Georgia Poverty and Income Distribution (In Two Volumes) Volume 1: Main Report

Transcription

1 Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Report No GE Georgia Poverty and Income Distribution (In Two Volumes) Volume 1: Main Report May 27, 1999 Poverty Reduction and Economic Management Unit Europe and Central Asia Region Document of the World Bank

2 CURRENCY EQUIVALENTS (as of May 26, 1999) Currency Unit Lari US$ Laris WEIGHTS AND MEASURES Metric System ABBREVIATIONS AND ACRONYMS ECA - Europe and Central Asia EF - Employment Fund EU - European Union FSU - Former Soviet Union GDP - Gross Domestic Product GDP - Gross Domestic Product GSIF - Georgia Social Investment Fund IDP - Internally Displaced Person NGO - Non Government Organization PAYG - Pay As You Go RARP - Regional Agricultural Reform Project OECD - Organization of Economic Cooperation and Development OXFAM - Oxford Committee for Famine Relief SDS - State Department of Statistics SIF - Social Insurance Fund SOE - State Owned Enterprises TACIS - Technical Assistance for Commonwealth of Independent States UNHCR - United Nations Higher Committee on Refugees UNICEF - United Nations Children's Fund GEORGIA'S FISCAL YEAR January 1 - December 31 Vice President: Johannes Linn (ECAVP) Country Director: Judy M. O'Connor (ECC03) Sector Director: Pradeep Mitra (ECSPE) Program Team Leader: Ana Revenga (ECSPE) Team Members: Ruslan Yemstov, Jim Lynch (ECSPE); Alexandre Marc, (ECSHD); Nora Dudwick (ECSSD); Ghislaine Delaine (Consultant); Revaz Tsakadze, Eteri Parsadanishvili, Nodar Kapanadze, and Zurab Sajaia (State Dept. of Statistics of Georgia)

3 Table of Contents Executive Summary and Policy Recommendations... i 1. The Profile of Poverty in Georgia... I 1.1 The Extent of Poverty in Georgia Who Are the Poor? Will Growth Alone Be Sufficient to Reduce Poverty? Macroeconomic Recovery and Its Impact on Poverty Chronic versus Transient Poverty What about the Future? Medium-Term Impact of the Level and Patterns of Growth Inequality and Income Distribution Measuring Inequality in Incomes and Consumption Mobility Explaining Inequality The Role of the Labor Market Labor Market Status and Poverty Output and Employment Trends What Has Happened to Earnings? Rural Labor Markets: Low Productivity and Low Incomes Social Assistance The Importance of the Informal Safety Net The State Safety Net does not Protect the Poor IDPs Receive a Disproportionate Share of State and Humanitarian Assistance The Voluntary Sector and Civil Society: Still in its Infancy Strengthening and Improving the Safety Net... 48

4 VOL. 1 TABLES Executive Summary and Policy Recommendations Table 1: Poverty in Georgia Table 2: Incidence of poverty by regions Table 3: Income, Consumption and Estimated Unreported Monetary Income by Sector of Employment Chapter 1 Table 1: Poverty in Georgia Table 2: Incidence and depth of poverty by location Table 3: Incidence of poverty by regions Table 4: Poverty profile - selected characteristics of the poor Table 5: Structure of total consumption by deciles, average Chapter 2 Table 1: Table 2: Table 3: Table 4: Economic and Social Indicators Average Poverty Shortfall for the Poor Characteristics of the Chronically Poor Chronic Poverty by Region Chapter 3 Table 1: Distribution of Current Monetary Incomes by Income Quintiles Table 2: Average Income and Consumption by Consumption Quintiles Table 3: Income, Consumption and Estimated Unreported Monetary Income by Sector of Employment Table 4: Distribution of consumption by consumption quintiles and inequality Table 5: Transitions between consumption quintiles Table 6: Inequality in Consumption by Sector of Employment ( 4 th quarter 1997) Table 7: Inequality by Education of Household Head Table 8: Inequality in Consumption by Regions ( 4 th quarter 1997) Table 9: Structure of Income by Consumption Quintiles Chapter 4 Table 1: Employment and Output Growth by Sector, Table 2: Employment by Type, Table 3: Urban unemployment rate by regions Chapter 5 Table 1: Georgia: Total Safety Net Expenditures in 1997 Table 2: Main (Budgetized) Social Safety Net Programs in Georgia, 1997 Table 3: IDP Program 1997 and 1998

5 VOL. 1 FIGURES Executive Summary and Policy Recommendations Figure 1: Georgia: poverty projections with different universal growth rates Figure 2: Georgia: poverty projections, universal growth of consumption and worsening of distribution to the level of Brazil's inequality Figure 3: Poverty in Georgia: future path under different levels of poverty benefit Chapter 2 Figure 1: Figure 2: Figure 3: Figure 4: Figure 5: Figure 6: Figure 7: Chapter 3 Figure 1: Poverty Rate by Quarters (new poverty line) Distribution of the Population by Per Capita Consumption: Comparing First and Last Rounds Georgia: poverty projections with different universal growth rates Georgia: poverty projections, universal growth of consumption and worsening of distribution to the level of Brazil's inequality Poverty in Georgia under with and without external debt restructuring Poverty by regions: sensitivity to growth (poverty incidence in 10 years from now under different scenarios) Poverty in Georgia: future paths under different levels of poverty benefit (Base case) Lorenz Curve for Monetary Income and Concentration Curves for Consumption and "Unreported" Monetary Incomes Chapter 4 Figure 1: Labor Force Participation, Employment and Unemployment by Age Figure 2: Trends in Employment and Output Figure 3: Unemployment Rate (survey-based) Figure 4: Real Monthly Wage, Figure 5: Land Holding and Productivity Chapter 5 Figure 1: Georgia: registered unemployed, Figure 2: Poverty in Georgia: future paths under different levels of poverty benefit (New poverty line) VOL. 11 TECHNICAL PAPERS I. The Profile of Poverty in Georgia 2. Labor Markets, Inequality and Poverty 3. Inequality and Income Distribution 4. A Qualitative Study of Impoverishment and Coping Strategies 5_ What Will Poverty Look Like In Georgia in 2007?

6 Executive Summary and Policy Recommendations A significantfraction of the population in Georgia is poor. Years of crisis and civil war caused the impoverishment of large section of the Georgian population. In the last five years, greater social and political stability, along with the resumption of economic growth, have brought about a significant reduction in poverty. Per capita consumption has increased in real terms for almost all Georgian households, and increased the most for those in the lowest quintile of the distribution. However, about 11% of the population remainspoor when measured by a realistic standard; and nearly 43% of the population can be considered poor if measured by the (much more generous) official poverty line.' The most important correlates of poverty in Georgia are employment status and ownership of productive assets. Those who are unable to work (the inactive, elderly or disabled), or do not have work (the unemployed) are much more likely to be poor. In general, urban poverty is more widespread, deeper and more severe than rural poverty, although recent trends suggest that this may change in the future. Table 1. Poverty in Georgia Average All Areas Urban Rural 40.1 _ Regional differences in living standards are very wide, with the incidence of poverty in the poorest regions being several times that in the richer areas. Regardless of the poverty line used, Adjaria and Samegrelo appear to have the lowest incidence of poverty while Imereti has the highest. No region is exempt from poverty or from extreme poverty. The region with the lowest incidence of poverty, Samegrelo, still has 4.3 percent of its population that can be considered extremely poor. The poorest region, Imereti, has three times as many. Table 2. Incidence of poverty by regions 2 A F Kakheti Tbilisi ;6 7.3 Shida Kartli/Mtskheta Mtianeti Kvemo Kartli Samtskhe Javakheti Adjara Guria Samegrelo Imereti/Racha Lechkhumi/L.Swaneti Total ' Poverty rates in Georgia are also very sensitive to assumptions made about economies of scale in the household. Our estimates for Georgia suggesthat to achieve the same welfare as a single person spending I lari, a family of three would spend only 1.8 lari -i.e. there are very significant economies of scale. But if we simply counted the number of people in a family and used a per capita measure (as is done in neighboring countries), we would get much higher measures of poverty. On a per capita standard, the poverty rate in Georgiat the end of 1997, under the new poverty line, would have been 32 percent. 2 Fourth quarter, Two sparsely populated administrative regions were included in largeregions during stratification. Racha-Lechkhumi and Lower Swaneti are included in the Imereti region, and Mtskheta-Mtianeti is sampled with the Shida Kartli region. Abkhaziand South Ossetia (Tskhinvali region) were not part of the sampling frame.

7 -ii- Most poverty is transient. At the new realistic poverty line, only about 2 percent of the population remains poor for a full year. Nearly 80 percent of the poor escape poverty over the course of a year. Employment status of the household is the strongest correlate of long-term poverty. Households where the head is inactive or unemployed face the highest chances of being chronically poor. The majority of the chronic or long-term poor are urban (73%). Nearly one half live in Imereti (48%), and another one-fifth (21%) in Tbilisi. Growth can have a big impact on poverty. Simple projections of poverty rates under different growth scenarios reveal that economic growth can have a very big impact on poverty. The elasticity of the poverty headcount with respect to growth in average per capita consumption is very high, and fast growth leads to a more than proportional reduction of poverty rates. Growth of 5-6 percent per year, for example, would lower poverty to one-half its current level in just 5 years. Sustained economic growth and stable macroeconomic performance are hence crucial components of any poverty alleviation strategy. Figure 1. Georgia: poverty projections with different universal growth rates 4 5 % %.- 405% 5 -- t -t 3 5 % c2 530% cx25% -~ =----~: a20% 00 e 10O% - 5% 0% A <, sv,"_,,,_,,, aa2 Growth rate for percapita consum ption -official poverty line -new poverty line 1 The downside of this high growth elasticity is that poverty is very sensitive to a slowdown in per capita consumption growth. A slowdown in GDP growth (because of the Russia crisis, for example) or GDP growth that does not translate into growth in per capita consumption, would have an immediate negative effect on poverty rates. The impact of growth can be swamped by rising inequality. Income inequality in Georgia is very high. In 1996, the Gini coefficient for money income was 0.59-a level comparable to the most unequal economies in Latin America. In this context, any worsening of the income distribution would undermine the positive impact of growth. If the distribution were to worsen to the same level as in Brazil (the most unequal country in the world, with an income Gini of 0.61), the economy could grow at an impressive 11 percent each year and it would still take almost 10 years to restore the incidence of poverty to present levels! The Government's capacity to tax and redistribute income will hence play a very significant role in determining poverty, outcomes in the future. Without a significant improvement in this capacity, thefuture prognosis for poverty in Georgia looks bleak.

