.0 60 - Land transport; transport via pipelines 3.3 4.5 2.7 2.2 1.9 2.7 3.1 3.1 3.2 61 - Water transport 0.8 2.4 1.1 0.8 1.5 1.8 2.5 2.6 3.6 62 - Air transport 12.0 20.1 4.7 3.3 2.5 10.4 9.4 6.4 6.1 63 - Supporting and auxiliary transport activities; activities of travel agencies 7.4 6.2 4.6 5.0 5.4 7.7 8.2 7.9 5.4 64 - Post and telecommunications 4.5 4.9 3.4 119.6 1.7 2.0 1.2 1.4 0.7 65 - Financial intermediation, except insurance and pension funding 16.9 33.7 22.2 27.9 9.3 14.2 18.7 1.9 2.7 66 - Insurance and pension funding, except compulsory social security 200.0 - 66.7 - 44.4 53.7 13.9 28.6 67 - Activities auxiliary to financial intermediation 24.5 17.5 15.7 15.7 13.7 12.9 13.5 10.5 12.7 70 - Real estate activities 3.7 6.9 9.6 8.9 7.3 7.8 13.4 14.6 14.4 71 - Renting of machinery and equipment without operator and of personal and household goods 17.1 17.2 37.5 17.4 19.5 20.0 19.0 30.2 21.1 72 - Computer and related activities 18.5 16.6 13.7 12.6 10.5 11.7 13.0 14.1 11.2 73 - Research and development 5.8 2.8 4.2 2.6 12.9 3.6 4.2 7.6 4.7 74 - Other business activities 15.1 16.2 16.1 12.3 10.1 12.9 12.4 11.9 11.2 75 ­ Public administration and defense, compulsory social security 80 - Education 12.0 12.8 13.5 10.1 13.4 11.2 10.6 10.9 9.7 85 - Health and social work 8.3 9.2 15.4 7.0 8.5 10.8 9.1 9.2 9.9 90 - Sewage and refuse disposal, sanitation and similar activities 3.4 5.2 3.0 4.9 2.9 6.2 4.3 5.4 3.5 91 - Activities of membership organizations n.e.c. 10.6 4.2 1.5 11.8 17.6 6.1 18.3 7.1 1.1 92 - Recreational, cultural and sporting activities 11.8 7.6 7.5 7.1 7.2 8.2 10.6 15.5 13.9 93 - Other service activities 5.4 6.0 6.5 5.3 7.2 12.1 8.3 6.9 5.6 227 Job Destruction 1996 1997 1998 1999 2000 2001 2002 2003 2004 - - 01 - Agriculture, hunting and related service activities 23.2 13.9 -7.1 -12.2 -16.2 -7.5 -8.8 -6.6 -6.6 02 - Forestry, logging and related service activities -1.4 -9.0 -1.4 -0.7 -0.5 -6.3 -2.1 -0.1 -2.3 05 - Fishing, aquaculture and service activities - - - - incidental to fishing 13.8 10.1 12.3 -11.1 -10.6 -7.9 -6.4 10.0 -9.6 - - - - - 10 - Mining of coal and lignite; extraction of peat 12.2 10.1 33.1 134.6 200.0 11 - Extraction of crude petroleum and natural gas; service activities incidental to oil and gas extraction - -3.8 -3.6 -9.3 - - - - 12 - Mining of uranium and thorium ores - - - - 13 - Mining of metal ores 45.5 - -43.6 -9.8 - 136.0 13.3 15.4 14 - Other mining and quarrying -3.2 -5.5 -3.4 -4.2 -5.2 -3.6 -3.1 -3.2 -4.0 15 - Manufacture of food products and beverages -8.3 -8.0 -5.8 -8.0 -4.4 -4.7 -4.3 -4.0 -5.4 - - 16 - Manufacture of tobacco products -4.4 -6.5 15.0 -24.2 -11.4 -1.1 -4.6 -0.8 18.6 - - - - 17 - Manufacture of textiles 12.2 22.7 -7.9 -8.9 -8.8 16.7 -16.0 12.8 -8.8 18 - Manufacture of wearing apparel; dressing and - - - - dyeing of fur 13.0 14.8 10.2 -5.9 -8.5 -5.9 -6.3 10.7 -8.4 19 - Tanning and dressing of leather; manufacture of - - - - luggage, handbags, saddlery, harness and footwear -7.8 10.3 10.1 -8.1 -11.1 16.3 -8.0 18.7 -8.0 20 - Manufacture of wood and of products of wood - - - and cork, except furniture; 11.3 10.5 -7.7 -10.8 -9.2 10.3 -8.0 -7.8 -9.2 - - 21 - Manufacture of paper and paper products 12.1 20.9 -7.3 -7.3 -3.8 -2.9 -6.3 -3.3 -4.1 22 - Publishing, printing and reproduction of recorded - media -7.8 -8.0 11.7 -10.3 -9.6 -6.4 -6.3 -5.1 -4.8 23 - Manufacture of coke, refined petroleum products - and nuclear fuel -2.1 20.5 -6.0 - - -1.3 -18.4 -6.5 -2.1 24 - Manufacture of chemicals and chemical products -4.6 -7.8 -7.7 -5.2 -8.0 -9.8 -4.6 -6.0 -5.3 - 25 - Manufacture of rubber and plastics products -9.3 -9.3 -11.1 -5.2 -5.8 12.6 -7.3 -9.2 -9.1 26 - Manufacture of other non-metallic mineral - - products 12.1 -5.4 -4.6 -3.3 -4.5 -7.7 10.2 -8.0 -7.9 - - - - - 