8 Figure 2. Georgia: poverty projections, universal growth of consumption and worsening of distribution to the level of Brazil's inequality 3 o % *~15 % 5 % 0% Now 03% 5% 7% 9% 1I.! growth growth growth growth growth growth for5 for5 for5 for5 for5 years years years years years - - Worsening inequality, new poverty line -I- Stable inequality, new poverty line Regional differences in poverty will not disappear with growth. The above scenarios all assume that growth is uniform across the whole economy. In reality, growth rates are likely to differ across sectors and consequently across regions, and poor regions may well lag the average in growth. Carrying out the projections while allowing growth to vary across regions in response to sectoral composition shows exactly that: poor regions such as Imereti or Guria grow more slowly than the average. As a result, existing regional differences in poverty do not disappear quickly with growth. Addressing these differences may require special "regional" interventions, particularly aimed at Imereti which is home to the bulk of the poor. Reported incomes do not accurately measure real consumption. There is a very large discrepancy between household cash consumption and monetary incomes, equivalent on average to about 40 percent of total consumption. The gap between consumption and income is lowest at both ends of the distribution (for the poorest and richest quintiles), and highest in the middle, where it represents about 45 percent of total consumption. Almost all of the observed gap is between reported total cash expenditures and reported total cash income, and represents underreporting of money incomes The magnitude of underreporting is positively associated with the degree of informalization of household employment, and with the prevalence of "gray" economic transactions. Underreported cash incomes are largest among those employed in restaurants and hotels, for whom unreported tips are an important fraction of income, and among health care workers, who rely on out-of-pocket payments by patients. But underreporting is also large for those employed in the public administration and in public utilities (perhaps reflecting widespread bribes or other petty corruption practices), as well as for those employed in education. As "formal" incomes tend to be reported accurately (estimates from the household survey of public sector wages and pensions are close to budget numbers), we assume that the gap between total monetary spending and incomes of households represent informal sector incomes. Using this methodology we arrive at an estimate of the share of informal economy equal to about 28 percent of GDP, or almost 1.9 bln. Iari in 1997.

9 -Iv- Table 3. Income, Consumption and Estimated Unreported Monetary Income by Sector of Employment (average per household for survey period) monetmy inotnes, AgricultuTe Manufacturing Electricity, gas and water Construction Trade Hotels and restaurants Financial services and real estate Public administration, defense Education Health and social services Poverty in Georgia is intrinsically linked to labor market status. The collapse of productivity and real incomes following independence was the main cause of impoverishment of the Georgian population. And inactivity, unemployment and lack of sufficiently remunerated employment remain the root cause of why families get stuck in poverty. Labor market adjustment following the collapse of output was achieved mainly through the growth of self-employment, and through the reallocation of labor towards small-scale agriculture. The resulting informalization of employment has dampened the impact of the crisis and served to protect the poor. But it remains a short-term response. Today, a large and growing fraction of the Georgian labor force still relies on self-employment as the primary means to earn an income. For some, this is an avenue for earnings mobility and growth; for the majority, however, self-employment remains constrained to low-productivity agricultural or trading activities, with little earnings stability and little potential for long term earnings growth. Prospects for future growth in living standards hinge critically on the economy's ability to generate new private employment, and to reallocate labor away from these low-productivity activities into higher value added sectors. The biggest constraints are the lack of investment and financing sources. The rural economy has played a crucial role as a safety net during the crisis years. Building on the almost universal access of the rural population to land, the rural economy has been able to absorb a huge inflow of labor released from other sectors. As a result, overall unemployment in Georgia has remained low. The down side, however, has been a sharp decline in labor productivity and consequently weak growth in rural incomes. In contrast to what has happened in urban areas, real rural incomes have remained practically stagnant during the recent high growth period, despite a record grain harvest and favorable weather conditions. This poor performance can be attributed to unequal access to inputs complementary to labor (fertilizers, tractors, capital equipment); barriers to land transactions and consolidation of holdings; lack of market access and information; scarcity of rural credit; and limited off-farm earnings opportunities. A small, but well targeted safety net, can play an important role in reducing poverty. The collapse of fiscal revenues in Georgia reduced the formal safety net to a bare minimum. An extensive informal safety net emerged in its place. However, the coverage of this informal safety net is limited, and many families slip through the cracks. The main challenge for the formal safety net for the near future will continue to be a lack of fiscal resources. Despite these constraints, however, the formal safety net has an important role to play in poverty alleviation. Given the small overall poverty gap, a well targeted poverty benefit even if small, can play a crucial role in dampening poverty. If the family allowance were to be kept at the very low level

10 observed in the 1997 budget (relative to GDP) but were accurately targeted, it would still be sufficient to reduce the incidence of poverty by several percentage points. And if the social assistance budget were to be increased more rapidly (to 1% of GDP by 2007), the impact on poverty would be much larger. Figure 3. 12% Poverty in Georgia: future paths under different levels of poverty benefit (Base case) L''':'''-'-'': --= = % % 0. 0.= 6% ~ i e 3%-_. 0% L -- Percent poor - initial level Poverty incidence- base case (no targeting of actual (1997) poverty benefit) Poverty incidence - base case, increase in poverty benefit to the planned 1998 levels and targeting Poverty incidence - base case, targeted poverty benefit gradually increasing to 1% of GDP by 2007 IDPs receive a disproportionate share of State and humanitarian assistance. The IDP program is one of Georgia's largest safety net programs, ranking second only to old age and invalidity pensions. However, IDPs do not appear to be more vulnerable than other population groups. In fact, IDPs that have resettled on their own or have integrated themselves into local communities face a lower risk of poverty than the average Georgian household (4 percent of them are poor as opposed to 1O percent of the total population). And they face the lowest risk of extreme poverty of almost any population group. The contrast between benefits received by IDPs and those received by other, often more needy, families highlights the need to improve the targeting of assistance to IDPs. In addition, in-kind benefits should be phased out, or at a minimum, their costs should be fully budgetized. How can the safety net be improved in the short and medium term? In the very short term (1999), actions can focus only on improving targeting and redirecting resources to the truly needy, mainly by: * Protecting the family allowance system. Ensure that the allocation in the 1999 budget is at least the same level as in the 1998 budget and that, unlike in 1998, it is fully paid out. * Eliminating tariff discounts to special population groups and other in-kind benefits. Replace with a direct payment (through a modest increase in the family allowance system) for those groups of recipients who are truly needy, mainly the disabled (categories I and II). * Eliminating direct payments from the budget for electricity. This can be phased in geographically, starting with Tbilisi, as electricity distribution companies are privatized. * Improving the targeting of assistance to IDPs. Trhrough self-targeted public works, proxy means testing or frequent registration requirements. Redirect released resources to the family allowance program or towards establishing public works programs as discussed below.

11 -vi- * Continue to support the integration of IDPs housed in institutions into local communities. The evidence from the household survey suggests that once this happen, IDPs are able on their own to dramatically improve their living conditions.. Over the medium term ( ), a more comprehensive strategy can be put in place, based on the following two principles: * Cash benefits should remain limited to individuals not able to work and not supported by families with members able to work-i.e. mainly pensioners living alone and the disabled. * The poor able to work should be supported through self targeted employment schemes organized to provide a minimum level of subsistence. Such programs would combine the best features of an Employment Guarantee Scheme with those of Social Funds aimed at promoting labor intensive projects in poor areas. To ensure that the money goes only to those who need it, without undermining the incentives to seek a normal job, the wage offered has to be low (at the level of local wage rate for unskilled manual labor in "normal" times).

12 1 The Profile of Poverty in Georgia Years of crisis and civil war caused the impoverishment of large section of the Georgian population. In the last five years, greater social and political stability, along with the resumption of economic growth, have brought about a significant reduction in poverty. However, not all groups have benefited equally: about 11 %of the population remains poor, and nearly 9% is extremely poor. The most important correlates ofpoverty in Georgia are employment status and ownership ofproductive assets. Those who are unable to work (the inactive, elderly or disabled), or do not have work (the unemployed) are much more likely to be poor. In general, urban poverty is more widespread, deeper and more severe than rural poverty, although recent trends suggest that this may change in the future. Regional differences in living standards are very large, with poverty rates in the poorest regions twice those in better-off areas. 1.1 The Extent of Poverty in Georgia 1. Just over half a million Georgians-I 1 percent of the population-are poor. By this we mean that they live in households where total consumption falls below a minimum subsistence level or poverty line. 2 Almost 9 percent of the population is extremely poor: their monthly consumption is below the cost of a "survival" food basket, which achieves only a minimum caloric intake. The majority of the poor (56%), live in urban areas. And, on average, urban poverty is higher, deeper and more severe than rural poverty, although there are signs that this ranking may reverse over time. 2. Any poverty rate depends critically on the poverty line chosen, and on the choice of methodology used to compute it. For this reason, it is often hard to compare poverty rates across different countries: these comparisons can be meaningless if the respective country poverty lines are widely different. In this profile for Georgia, we use three lines: an official line based on the (unrealistically) high nutrition norms inherited from the pre-independence period (equal to 102 lari per month per equivalent adult); a new poverty line based on a revised minimum food basket and derived from the actual consumption patterns of the population (equal to 52 lari per month per equivalent adult); and an extreme poverty line, drawn at 40 percent of median consumption, and equivalent to the cost of a "minimum survival" food basket (of 40 lari per month per equivalent adult). Table I summarizes the incidence, depth and severity of poverty for all three lines. 3. Not surprisingly, there are very large differences in the incidence of poverty depending on whether we use the official or the new poverty line. The official line is almost exactly double the new line, so we would expect to obtain a higher poverty rate. But the poverty headcount is almostfour times higher when we use the official line. This indicates that households are "bunched" between these two lines; e.g. almost a third of the population could be regarded as poor by a generous standard (official), but not poor from a more realistic point of view (new line). This comparison also reveals that about 30 percent of the population that is currently non-poor is vulnerable to poverty. 3 This chapter is a summary of the more detailed analysis presented in The Profile of Poverty in Georgia, Technical Paper 1, Volume 11 of this study. Data, methodology and findings are discussed there at length. 2 These figures correspond to end- 1997, and are drawn from the quarterly Survey of Georgian Households which was conducted by the State Department of Statistics of the Republic of Georgia (SDS) during June, 1996 through December, Their consumption is above the poverty line, but less than double of the poverty level; e.g. these people are close enough to poverty to feel themselves insecure.

13 -2- Table 1. Poverty in Georgia 4 YeT7-Ph e O#ia lpoeitline r New lines, -e~r~ ~i~ Poverty incidence (% of population that is poor) Poverty Depth Average shortfall of the poor Severity of poverty As illustrated by Table 1, the incidence of poverty depends critically on the choice of a poverty line. But poverty rates are also very sensitive to assumptions made about equivalence scales and economies of scale in the household: whether we simply count the number of people in a household, or try to adjust in some way for the composition and size of families can have a huge impact on the measurement of poverty. In the case of Georgia, for example, there are solid reasons for believing in sizeable economies of scale. 5 But if we simply counted the number of people in a family and used a per capita measure (as is done in neighboring countries), we would get much higher measures of poverty. On a per capita standard, the poverty rate in Georgia at the end of 1997, under the new poverty line, would have been 32 percent-almost three times as high as the actual estimate obtained using economies of scale! Poverty in Georgia would still be lower than in neighboring countries, but less dramatically different. So, what measure is one to rely on? One that is comparable with those of other countries or one that is country-specific, but realistic? It depends on the purpose for which poverty is being measured; if the goal is setting policy, then what matters most is to construct a measure of poverty that is the most accurate for that particular country, and best reflects its specific characteristics. Table 2. Incidence and depth of poverty by location (>fflvinl >ver line,, r.i, povrwty vtr...., _0 -;1k -wv po UJrban Areas: Poverty incidence Average shortfall of the poor Poverty depth Severity of poverty Rural areas: Poverty incidence Average shortfall of the poor Poverty depth Severity of poverty Rural versus Urban. Regardless of which poverty line one chooses, a higher proportion of the urban population lives in poverty. Urban poverty is also deeper and more severe. From this perspective, Georgia looks similar to many other FSU countries. Rural areas have lower incidence of poverty because the rural population has much better access to at least one 4 These figures are for the fourth quarter of An individual is classified as poor if he/she lives in a household with consumption below the poverty line. Poverty incidence is percentage of the population that is poor; the shortfall of poverty is the distance between the average consumption of the poor and the poverty line (as percent of the poverty line); the index of poverty depth distinguishes among the poor according to how far below the line their consumption falls; it is the poverty shortfall averaged for the whole population (zero for non-poor); severity ofpoverty gives relatively greater weight to poverty shortfalls that are way below the line (e.g. it captures the degree of inequality among the poor). Data are from State Department of Statistics, Survey of Georgian Households. Estimates use sampling population weights. 5 Our estimate of theta is 0.6, meaning that to achieve the same welfare as a single person spending I lari, a family of three would spend only 1.8 lari. These methodological issues and the sensitivity of our results to economies of scale are discussed in detail in Technical Paper I and Annex I, Volume ll of this study.

14 -3- productive asset, namely land. 6 However, rural poverty is also highly sensitive to seasonality; in other words, the incidence of poverty in rural areas varies with the agricultural cycle. During the slack agricultural season, rural poverty increases; after the harvest, the situation improves. As we use consumption, not income, to measure poverty, this time pattern suggests that rural households are unable to spread their consumption evenly throughout the year. Comparing Absolute Poverty People living at less than one dollar a day at purchasing power parity (in 1985 prices) (percent of population) Country Poverty incidence Year GEORGIA Armenia Bulgaria Estonia Kyrgyz Republic Poland Turkey Source: all countries, except Turkey, Armenia and Georgia from Worldi Bank, WDI(1998); Turkey, Armenia, Georgia own calculations estimates using household survey data; PPP - World CPenn tables. 6. Regionalpoverto. Regional differences in liviing standards are very wide, with the incidence of poverty in the poorest regions being twice that in the richer areas (see Table 3). Regardless of the poverty line used, Adjaria and Samegrelo have the lowest incidence of poverty; Imereti the highest. Tbilisi has a poverty incidence that is slightly below the national average for lower lines and slightly above for the official line. No region is exempt from poverty, even from its extreme form. The region with the lowest measured poverty, Samegrelo, still has 4.3 percent of its population in extreme poverty. The poorest region, Imereti, has three times as many. Imereti is also the most populated region, hosting almost 25 percent of the surveyed population of 6 Note, however, that we base our comparisons on total consumption. Once incomes or monetary expenditures are used, rural areas rank much poorer than urban areas. This impliets that in rural areas most of the products are not sold at the market, but consumed in-kind.

15 -4- Georgia; as a result, is home to a large share of the poor (32 percent of the total poor population of Georgia lives in Imereti). Imereti is also the region where the poverty is the deepest: most of the poor are extremely poor. Table 3. Incidence of poverty by regions 7 Kakheti Tbilisi Shida Kartli/Mtskheta Kvemo Kartli Samtskhe Javakheti Adjara Guria Samegrelo Imereti/Racha/Swaneti Total In large part, these regional disparities reflect differences in initial conditions or endowments, aggravated in some cases by the vastly differing opportunities generated by growth and transition. The inherited structure and specialization of production is one explanatory factor. Regions with high poverty rates were predominantly former industrial centers. Today, most of their factories and mines are obsolete, inefficient and largely inactive. Regions with low and medium poverty rates are more dynamic and diversified, agro-industrial areas or trading centers (the capital and main ports). On the other hand, geographical factors also play a role. Among all survey regions, Imereti has the highest proportion of the population living at high altitudes. Due to neglect of road maintenance, many of these sites are now isolated for many months of the year. 1.2 Who Are the Poor? 8. Table 4 summarizes poverty rates by demographic and economic characteristics of household and individuals. While demographic composition of the household is somewhat correlated with poverty, the biggest determinants of poverty risk in Georgia are clearly the economic variables of the household: factors such as the employment status and earnings of the household head, and ownership of productive assets (primarily land). 9. Poverty and Employment Status. Households with an inactive or unemployed head face the greatest risk of being poor. Almost a third (3 0%) of all households in which the head of household is not employed (unemployed or inactive due to old age, disability or having dropped out of the labor force) are poor, and 26 percent are extremely poor. These families, moreover, constitute a large share of all the poor (28%), and nearly one third (32%) of the total extremely poor population. This inactive population is disproportionately located in urban areas: 86 percent of households with an inactive or unemployed head are urban. However, rural households with an inactive head of household are at a particular high risk of poverty. These are primarily households which either have no land or are not able to work it because of old age or disability. 7 Fourth quarter, Two sparsely populated administrative regions were included in largeregions during stratification. Racha-Lechkhumi and Lower Swaneti are included in the Imereti region, and Mtskheta-Mtianeti is sampled with the Shida Kartli region. Abkhaziand South Ossetia (Tskhinvali region) were not part of the sampling frame.

16 In both urban and rural areas, inactivity is the strongest predictor of poverty. However, in rural areas we find many people working in agriculture who are also poor. About 15 percent of households in which the head is self-employed in agriculture are poor (as compared to a national average of percent), while 9 percent are extremely poor. This is linked to low levels of productivity in agriculture and low level of earnings for farmers (see below and Chapter 4). Surprisingly, there are no clear differences in poverty incidence between those employed for wages in the state versus the private sector, or relative to those who are self-employed outside of agriculture (mainly trade). There are, however, signs that these sectoral patterns may be changing: by end (our last observation point), those employed in private firms clearly had the lowest risk of poverty at any line, suggesting that they are benefiting the most from economic growth. 11. Poverty and Ownership of Assets. After the economic collapse of industry, agriculture emerged as the most important economic activity in the country, and remains so to this day. At present, incomes from land (agricultural self-employment and subsistence gardening) constitute just about half of all incomes of the population (including the urban population), and over 70 percent of the incomes in rural areas. In Georgia today, agriculture is the rural economy; it is also a substantial part of the urban economy! 12. The most important asset in termns of affecting the risk of poverty in rural areas is ownership of land. Landless households have a higher poverty risk and depth of poverty than households with land; and three times the poverty risk and depth of poverty than households that own more than 1 hectare. However, only a relatively small proportion of the rural population is landless: about 8 percent of rural households do nol: own land, or do not use any land. Half of these rural landless poor do not have any employed members at all, and are predominantly single elderly. 13. The bulk of the rural poor do own land, but own less than the non-poor and use it less effectively. On average, the rural poor own about 0.5 of a hectare, compared to 0.7 owned by the non-poor. More importantly, the poor spend only 2 lari every month on farming expenditures, whereas the non-poor spend 8.5 lari. This translates into lower yields and less diversified crops for the poor. On average, the non-poor are cropping three different cultures; the poor one or one and a half. Cultivation of technical plants (sunflower, tobacco etc.) is almost exclusively done by non-poor households. This apparent link between cropping patterns and poverty suggests that agricultural specialization is an important factor behind inter-regional differences in poverty risks. For example, corn and wheat are now relatively pro-fitable crops; as a result, cereal growing regions have a much lower number of poor households than citrus and grape growing areas. 14. When a rural household has other assets (such as draft animals or cattle), its poverty risks are greatly reduced. These assets substantially increase the productivity of agricultural employment, they also allow for the diversification of sources of income, and hence reduce the risk of poverty. 15. For urban households, the link between poverty and ownership of assets is less clear. Owning some financial assets (such as a bank deposit) is correlated with a very low (or zero) risk of poverty, but only a miniscule fraction of the population reports owning these assets. Owning a house versus renting also appears to slightly reduce the risk of poverty, but on the other hand, the majority of the population (and hence of the poor) own their house. Ownership of durables (refrigerators, TVs) is no different between the pooi and non-poor, or between urban and rural areas. However, owning a car, one of the few durables that can be put to productive use, clearly reduces the risk of poverty.

17 -6- Table 4. Poverty profile - selected characteristics of the poor I. HBousehold demwgraphic profile Couple with Children 6.2% 1.2%/ 0.5% 7% Extended Sigl Wokn with no Ag Children Adul... t 8.1% 32.1% ' '.9% 20.9%.1%Ae. 12% Extended with Children 18.8% 3.6% 1.5% 66% Pensioners with Children 17.1% 13.1% 10.0% 0% Single Working Age Adult 10.0% 3.9% 2. 1% 1% Single Adult+ Pensioner 9.3% 3.9% 2.6% 5% Single Mothers 12.0% 1.6% 0.4% 1% Single Pensioner(s) 12.3% 3.8% 2.0% 7% 2. Household head enployment: Inactive or unemployed 30.1% 8.9% 4.8% 28% Self-employed in agriculture 15.2% 2.5% 0.9% 30% Employed in budget sphere 10.1% 2.5% 1.4% 13% Employed in SOE 11.1% 2.3% 0.9% 10% Employed in private firm 11.0% 3.1% 1.2% 10% Self-employed outside agriculture 7.7% 0.8% 0.3% 14% 3. Education (for persons over 15years old): Elementary 22.9% 3.4% 1.4%. 9% Secondary incomplete 14.3% 3.0% 1.4% 10% Secondary 13.9% 3.0% 1.3% 30% Post secondary 10.9% 2.8% 1.2% 12% University 7.8% 1.5% 0.7% 9% Total 13.4% 2.9% 1.3% 1 100% 16. Dependency and Poverty. The demographic structure of households in Georgia has been sharply altered by the fall in living standards over the crisis years. In response to the collapse of wages and incomes, nuclear households have expanded to form larger, extended households which provide a better safety net. Today, extended multigenerational households represent over one half of all households in both urban and rural areas. Few single households or single parent households remain-although those who do are at a very high risk of poverty, since they tend to be the ones without relatives to rely on or to take them in. Nuclear families-couples without children, or with children but no other dependents-now represent only 40 percent of all households (down from nearly 70 percent), and have a low risk of being poor. The fact that they can choose to live as a nuclear family is in itself a sign that they are better off than those who nust group into extended family homes. 17. The majority of the Georgian poor live in extended, multigenerational families with children. Almost 19 percent of such families are poor; and since there are many of them, they also constitute the bulk of the poor population (66%). A key factor of poverty appears to be the dependency ratio-i.e., how many persons have to live on each breadwinner's income. The poorest households are those in which no one has gainful employment, or those in which more than three dependents are supported by each employed member. 18. Age, Education and Gender. There are no large differences in terms of poverty risk between age groups, with the exception of single pensioners, who face a high risk of being poor. About 17 percent of urban single pensioners are poor; and over 80 percent of these poor single pensioners are also extremely poor. In comparison, 8.5 percent of urban working age males and 8 percent of urban working age females are poor. Children do not appear to be at especially high risk: 9.8 percent of all children of school age are poor (the same proportion in urban and rural areas). Young children, who often are at higher risk because childcare demands prevent the women in the household from working, also do not face a poverty risk that is higher than