27 - Manufacture of basic metals 10.5 12.3 -3.1 10.1 18.7 38.4 -8.2 -7.7 -8.2 28 - Manufacture of fabricated metal products, except - - - machinery and equipment 12.7 -6.4 -5.6 -4.9 -7.0 11.4 10.7 -14.4 -11.4 - - - 29 - Manufacture of machinery and equipment n.e.c. 15.4 10.2 -6.9 -4.2 -6.7 11.0 -9.8 -8.9 -7.4 30 - Manufacture of office, accounting and computing - machinery -7.1 -7.3 -7.8 -7.8 -3.9 44.9 -7.1 -5.5 -6.5 31 - Manufacture of electrical machinery and - - - apparatus n.e.c. 10.9 12.6 10.3 -11.3 -3.0 -4.8 -5.9 -5.8 -5.0 32 - Manufacture of radio, television and - communication equipment and apparatus -6.8 -7.9 15.2 -20.2 -10.8 -9.1 -6.2 -5.6 -2.3 33 - Manufacture of medical, precision and optical - - instruments, watches and clocks -9.9 11.5 13.0 -10.7 -10.0 -3.9 -5.7 -4.7 -3.8 34 - Manufacture of motor vehicles, trailers and semi- - trailers -6.5 -2.8 12.8 -8.4 -32.1 -7.6 -0.7 -4.1 -5.3 35 - Manufacture of other transport equipment -7.5 -8.8 -3.3 -4.2 -6.1 -3.0 -1.7 -3.6 -6.8 228 - - - 36 - Manufacture of furniture; manufacturing n.e.c. 14.8 -7.6 10.3 -10.9 -13.3 -9.7 -10.2 10.0 -5.9 - - 37 - Recycling 16.2 14.7 -6.0 -8.1 -6.3 -4.1 -5.8 -4.0 -1.6 40 - Electricity, gas, steam and hot water supply -0.3 -0.3 -0.6 -0.6 -0.1 -1.4 -1.9 -0.4 -0.5 41 - Collection, purification and distribution of water -1.2 -0.8 -2.4 -1.1 -2.5 -1.9 -3.4 -2.2 -0.6 - - - 45 - Construction 13.2 10.3 -8.6 -12.9 -12.2 13.0 -6.1 -6.0 -6.1 50 - Sale, maintenance and repair of motor vehicles - - - and motorcycles; retail sale of automotive fuel 12.4 17.4 11.3 -11.5 -9.9 -7.7 -6.1 -5.9 -7.4 51 - Wholesale trade and commission trade, except of - - - motor vehicles and motorcycles 14.3 11.3 10.8 -11.5 -9.9 -9.1 -7.4 -7.3 -7.0 52 - Retail trade, except of motor vehicles and - - - - motorcycles; repair of personal and household goods 16.1 11.7 10.6 -12.0 -9.6 11.9 -6.4 -9.0 -6.5 - - - 55 - Hotels and restaurants 11.3 10.2 10.1 -12.5 -6.0 -6.1 -8.0 -8.0 -7.3 60 - Land transport; transport via pipelines -2.7 -3.2 -7.0 -7.0 -5.2 -5.0 -7.7 -5.6 -4.4 61 - Water transport -2.6 -2.7 -4.8 -7.4 -5.0 -6.6 -2.7 -4.9 -2.3 62 - Air transport - -0.1 -0.2 -0.3 -0.4 -0.4 - -0.6 -0.3 63 - Supporting and auxiliary transport activities; - activities of travel agencies -6.4 -7.0 11.9 -8.0 -5.8 -4.5 -4.5 -3.8 -5.6 - 64 - Post and telecommunications - - -0.0 -0.5 -0.0 -1.3 -3.4 10.5 -2.2 65 - Financial intermediation, except insurance and - - pension funding -6.8 -7.8 12.7 -24.0 -21.3 13.6 -9.5 -1.0 -0.8 66 - Insurance and pension funding, except - compulsory social security - - - -54.5 22.2 - - -9.5 - 67 - Activities auxiliary to financial intermediation -9.6 -8.1 -7.9 -9.8 -12.0 -8.2 -10.3 -9.3 11.0 - 70 - Real estate activities -3.1 -1.7 -4.1 -2.5 -3.6 12.2 -10.5 -5.8 -9.2 71 - Renting of machinery and equipment without - - - operator and of personal and household goods 13.6 22.6 -8.6 -8.2 -7.7 12.5 -14.1 -9.9 -6.9 - 72 - Computer and related activities 10.8 -7.0 -9.6 -10.0 -8.2 -7.9 -7.7 -5.9 -6.4 73 - Research and development -8.7 -5.6 -4.8 -7.0 -6.0 -4.8 -2.2 -2.0 -3.0 - - 74 - Other business activities 11.1 -8.7 11.8 -12.2 -9.1 -7.5 -6.4 -6.2 -6.0 75 ­ Public administration and defense, compulsory - social security -4.8 -0.9 19.2 - 80 - Education -5.9 -4.6 -7.5 -11.9 -8.7 10.6 -6.2 -7.6 -7.3 85 - Health and social work -6.1 -3.1 -9.4 -11.7 -4.3 -7.2 -4.0 -3.8 -3.1 90 - Sewage and refuse disposal, sanitation and similar activities -0.9 -2.4 -2.0 -3.5 -1.9 -1.1 -3.3 -1.4 -1.7 91 - Activities of membership organizations n.e.c. -5.3 -1.4 -4.5 -2.9 -5.9 -1.5 -3.9 -0.8 -5.0 92 - Recreational, cultural and sporting activities -5.4 -3.9 -3.4 -4.4 -2.2 -2.5 -5.6 -2.8 -2.5 93 - Other service activities -6.8 -6.3 -6.9 -6.7 -6.4 -8.8 -5.3 -6.0 -7.9 229