18 -7- average-at least in urban areas, where just under 9 percent of children under 7 are poor. But they do face a somewhat higher risk of poverty in rural areas, where 11.5 percent of children under 7 are poor. This could reflect the lower availability of childcare options in rural areas. 19. As we may expect, education is inversely correlated with poverty. Having only an elementary education significantly increases the risk of being poor, while a higher education clearly reduces it. However, we do find poor individuals at all levels of education (and individuals with a higher education represent an equally large share of the poor as do those with only an elementary education) Female-headed households are more likely to be poor than male-headed households. And when we consider jointly marital status, gender and poverty of individuals, the underprivileged position of women become more evident. Single women and women in unregistered wedlock have much higher risks of poverty than men in similar situation; and especially, single female pensioners (widows) are particularly vulnerable to poverty. 21. "Vulnerable Groups". A number of population groups were perceived under the former Soviet system to be particularly vulnerable, and were granted special privileges, some of which persist in Georgia today (such as discounts on electricity tariffs, transport etc). Some of these groups were identified in the survey instrument. Comparing their risk of poverty with the average incidence for the population as a whole reveals some interesting findings. The disabled (especially the seriously disabled, groups I and II) at are a high risk of poverty; a large share of them are also extremely poor. Special pensioner categories (veterans, defense employees and others entitled to special pension supplements) have a lower risk of poverty than non-special pensioners and than the population as a whole. IDPs covered by the survey (those that have resettled into households as opposed to remaining institutionalized) also fare much better than the population on average (only 4 percent of such IDPs are poor). 9 Basic Needs 22. Consumption patterns. As revealed by the share of food in total consumption (almost 70 percent), the population on average is rather poor. In this context, the poorest decile of the population, which devotes three-quarters of its total monthly consumption to food, is not dramatically different from the richest, who still have a food share of 60 percent. What is rather unusual is the very high share observed, evenfor rich households, of food in the form of own agricultural produce consumed in-kind. This reveals that even the relatively affluent depend on their garden plots to guarantee food security. 23. Outlays for housing, health care, transport, clothes, electricity and fuel are the major nonfood budget items for the poor and non-poor alike (Table 5). But jointly they claim only about 25 percent of all consumption expenditures. Education has a substantially lower weight in the budget, but unlike for other non-food items, the non-poor spend substantially more than the poor. s A more detailed discussion of the links between education, incomes and poverty can be found in Chapter 4, and in Technical Papers I and 2 in Volume By definition the sampling frame in the survey does not cover institutionalized population, so IDPs living in dormitories, hostels etc were not sampled.

19 -8- Table 5. Structure of total consumption by deciles, average (percent of total consumption) I(poorest) (richest) Total Note: deciles are drawn based on consumption per equivalent adult 24. Nutritionalstatus. Georgian households do not face a problem of dire caloric deficiencies, but do face problems of adequate composition of food. The SDS survey found that the average food consumption per equivalent adult in the lowest quintile was 1700 Kcal per day (as compared to the recommended minimum of 2000 calories). This level is low, but not so low as to lead to serious nutritional problems. However, it does have to be put in context as traditionally Georgia had high consumption standards, both in terms of quantity and quality of food consumed. It is not surprising then to find that a sizeable proportion of households assess their nutritional status as deficient: according to a Vulnerability Assessment conducted in March 1996, 23 percent of urban households and 17 percent of rural households evaluated their nutritional status as bad or unbearable.' The assessment also found that about 1 percent of all households can afford to consume only two food items on a regular day (most often bread and potatoes or beans). There are clear differences in the diversity of food basket between different groups: 5 percent of single pensioners and 2.3 percent of single parents can afford to eat no more than two food products regularly. 25. Basic services. The link between access to basic services and poverty in Georgia is very weak. During the years of crisis, the breakdown of municipal services pushed the standards of living for all households to very low levels. Today, the simple fact of possessing a certain facility does not necessarily mean that it is used by the household. For example, while almost all Georgian households (97%) have an electricity connection, supply has been unreliable and most households have received no more than 2 or 3 hours per day. Gas supplies have been reduced to very low levels, or stopped altogether. Lack of electricity and gas supplies have made it difficult to heat dwellings or prepare hot meals, and have forced the majority of Georgian households to rely on kerosene and wood as their main heating fuels. 26. According to the SDS household survey, the link between poverty and basic services is not uniform, but depends on area, region, and the type of service. There are, for example, strong differences in access to water and sewerage between urban and rural areas: about 80 percent of urban families have hot water (either centrally or individually supplied), whereas in rural areas less than 10 percent of households have this utility. Only every fifth rural household has a telephone; in urban areas this service is provided to near half of all households. Similarly, piped water reaches almost 100 percent of urban dwellers, but only two-thirds of the rural population. Taking into consideration these urban-rural differences, we find that the poor are clearly worse off than the non-poor in using gas (individual or central connection). Households that do not use "' L. Dershem, D. Gzinshuili, A. de Roos, and D. Venekamp. "Food, Nutrition, Health and Nonfood Vulnerability Georgia, 1996: A Household Assessment", ECHO, Henceforth referred to as "Vulnerability Assessment". in

20 -9- gas at home are more likely to be severely poor than others. Other services, however, do not render themselves useful to distinguishing the living conditions of the poor from the non-poor. 27. Access to health care. Access and use of health care in Georgia are closely correlated with poverty status, and there are growing signs that the poorest groups of the population are being excluded from the health care system. Moreover, in the absence of effective health insurance, health risks are emerging as one of the important factors of vulnerability. 28. Morbidity in Georgia is quite high. According to a recent SDS survey, a majority of all households (close to 80 percent) reported having at least one ill member over the 3-month survey period; 14 percent of households reported having a member with serious physical limitation impairing her or his daily life." Despite this, actual use of the health care system is neither high nor universal. According to the SDS survey, only 46 percent of individuals who were sick sought professional treatment; the rest treated themselves or sought non-traditional forms of help. Out of those who were sick, but treated themselves, about 20 percent did so because they were not able to afford professional treatment." 2 Among those in the poorest quintile, the share of those with health conditions who said they could not afford professional help, was twice as high as in the richest quintile. 29. As the trade-off between health care and food for poor families is so extreme, this group faces difficult choices in trying to mobilize funds necessary for treatment. When they do turn to doctors for help, medical spending becomes a very big item for the poor, and the total amounts are almost comparable to what the non-poor spend. Poor households seeking curative care spend on average 7.8 lari per month on medicines (compared to 11.5 lari for non-poor), 5 lari on visits to the doctors (as opposed to 18 lari for non-poor), 36 lari for hospital care (as opposed to 50 lari for non-poor), 70 lari for child birth (140 for non-poor). Thus, a single case of hospitalization can eat up the full monthly earnings of a poor family. 30. Out-of-pocket payments have long become a condition of receiving treatment even in public health facilities. And without effective health insurance, the illness of a family member emerges as one of the most serious risks of slipping into poverty. There is a sizable number of households in the survey who report spending all their cash on medical care: almost 5 percent of poor households and 1 percent of non-poor households surveyed had no more than 5 lari left after health payments, to meet all other monthly needs. As a result, health care emergencies are an important cause of indebtedness for households. Nearly half of the households in the poorest quintile borrow or sell property to finance health expenditures Real public expenditures of the Ministry of Health declined over the years of transition ( ) by more than ten times. At present, only 22 percent of total finance for health care is coming from the State central and municipal budgets, or from medical insurance. The overall constraint on resources has meant diminishing funds, low salarieg 4 and no maintenance of equipment or facilities. It has forced health facilities to rely increasingly on out-of-pocket payments, which have effectively priced the poor out from hospital care. According to the SDS 1 SDS, Survey of the Demand for Health Care Services, Summer ' The Vulnerability Assessment findings are very similar: 21.7% of all individuals with health problems did not contact a health care provider because of the inability to pay. '3 SDS Health Demand Survey. Confirmed by the SDS household survey, which shows that 40% of the poor with medical expenses borrow and/or sell their property. 14 Health care has one of the lowest monthly salaries (about 1/2 of the national mean) in the Georgian economy both according to official data and to SDS household survey. However, as an indicator of importance of informal payments, the poverty risks for those employed in health care are not higher, and are even slightly lower, than for other employees in state and private firms.

21 -10- health survey, only 10 percent of hospital care users are from the poorest quintile of the population. This usage pattern highlights an anti-poor bias in the allocation of health expenditures: over 50 percent of the total expenditures of the Ministry of Health and State Medical Insurance Company go towards financing inpatient curative health care yet, as the SDS survey reveals, this type of care is rarely used by the poor. 32. Access to education. Education is inversely correlated with the level of poverty, but the link is strong only for higher education. Secondary education in particular, seems to provide poor shelter against the risk of poverty. This, combined with less affordable access to higher education, have weakened the incentives to obtaining an education among the poor. The fall in demand has been accompanied by the collapse of public financing of education (which declined by a factor of 11 times between 1991 and 1996), and by a sharp deterioration in quality. 33. Total gross enrollments have fallen by about 18 percent during transition years (by 30 percent in urban areas and by 10 percent in rural areas). This is partly due to demographic factors, and partly to the reduction of the population covered by the existing systemr 5, but there is also a clear deterioration in the effective use of the education system by the population. Primary education is, in theory, free and universal. However, net enrollment rates have dropped sharply, from 91 percent in 1990 to 74 percent in 1994; gross enrollments have fallen from 95 percent to about 80 percent. Secondary school is still compulsory and free up to the 9th grade. However, enrollment rates there have also fallen: from 95 in 1990 to 76 percent in For 10th and I Ith grades there are formal enrollment fees that families have to pay. Formal fees are modest (10 lari per month in 1997), but for a poor family it can be unaffordable (average consumption per capita in poor families is only 20 lari per month). Apart from these formal fees, there are ad-hoc informal payments collected from parents of pupils at every grade, usually to pay for school heating, urgent repairs etc. And there are other significant costs at all grades for books, writing materials, clothes, pocket money for school breakfasts and transport. 34. Education expenditures are the most unequally distributed item in the structure of consumption in Georgia (Table 5). On average a poor family sending a child to school spends annually about 15 lari, or just over 2 percent of their cash budget. The non-poor spend about 140 lari or over 5 percent of their total cash expenditures." 6 Two factors seem to be driving the inequality in private educational spending: one is textbooks (which represent about a quarter of total private spending on education); a second are private lessons (which represent about 10 percent of all private spending). Almost no poor household in the survey reported taking private lessons. Given the poor quality of secondary instruction in Georgia, no private lessons and lack of textbooks effectively act as barriers preventing children from poor families from accessing higher education. It also translates into unfair differences in quality of education even for the same level and same school, and thus creates very discouraging expectations among poor students, who believe that they will never be accepted at higher education establishments" The resulting unequal access to post-compulsory education is reflected in the SDS survey: only 20 percent of individuals aged in poor families are students, as compared to 78 percent in non-poor families; and out of all families that have a child enrolled in higher education, only 6 percent were poor. Public expenditures on higher education clearly benefit the rich more than the poor." 5 Due to the breakaway of some regions. 16 Data from SDS household survey. 17 See N. Dudwick, Georgia: A Qualitative Study of Impoverishment and Coping Strategies, Technical Paper 4, Volume II of this study. 18 In 1997, the share of higher education in the Ministry of Education's budget was almost 50%, suggesting that spending is biased against the poor.

22 -11- Will Growth Alone Be Sufficient to Reduce Poverty? The rise of widespread poverty in Georgia is a direct consequence of the economic collapse and civil strife following the breakup of the Former Soviet Union. The recovery of growth has done much to reduce the number of the poor; but not all population groups have benefited equally from growth. Hyperinflation, macroeconomic crisis and weak governance have combined to cause a sharp rise in inequality among households. While sustained growth is an absolute priority for continued poverty alleviation in Georgia, it is not clear if growth alone will suffice. The existence of large inequities between households, groups and regions suggests that pockets of more structural poverty will persist even in afast-growing economy. The quality and patterns of growth-and in particular, whether growth translates into widespread increases in productivity and labor incomes, or whether it results in increased incomes for only afew-will be the key determinants of the evolution ofpoverty during the next fewyears. 2.1 Macroeconomic Recovery and Its Impact on Poverty 36. In early 1994, Georgia emerged from two years of armed conflict with a collapsed economy. Officially estimated GDP had dropped to 30 percent of its 1990 level; government revenues were barely 2 percent of GDP; and the economy was ravaged by hyperinflation and the ensuing depreciation of the domestic currency. These developments had a severe impact on the population: real wages fell by 90 percent, and social indicators worsened dramatically. As a result of the collapse in real incomes, household expenditures became severely compressed, with a large part of the population relying on subsistence gardening and barter trade to provide for their most basic needs. At the same time, the fiscal collapse made it impossible for the Government to finance its basic functions, let alone an effective social safety net to protect the most vulnerable. Family networks, private support mechanisms and Georgians' traditional entrepreneurship became the only reliable mechanisms to guarantee survival. Table 1. Economic and Social Indicators Real GDP (% change) CPI inflation (year end) 7,488 6, Employment ('OOOs) State sector 1, , , Private sector , ,617.6 Unemployment ('000s) Registered o/w receiving benefits Avg. monthly wage (laris) Source: IMF 37. Since 1994, the economic situation has improved dramatically. The economy has been growing steadily since 1995, at an average of 8 percent per annum. Inflation has come down to moderate levels and the fiscal situation, while it remains extremely fragile, has been brought under control, allowing for a modest improvement in the financing of social expenditures. These trends have underpinned an improvement in living standards, which nevertheless remain significantly below their pre-transition levels. 38. What has been the impact of growth on poverty? It is hard to carry out a rigorous assessment since the SDS survey data starts only in However, some useful insights can be 1 This Chapter draws extensively from Technical Papers I and 2 in Volume II.

23 -12- gained from looking at some earlier surveys and poverty work. For example, a UNICEF study found that in 1994 an overwhelming 85 percent of the population was poor. A survey conducted in early 1996, on a sample of 1200 households, found that 71 percent of urban and 43 percent of rural households were below the (official) poverty line? Applying similar methods and poverty line to the SDS household survey, we would obtain a poverty incidence of about 55 percent of the population by end-1996/early Thus, there has been some progress in reducing the number of poor in the years after the stabilization of economy ( ). 39. The data from the SDS survey also allow us to estimate trends in poverty over the recent high growth period ( ).4 In urban areas, there has been a clear reduction in poverty rates (Figure 1). In rural areas, the improvement is much more muted (and it is difficult to separate trend from seasonal variations). 5 Figure 1. Poverty Rate by Quarters (new poverty line) 25% -. _ 20% -0 Urban!0o/o -U~~~~~~~~~~~~~~~~~~~~~~~~~~~Ruiral Percent poe r I 5% I 0%- I I Summer 96 Fall 96 Winter 97 Spring 97 Summer 97 Fall 97 Quarters 40. The differential performance of rural and urban areas reflects primarily very different patterns of growth across sectors: value added in transport, trade and services, all of which are primarily urban, grew by percent in real terms during ; in contrast, value added in agriculture, which represents the bulk of the rural economy, grew by only 2 percent. At the same time, agriculture absorbed much of the employment released in other sectors of the economy; the result was much lower growth in productivity per worker, and consequently lower growth in real incomes (and consumption) for those employed in the sector' At end- 1997, value added per worker in agriculture was only about 100 lari per month; four times lower than value added per worker in trade, and five times lower than in industry or construction. 2 Based on the official family budget surveys data, referred to in "Demand for health care services and expenditures" UNICEF, ECHO, Food, Nutrition, Health and Nonfood Vulnerability in Georgia 1996: A Household Assessment (also referred to as "Vulnerability Assessment"). The poverty line used was a predecessor of the official poverty line established by Law in 1997; at the time of the survey is was 86 lari per equivalent adult. 4 There is a methodological break in the series that makes measures of poverty incomparable between the initial four rounds and the last two. But one can make comparisons within the first four and within the last two, and infer a trend. 5 An extreme example of these urban-rural differences is found in Adjaria: households that were interviewed for four consecutive quarters in urban areas reported their consumption had doubled; in rural areas, average per capita consumption in households that participated in the survey for four consecutive quarters actually dropped. This translates into an astonishingly large discrepancy between urban and rural poverty in this region: while less than 1% of the urban population in Adjaria is below the poverty line, as much as 30% of rural population is poor. 6 Value added in agriculture grew by 2% in real terms between 1996 and During the same period, agricultural employment grew by 16%.

24 On the whole, real GDP growth during has resulted in increases in per capita consumption for almost all households. This is reflected in Figure 2 as a rightward shift of the consumption distribution. Comparing the situation i.n Fall, 1997 to that in Fall, 1996, we find that per capita consumption increased in real terms for all quintiles of the consumption distribution, and in fact, increased the most for the lowest quintile. Per capita consumption for the lowest 20% of the distribution increased by nearly 17 percent in real terms, while real per capita consumption for the top quintile increased by only 2 percent. However, for those households that stayed poor throughout the full year, real per capita consumption did not increase at all. And in fact, the average poverty shortfallfor the poor actually increased over time! (Table 2). Thus, there is a small fraction of the population who appears to be stuck in poverty and getting poorer. These are the chronic or long-term poor. Figure 2. Distribution of the Population by Per Capita Consumption: Comparing First and Last Rounds 45% 40%.. 35%/ 30%0jo 25% 1 20% 15% I 10% 5% Lari pr c4sits -3rd quarter th quarter_f 1997 Source: SDS household data; log scale; kernel procedure to obtain density estimates. 2.2 Chronic versus Transient Poverty 42. Much of the early poverty analysis for the transition economies was based on the assumption that poverty in these countries was both shallow and transitional in nature. It followed naturally that rapid and sustained growth would resolve the poverty problem. However, increasingly the evidence throughout the region points to the emergence of a class of long-term poor, who are not reached by growth. Recent estimates suggest that in Russia, for example, 17 percent of the population is chronically or long-termn poor! Even in successful transition economies such as Poland, there is evidence of chronic poverty: households that are poor and stay poor, or even get poorer over time. The risk is that these chronically poor become an underclass, with distinct characteristics that set them apart form the non-poor, and that doom them and their children to live in poverty. 43. As seen above, there is a fraction of the population in Georgia which is persistently poor, and who are not benefiting from growth. But how large is this fraction? What are their characteristics? Are they different from other poor households who are able to escape poverty over time? To analyze this, we explore the panel dimension of the SDS household survey and examine cohorts of households during a full year. 7 Poverty Transitions in Russia: Moving Up or Bottoming Out in , The World Bank, November, 1998.

25 This exercise reveals that the majority of the poor are poor for only part of the year. At the new poverty line, only about 2% of the population remains poor for the full year of observation. Nearly 80 percent of the poor escape poverty during the year. If we apply the official poverty line rather than the new one, then 18 percent of the population is chronically poor' Alternatively, Table 2. Average Poverty Shortfall for the Poor we may consider the long-term poor to be those households whose -- ;M M.. average consumption throughout Summer, the year was below the poverty line. Fall, About 6%, or close to half of all the Winter, poor, are poor on average. Spring, However, even in this case, the Summer, majority of the poor are transient Fall, rather than chronic poor. 9 Source: SDS Household Survey 45. The fact that a large proportion of poverty is transient indicates that most of the poor are able to lift themselves out of poverty in a relatively short period of time. This is good news: it means that the bulk of the poor are not likely to fall into more structural, and more permanent, forms of poverty. And it suggests that growth can potentially have a very large impact on poverty. On the other hand, it also means that many households which are not poor today face the risk of falling into poverty at some point, which makes targeting assistance much more difficult. It also means that most of the population has had an experience of poverty at some point in time, and that many households find themselves vulnerable to poverty. In the winter months, for example, about 20% of the population that is normally not poor falls into poverty. On average, for all panel waves, we find that one-third of the population has had a spell of poverty over the period of observation (4 quarters). The risk of slipping into poverty is very high for the average Georgian household. Table 3. Characteristics of the Chronically (ne~rline) fli~e 1. Location ot~kw) (new line) Urban 17.7% 2.1% 50% 73% Rural 19.3% 0.9% 50% 27% 2. Household head employment: Inactive or unemployed 29.0% 6.6% 18% 52% Self-employed in agriculture 19.7% 1.6% 26% 25% 3. Household labor profile One or more unemployed 23.5% 3.3% 21% 36% No one employed for full year 35.7% 8.3% 10% 28% All adults employed for full year 13.8% 0.5% 10% 4% Some members lost their job 19.7% 1.6% 56% 57% Total 18.4% 1.5% 100% 100% We call chronically poor households which have been below the poverty line during the whole year of observation. It may not be comparable to the definition of chronic poverty used in other countries where longitudinal data are available. 9 However, some of these households may then fall back into poverty. Since we observe households for only a year, we cannot accurately track these transitions. As such, our results probably overstate the escape rate form poverty - and underestimate the true magnitude of chronic poverty.

26 Can we identify some of the root "causes" of chronic or long-term poverty? As is the case for overall poverty risk, employment status of the household emerges as the strongest correlate of (chronic) poverty. Households where the head is inactive or unemployed face the highest chances of being chronically poor. The number of employed persons in the household and the stability of their employment also matter. The presence of unemployed members, even in households with a working household head, considerably increases the chances of the household being chronically poor. The risk is also high for households with temporary or fluctuating employment (mainly households where employed members suffer spells of non-employment in between jobs). Although average budgetary wages are low, employment in the Government significantly reduces the risk of being chronically poor. This suggests that budgetary employment (and the additional sources of income it may generate) acts as an implicit safety net that prevents families from falling into permanent poverty. Reflecting the importance of inactivity and unemployment as the major correlate of long-term poverty, the majority of the chronic or long-term poor are urban (73%). Nearly one half of the long-term poor live in Imereti (48%); and another one-fifth (21%) in Tbilisi.' Table 4. Chroniic Poverty by Region Kakheti _ 1.0, 0.6% 0.1% 11% 5 / 1 / Tbilisi 14.3% 1.4% 0.6% 18% 21% 20% Shida kartli/mtskheta 12.9% 0.7% 0.0% 7% 5% 1% Qvemo kartli 25.7% 1.5% 0.5% 13% 10% 7% Samtskhe-javakheti 16.2% 0.0% 0.0% 3% 0% 0% Adjara 11.6% 1.0% 0.1% 6% 6% 1% Guria 36.6% 1.0% 0.1% 8% 3% 1% Samegrelo 8.5% 0.5% 0.5% 5% 3% 7% Imereti/racha/Swaneti 28.8% 3.8% 2.5% 29% 48% 63% Total 18.4% 1.5% 0.7% 100% 100% 100% 2.3 What about the Future? Medium-Term Impact of the Level and Patterns of Growth 47. Some simple projections of poverty rates under different growth scenarios reveal that, assuming inequality stays constant, economic growth can have very big impact on poverty (Figure 3). The elasticity of the poverty headcount with respect to growth in average per capita consumption is very high, and fast growth leads to an even faster reduction of poverty rates. Sustained economic growth and stable macroeconomic performance are hence crucial components of any poverty alleviation strategy. 48. The simulations also show, however, that a worsening of the income distribution would undermine any positive impact of growth (Figure 4). If the distribution were to worsen to Brazilian levels, the economy could grow at an impressive 11 percent each year and it would still take almost 10 years to restore the incidence of poverty to present levels! Hence, the Government's ability to tax and redistribute income through expenditure and taxation policies rould play a very significant role in determining poverty outcomes in the future. In this regard, Georgia's currently weak tax enforcement capacity poses a serious concern for the future, not only because of its implications for macroeconomic stability and growth, but also because of its potential impact on the distribution of income. 10 See Technical Paper 1, Volume 11, for details.

27 -16- Figure 3. Georgia: poverty projections with different universal growth rates 45% r{v- { ca c35% c4 5% r _ _... _... _ % 20% _ 15% I0 0/ t 0%~~~~~~~~~~~~~ andwoe in of distrit tio the l of linequ ai Figure 4. Georgia: poverty projections, universal growth of consumption and worsening of distribution to the level of Brazil's inequality 40% % &30% E 25% 5 20% 15% 5% t ~~ 0% Now 0 3% 5% 7% 9% 11% growth growth growth growth growth growth for 5 for 5 for 5 for 5 fo r 5 r -~ years years years years years -u-worsening inequality, new poverty line I Stable inequality, new poverty line 49. The need for debt relief/restructuring. The above simulations show a strong link between per capita consumption growth and poverty incidence. However, the link between overall economic growth and consumption growth in a highly indebted country like Georgia, is not simple. With a heavy debt burden, and tight external financing constraints, positive output growth can easily co-exist with relatively modest, or even negative, consumption growth. Figure

28 -17-5 illustrates what happens to poverty rates with and without restructuring of Georgia's debt service payments. Percent poor Figure 5: Poverty in Georgia with and without external debt restructuring 12% 91% 6% 3% Percent poor - initial level Poverty incidence- base case U Poverty incidence- base case and external debt restructuring 50. Unfortunately, both scenarios offer very limited improvement in poverty rates for the short term. This is because, despite moderate but stable GDP growth rates, projected per capita consumption growth for under both plausible scenarios is very low. The picture improves considerably over the medium-term, particularly under the debt restructuring scenario. A restructuring of Georgia's considerable burden of debt payments (including its repayments to Turkmenistan) allows for a substantial increase in the rate of growth of private consumption visa-vis the base case, with an immediate impact on poverty. Moreover, if the resources released through restructuring were to be targeted in any way (via, for example, a Government transfer to the poor) the impact on poverty would be even larger. 51. Regional Differences. The above scenarios all assume that growth is uniform across the whole economy. In reality, growth rates are likely to differ across sectors and consequently across regions, and poor regions may well lag the average in growth. Carrying out the projections while allowing growth to vary across regions in response to sectoral composition shows exactly that: poor regions such as Imereti or Guria grow more slowly than the average. As a result, existing regional differences in poverty do not disappear with growth (Figure 6). In a country as regionally and ethnically fragmented as Georgia, this could undermine overall social and civil stability. It suggests that dealing with regional poverty may require special "regional" interventions.

29 -18- Figure 6. Poverty by regions: sensitivity to growth (poverty incidence in 10 years from now under different scenarios) 20.0% % r9 3 i +~~~~~~~~~~~~~~~~Kakheti '= 140%.. ^.. ~~~~~~~~~-.- \,Tbilisi ~ 120%..... t... A Shida.. \ kartli... o 10.0% Qx Qvemo 8.0% -- Samckhe javakheti E!D Adjara kartli 0 Guria Samegrelo Now Zero growth Base case - Medium case- High case- -' Imereti RMSM model extrapolation extrapolation forecast of of 1997 trend s urvey Figure 7. Per cen 12% Poverty in Georgia: future paths under different levels of poverty benefit (Base case) Poo r in the 9%. ulat ion 6% F---- 0% P-e-Perfcent poor - initial level Poverty incidence- base case (no targeting of actual (1997) poverty benefit) a Poverty incidence - base case, increase in poverty benefit to the planned 1998 levels and targeting Poverty incidence - base case, targeted poverty benefit gradually increasing to 1% of GDP by The Impact of State Transfers. How could state transfers or other targeted interventions affect this picture? Because of its limited resources, currently the state can play only a small role in poverty alleviation. However, the size of poverty gap is not large (about 2 percent of GDP) a,nd is likely to fall further if growth continues. In this context, even a small, but well-targeted poverty benefit, could play a crucial role in dampening poverty. This is illustrated in Figure 7, and further elaborated in Chapter 5. Further improvements in the state's ability to tax, and hence transfer resources to the most needy through appropriately targeted expenditures, would strengthen the reduction in poverty rates.

30 -19-3 Inequality and Income Distribution' Income inequality in Georgia is very high. The Gini coefficientfor money income in 1996 was a level comparable to the most unequal econornies in Latin America. However, measuring only money incomes in Georgia can be misleading, as there is evidence of a very large discrepancy between reported incomes and actual consumption. Inequality in consumption is less severe than for incomes, with a Gini coefficient of Growth has notyet had a strong impact on consumption inequality per se, but we find evidence that during consumption increased at almost all levels of the distribution. During the same period, there was significant income mobility, exceptfor those at the very bottom or the very top of the income distribution. For the latter, economic success appears to be closely associated with labor market status, ownership ofproductive assets and resulting earnings opportunities. 3.1 Measuring Inequality in Incomes and Consumption 53. Georgia is, in many ways, a highly unequal society. If we examine the distribution of monetary incomes we find evidence of extreme inequities, with the lowest 20 percent of population commanding only about I percent of total income, while the richest 20 percent of individuals account for nearly 60 percent. The Gini coefficient estimated for end-1996 was 0.59, just short of that observed in Brazil, which in 1995 had a Gini for per capita income of Table 1. Distribution of Current Monetary Incomes by Income Quintiles Individuals First quintile (poorest) I Second quintile Third Quintile Fourth quintile Fifth Quintile (richest) Coefficientofvariation Gini coefficient Theil entropy measure Theil mean log deviation measure 54. With growth there has been some improvement in the distribution of income: during 1997 the lowest quintiles experienced an unambiguous increase in their share of total income, while the top quintile lost around 10 percent. Indeed, inequality indices sensitive to different parts of the distribution show that the improvement affected the entire spectrum of incomes, and not just the top and bottom. Nevertheless, inequality in money incomes remains extremely high. 55. Given the degree of informalization and demonetization of the Georgian economy, measuring only reported money incomes may give a misleading picture. Income is traditionally underreported in household surveys, even in established market economies, and Georgia is no exception. If we compare income and consumption data using the SDS survey, we find a very large discrepancy, equivalent to about 40 percent of consumption on average. The gap between consumption and income is lowest at both ends of the distribution (for the poorest and richest This Chapter draws extensively from Inequality and lncome Distribution, Technical Paper 3, Volume II. See the full paper for more detailed exposition and analysis.

31 -20- quintiles), and highest in the middle, where it represents about 45 percent of total consumption (Table 2). Table 2. Average Income and Consumption by Consumption Quintiles Mean income per capita, lari/month Median incone per capita, lari/month Mean consunption per capita, lari/month Median consumnption per capita, lari/month Distribution of income per capita by quintiles, lari/month First consunption quintile (poorest) Second quintile Third quintile Fourth quintile Fifth consumnption quintile (richest) Distribution of consutption per capita by quintiles, lari/month First consumption quintile (poorest) Second quintile Third quintile Fourth quintile Fifth consumption quintile (richest) Discrepancy, income - consunption (lari per capita and percent of consumption on average) First Quintile (poorest) % Second Quintile % Third Quintile % Fourth Quintile % F ifth Quintile (richest) % Source: SDS Household Survey database. Note: due to changes in methodology of collecting data on in-kind income and consumption, data for the last two quarters are not fully comparable to previous quarters. For the fill sample in each round, quintiles are based on total nominal (non-deflated) consumption per capita, including consumption of food in kind; incomes also include consumption in kind. All quarterly data are weighted averages uising survey weights, average over the period of observation is an un-weighted mean for all quarters. 56. Almost all of the observed consumption-income gap is between reported total cash expenditures and reported total cash income. It is not, as one may have thought, related to consumption of products in kind, which are imputed and included in the total reported income measures. Further examining the evolution of the gap between observed cash spending and reported cash income in the panel component of the data suggests that it is linked to actual underreporting of money incomes, and is not a reflection of accumulated cash holdings by households. To understand the patterns of underreporting better, we decompose the consumption-income gap by sectors of employment. We find that underreporting of incomes is positively and closely associated with the degree of informalization in each sector, and with the prevalence of "gray" economic transactions (Table 3). Underreported cash incomes are largest where you would expect: among those employed in restaurants and hotels, for whom unreported tips are an important fraction of income, and among health care workers, who rely on out-ofpocket payments by patients. But underreporting is also large for those employed in the public administration and in public utilities (perhaps reflecting widespread bribes or other petty corruption practices), as well as for those employed in transport activities and in education. Y7. To further explore the discrepancy between consumption and income, we compare the data from the survey with information from the national accounts. Due to limited information on personal incomes, we have to limit our comparison to GDP aggregates. However, the comparison is revealing: average per capita consumption for 1997, as observed in the survey, almost perfectly matches GDP at factor cost per capita; in contrast, the income data from the survey falls short of any GDP-based measure.

32 All of this evidence suggests that consumption provides a much more accurate measure of welfare than income. Moreover, it suggests that any analysis of income, income inequality or income distribution must make some form of adjust:ment for underreporting of incomes. Our approach is simple: for each household, we impute "unrecorded" income as the difference between total money spending and total money revenues whenever the former exceeds the latter. This adjustment affects about 90 percent of households in each round, half of them quite substantially. Table 3. Income, Consumption and Estimated Unreported Monetary Income by Sector of Employment (average per household for survey period) All are unemployed or inactive 279 J41 95 Agriculture Mining Manufacturing Electricity, gas and water Construction Trade Hotels and restaurants Transportation, communications Financial services and real estate Public administration, defense Education Health and social services Other personal services Using imputed rather than reported income data has profound consequences for incomebased measures of inequality. Since most of "unreported" income, as we have seen, is concentrated in the middle of distribution, the adjustment has a strong equalizing effect. This is illustrated in Figure 1. The concentration curve for "unreported" incomes lies above the equality line, suggesting that "hidden" income is a progressive component of total income. The resulting concentration curve for total (including imputed) income is very close to the (shown) concentration curve for consumption. 60. Inequality in consumption is significantly smaller than the estimates of income inequality would suggest. By end-1997, the Gini coefficient Ior per capita consumption was 0.36, and for total household consumption it was 0.38 (Table 4). This is comparable to consumption inequality levels observed in Indonesia or Vietnam, but higher than in most European and FSU countries. Consumption inequality in Georgia is just below the levels observed in most Latin American countries, which have consumption Ginis starting at about During there was a slight improvement in the distribution of consumption, accompanied by an increase in consumption over almost the whole range of the distribution. The improvement in the distribution appears to be driven mostly by losses at the very top (top 5%), acd by developments in the mid-part of the distribution. The improvement, however, is not monotonic, and demonstrates some variability. An issue of concern is that the share of the very bottom of the distribution (bottom 5%) has not changed at all; moreover, for the very bottom of the distribution there has been no real consumption growth during the period of observation. This represents a small, but nevertheless significant, fraction of households who seem to be mired in chronic or long-term poverty, and whose position relative to the median of the distribution is actually deteriorating!

33 -22- Figure 1. Lorenz Curve for Monetary Income and Concentration Curves for Consumption and "Unreported" Monetary Incomes 100% ~~XCumulative percent I /// 1 of monetary incomre 75%. Art_,/ X [ - -_Cumulative /_ 50% / ; / /. 7,{ of consumption precent Cumulative percent 75% -/ Z / i of "unreported" 25% i,ncome 0% 0% 25% 50% pl fi 75% 100% _Equality line Source: SDS household survey data, pooled for all quarters of the survey, using only first interviews Table 4. Distribution of consumption by consumption quintiles and inequality 7 ~.' ~ ~ ~ ~ r.,,,? 7r ' n 9 Fa1l97 Individuals (percent to total) Bottom 5 percent I I I I I First Quintile (poorest) Second Quintile Third Quintile Fourth Quintile Fifth Quintile (richest) Top 5 percent Inequality Coefficient ofvariation Giini coefficient 'Theil entropy measure 'Fheil mean log deviation measure Mobility 62. We can look more closely at the dynamic properties of the distribution by examining our three panel waves of households that were surveyed continuously for a full year. This reveals a picture of significant income mobility, except for those at the very bottom and very top Table 5 gives us a picture of mobility across consumption quintiles in one year. It shows that for a typical panel wave, only percent of households starting off in quintiles 2-4 are in the same quintile by the end of the year. For these quintiles, mobility is not limited to moving to a neighboring quintile, but can involve rather large shifts. However, there is much less mobility at the very top, and especially at the bottom. About 52 percent of households that started off in

34 -23- the lowest quintile are in the same position one year later. Similarly, very few of those who started off at the top fall below the median within a year. What explains mobility? Do the persistently rich have different characteristics from the poor? Can this help us draw a picture of "winners" and "losers" during the transition? Table 5. Transitions between consumption quintiles (second wave of the panel) Final position (last round, 4 quarters later) ~~ tii~~o~~st) 52% 19% 13% 8% 7% 100%,t~1e 26% 28% 22% 13% 10% 100% Startingposition, 1st round 19% 25% 30% 15% 11% 100% le 11% 12% 21% 30% 26% 100%1 5tb~~' (~l~ 4% 6% 20% 25% 44% 100% Note: Quintiles are based on per capita consumption, sample restricted to panel households 64. If we examine households that remain in the top quintile for the full year, we find that they tend to be located in richer regions, and predominantly in urban areas. For example, in urban Adjaria, as many as 31 percent of all households remained in the upper quintile for the full year. As a result, 12 percent of the better-off in Georgia reside in urban Adjaria (even though its share in the total population is less than 5 percent). In Samegrelo, about 12 percent of households are in this group of persistently "rich" (and found in both urban and rural areas). Contrary to what we may expect, the rich are not exclusively concentrated in Tbilisi: only some 20 percent of all persistently rich households live there. 65. Most successful households have a head that is self-employed outside agriculture, with individual entrepreneurs, employers or collective entrepreneurs (cooperative members) representing the bulk of the "rich". About 35 percent of all "rich" head of households are selfemployed outside agriculture; and 48 percent of all entrepreneurs remain in the top quintile for the full year. Among wage earners, Government employees represent the largest share of households that managed to stay in the top quintile for the full year. Interestingly, the share of "rich" households where the head is employed by the Government is higher than their share in the population, indicating that budgetary employment can be relatively lucrative. Those employed in private firms have an even higher chance to be among the "winners" but given the small size of the new private sector, account for only a relatively small share of the rich (about 5 percent). 66. There is a clear positive association between education and being in the top quintile on a sustained basis: the share of those with higher education among the persistently "rich" is 34 percent, far above their share in the population. In rural areas, successful households are associated with larger land holdings and more diversified crops (including wheat and corn), which yield higher money incomes. Incomes from farming for "rich" rural households average some 145 lari per month, four times higher than for a typical rural household. 67. Contrary to what we may have expected, the income structure of "rich" is not dominated by "unreported" income; in fact, its share for the persistently rich (25 percent) is lower than for the population on average. Economic success appears to be closely associated with a greater "visibility" of employment and income and higher reliance on formal, rather than informal, sources. 68. The analysis of households that are persist:ently at the top of the distribution reinforces the conclusions of the poverty profile: labor market status, ownership of productive assets and

35 -24- resulting earning opportunities are the key determinants of households ' absolute and relative position in the income distribution 3.3 Explaining Inequality 69 There are two ways to analyze the factors driving inequality in Georgia. One is to examine inequalities between different groups of the population, decomposing the contributions of the between and within factors. A second is to look at income sources, and analyze which contributes the most to increasing inequality. 70. Decomposition of inequality by groups. Previous analysis suggests that the most important determinant of living standards and income levels is employment. Hence we begin by decomposing inequality (in consumption) by sector of employment. These results are presented in 'rable 6, which shows average per capita consumption for each group, the extent of inequality by group, and the contribution of each group to total inequality. It also shows what share of total inequality is explained by differences in means between groups. Only I percent of the overall inequality can be attributed to differences between sectors of employment; most of the inequality is wvithin sectors. Self employment, whether in agriculture or outside, is the biggest contributor to overall inequality, accounting for over one half of the total. Table 6. Inequality in Consumption by Sector of Employment (4th quarter 1997) Inactive or unemployed SelIf-employed in agriculture Employed in budget sphere Employed in SOE Employed in private firn Self-employed outside agriculture Between sectors-> I Total [ The increase in the dispersion of earnings was expected to be an integral part of transition, as returns to education and skills moved to align with market forces. There are signs that education is indeed playing an increasing role in determining wage differentials (see Chapter 4), so we may expect education to be a contributing factor to increased inequality. However, differentials between education groups capture only about 2 percent of overall inequality (Table 7). Inequality within education groups, and primarily among those with secondary education, is a much more important determinant of inequality. In this sense, increased returns to skills are been played out as rising dispersion within education groups, where an intensive process of sorting '*vinners" from "losers" is going on. 72. Next we turn to examining intra-regional and inter-regional inequality (Table 8). We find that, again, variance within regions contributes more to overall inequality than differences across regions. Inter-regional differences account for only 3 percent of total inequality (up to 6 percent if we treat urban and rural areas in each region separately).

36 -25;- Table 7. Inequality by Education of.household Head (average for survey period) No one is employed % Less than elementary % Elementary % Uncompleted secondary % Secondary % Post Secondary % University and higher % Between education levels 2% Total % Table 8. Inequality in Consumption by Regions ( 4 th quarter 1997) Kakheti % l Tbilisi % l Shida kartli % Qvemo kartli % Samckhe-javakheti % Adjara % Guria % Samegrelo % Imereti % Between regions: 3% Separating urban/rural: 6% Total % 73. Decomposition of inequality by sources of income. The structure of incomes has changed dramatically from pre-transition times. The most pronounced change has been the collapse in the wage share, from about 60 percent pre-transition to less than a third of total monetary incomes by In contrast, entrepreneurial income and self-employment income have increased in importance, as has in-kind farm income (imputed income from farm products consumed in-kind). The top panel of Table 9 presents a snapshot of this income structure. 74. Several features stand out. First, in-kind farm income is the most important component of income for rural areas, far exceeding the importance of monetary farm incomes. Surprisingly, this is true for all quintiles; in fact, the share of in-kind income for the top quintile is larger than for the lower quintiles. Second, "unreported" income is important in both rural and urban areas, but particularly so in the latter, where it accounts for 40 percent of total income. Third, transfers (both State and private) play an important role for the poorest quintile, but somewhat differently in urban versus rural areas. The rural poor have much lower private transfers than the urban poor.

37 The bottom panel of Table 9 examines how each source of income is distributed between quintiles. By comparing the shares of quintiles for each source of income with the distribution of total income between the quintiles we can tell whether any particular type of income contributes to increasing overall income inequality or to decreasing it. In urban areas, the most unequally distributed components of income are asset and financial income and farm income, but these represent a relatively small part of total urban incomes. The main factors driving inequality in urban areas are actually monetary reported earnings (primarily wages and self-employment) and "unreported" income. In rural areas, the main determinants of inequality are income from farrning and in-kind consumption of agricultural products, both of which are very unequally distributed. The only progressive components of income for both rural and urban areas are state transfers (mainly pensions); but they represent only a miniscule percent of total incomes on average. 76. All monetary reported earnings are quite unequally distributed with respect to total income. The concentration coefficient for wages is 0.39, for self-employment income it is 0.49, and for farm monetary incomes it is Wage inequality reflects primarily differential returns to skill and human capital, but also other non-negligible factors such as inter-sectoral differences in hours worked and productivity, industry-specific wage differentials, and the prevalence of arrears. Surprisingly, regional factors play a relatively minor role (see Chapter 4 for a more detailed analysis). The determinants of inequality in earnings from self-employment may be analyzed in a similar way. In this case, however, inter-regional differences appear to be very important, accounting for nearly one half of total inequality in self-employment earnings. Presumably they reflect different underlying opportunities for business start-ups in different regions. There are also large differences by sector of employment, with construction yielding the largest self-employment incomes. And there is evidence of a large gender gap (as for wages) of about 35 percent, after controlling for job and individual characteristics. 77. Farm income is one of the most unequally distributed sources of income in both urban and rural areas. The concentration coefficient for total income from land (in-kind and farm income) is a very high This, together with its large share in total income, makes it the largest single contributor to overall inequality. Inequality in farm income, however, does not appear to be linked purely to patterns of land ownership. If we concentrate only on relatively large farmers (over 1/2 ha of land), we find the same picture of persistent inequalities in income from land than for the population as a whole (the own Gini coefficient for land income among relatively large farriers is as high as 0.55). At the same time we do not find big differences between rural households in the size of land holdings. The main factors driving inequality in farm incomes are differences in production patterns and in access to information and capital (see Chapter 4). 2 -The concentration coefficient of an income component measures how evenly or unevenly that component is distributed over total household income; negative values tell us that it is progressive (i.e. reduces inequality); a value of 0 tells us that it is neutral; and a positive value suggests that it is inequality increasing. An extreme value of I would reflect that the richest household gets all the income. For details of decomposition see technical annex 4.1.

38 -27- Table 9. Structure of Income by Consumption Quintiles (percentages to total) First quintile (poorest) Second quintile Third Quintile Fourth quintile Fifth Quintile (richest) ALL URBANT J 00g.0 M2H <g: i 0Q00 $S0UU:ML:. 0 0 = M 0 S0-3^ First quintile (poorest) Second quintile Third Quintile Fourth quintile Fifth Quintile (richest) ALLRURAL Distribution of Income Sources by Consumption Quintiles (percentages to total) First quintile (poorest) Second quintile ThirdQuintile Fourth quintile Fifth Quintile (richest) TOTAL URBAN First quintile (poorest) Second quintile ThirdQuintile Fourth quintile Fifth Quintile (richest) TOTAL RURAI * For non-agricultural activities only

39 -28-4 The Role of the Labor Market' The labor market is the main channel through which growth affects poverty. Growth reduces poverty through rising employment, increased labor productivity, and higher real wages. In Georgia, the labor market has shown outstandingflexibility during a period of severe political and economic turmoil. Thisflexibility has been achieved mainly through the informalization of employment, and through the reallocation of labor towards small-scale agriculture. Informalization has dampened the impact of the crisis and served to protect the poor. However, it remains a short-term response. 7oday, a large and growing fraction of the Georgian labor force relies on self-employment as the primary means to earn an income. For some, this is an avenue for earnings mobility and growth, for the majority, however, self-employment remains constrained to low-productivity agricultural or trading activities, with little earnings stability and little potentialfor long term earnings growth. Prospects for thefuture hinge critically on the economy's ability to generate new private employment, and to reallocate labor awayfrom these low-productivity activities into higher value added sectors. 4.1 Labor Market Status and Poverty 78. In Georgia, poor and in non-poor alike derive most of their income (up to 80 percent), from labor market sources, either from wages or from self-employment earnings. As a result, the risk of poverty at any moment is closely associated with the extent to which a household participates in the labor market, and with the way in which the market remunerates its labor. Changes in a household's poverty status over time are tightly linked to changes in the labor market status of its members. This is reflected in strong differences between the poor and non-poor in participation patterns, unemployment, sector of employment and remuneration. 79. The poor are more likely than the non-poor to be inactive or unemployed. Labor force participation rates for the poor are below those for the non-poor at all ages (Figure 1). Similarly, unemployment rates are higher among the poor for all age groups. These differences are fourd for both men and women, and if anything, are even more pronounced for the latter. 80. Among those who are working, the poor differ from the non-poor primarily in that they have more unstable employment-e.g. are more likely to hold casual or temporary jobs, be on involuntary leave or work part time. As a. result, the poor work less hours, on average, than the non-poor (36 versus 40 per week). About 66 percent of the working poor are self-employed in agriculture, as compared to 50 percent of the non-poor. Outside of agriculture, they tend to be concentrated in sectors that pay lower wages, such as education or domestic help. Interestingly, the poor are much less likely than the non-poor to be self-employed outside of agriculture, perhaps reflecting that they lack the access to even the small amount of capital (or social networks) needed to establish a small business or activity on their own. The poor are also more likely to be owed backwages: on average 22 percent of poor wage earners were owed wages by thei.r employers, while only 8 percent of the non-poor reported delayed or irregular wage payments. Hence, wage arrears may be one factor driving families into poverty. 81. Not surprisingly, there is a large wage gap between the poor and the non poor: on average, the poor earn only 52 percent of the wages of the non-poor. In part, this differential is explained by endowments: about one-third of the wage gap is explained by differences in education, age, potential experience, gender, sector of employment and location. But two-thirds of the gap is unexplained, and probably reflects the concentration of the poor into less desirable, lower productivity occupations within sectors. 1 This Chapter draws extensively from Labor Markets, inequality and Poverty, Technical Paper 2. Volume II. A more detailed presentation and discussion of the results can be found there.

40 -29- Figure 1. Labor Force Participation, Employment and Unemployment by Age Non-poor in labor force Poor in labor force ~~~~~~~~~~~~~~~~ a UnempJoyed Source: SDS household survey MEmployed EDOut of labor force U Unemployed EsEmployed D Out of labor force 4.2 Output and Employment Trends 82. Following independence Georgia experienced a sharp collapse in output, followed by an incipient recovery from 1995 onwards. How did these developments play themselves out in the labor market? Figure 2, which plots cumulative changes in output and employment, gives us an aggregate picture. As in much of the FSU, changes in employment appear to have greatly lagged those in output, leading to a sharp drop in aggregate productivity per worker. This was mirrored by a sharp drop in real wages which by 1995 had fallen to about one-tenth of their pre-transition levels. An additional price adjustment mechanism used by enterprises to counter output shocks was the delay of wage payments to workers. According to the SDS survey, 23 percent of all workers who were surveyed during a full year reported being subject to wage arrears at least once during that period. The incidence of wage arrears is highest in education, construction, mining and industry, where 30 percent of all workers rerport suffering from delayed wage payments. However, in contrast with what is observed in Russia and other FSU countries where workers suffer from chronic or persistent wage arrears, in Georgia the majority of workers report suffering only one episode of delayed wage payments during the year. 83. Aggregate employment figures are somewhat misleading since they hide a significant adjustment in hours worked: in 1996, about 20 percent of all wage employees reported working less than 15 hours per week. Hence, there has been a significant adjustment in total labor input. Aggregate figures also hide a significant reallocation of employment across sectors (from industry into agriculture, trade and services), and between types of employment (from salaried jobs into self-employment). Employment in industry has fallen sharply, and almost matched the decline in output (both have fallen by nearly 80 percent). In contrast, employment in agriculture has expanded rapidly, and almost doubled relative to its 1990 level (even though total agricultural production by end-1997 stood at about 70 percent of its pre-transition level!). This has occurred in a context of radical institutional and ownership changes in Georgian agriculture, where the dominant production form has become small, private (about hectare) plots. Due to the breakdown of industrial and trade links with other FSU countries, however, there has been a sharp reduction in the use of fertilizers, tractors and other capital equipment, with a consequent negative effects on agricultural productivity. By 1997, value added per worker in Georgian agriculture was only 100 lari per month, much lower than in any other sector of employment.

41 Figure 2: Trends in EMDloyment and OUtpt g60\ l I H--GDP -Total employment Source: IMF; SDS data on employment before Note: Employment series show a break because of the change in methodology. 8L. The reallocation of employment between sectors is further illustrated by Table 1, which presents changes in GDP and employment by sector for (computed from the SDS survey data). Agriculture is the dominant employer, and reported the highest employment growth during Most of this growth has taken the form of a rise in self-employment, and within self employment, as an increase in the number of unpaid family workers. Trade and transport are the other growing sectors. Despite a recorded increase in industrial output, enmployment in industry continued to fall during , reflecting that there was still significant hcarded labor. Table 1. Employment and Output Growth by Sector, Agriculture ,210 Industry Construction Trmnsport and communications Trade and catering Other Total ,233 Source: Official SDS publication, SDS household survey for employment by branch. Note: GDP at factor costs. 85. There has been a continuing reallocation of labor between different types of employment. Most net employment growth has taken the form of self-employment. In contrast, wage employment has shrunk, both in absolute levels and as a share of total employment (Table 2). This reflects a trend towards growing informalization of the labor market. This trend is visible even if we look only at urban markets. Between Summer, 1996 and Fall, 1997 (first and last survey rounds), the share of self-employment in total urban employment grew from just over 20 percent to over 40 percent. The largest and most rapidly sector for self-employment (even in urban areas!) was agriculture, followed by trade (which by end-1997 accounted for about 40

42 -31- percent of all self-employed in urban areas). 2 Together, trade, agriculture and transportation cover 90 percent of all self-employment in urban areas. In addition to those who have selfemployment as a sole and primary occupation, 20 percent of wage employees have to supplement their meager wages with some kind of self employment. In most cases this is subsistence gardening. Table 2. Employment by Type, Wage employment State sector Cooperatives Private firms Self-employment 1,324 1, Total 2,036 2, Source: Official SDS publications, IMF(1998); employment classification is according to primary employment 86. In contrast to the dynamic growth in urban self-employment, total wage employment in urban areas has remained stagnant. Wage employment in new private firms actually expanded by nearly a third between mid-1996 and end-1997, but this growth did not compensate for the fall in employment in Government, SOE and privatized companies. Among the latter, cuts were particularly concentrated in education, electricity, gas and water supply. New private wage employment is most important in Tbilisi, Adjaria and Samegrelo. Jointly, these three regions account for nearly 60 percent of all private wage employment in Georgia; in contrast, the two poorest regions, Imereti and Guria account for only 14 percent. 87. As a result of the sluggishness in the contraction of employment, unemployment is relatively low. Official unemployment, based on the number of registered jobseekers, has fluctuated between 2.5 and II percent of the labor force. However, these data are deceptive, since there are few incentives or reasons for the unemployed to become registered' 88. A much better picture of unemployment can be obtained from the SDS household surveys (Figure 3). According to these data, in mid-1996 unemployment stood atl2.8 percent of the labor force (based on standard ILO/OECD criteria for defining the unemployed). Since then unemployment has fallen rapidly to less than 5 percent of labor force. Using a broad or "soft" definition of unemployment, which includes discouraged job seekers, yields a similar picture, with total unemployment dropping from about 20 percent to 8 percent of the labor force during The SDS survey also reveals that unemployment is becoming a phenomenon that affects very specific groups of the population: especially new entrants into the labor market, and women more than men. 2 Some of the expansion in trade represents subcontracting by large wholesalers who economize in this way on various costs, including labor (see Technical Paper 4, Volume II). 3Benefits are low: in 1996, about 30 percent of the average national; by end 1997, about 20 percent. And are paid for a short period of time (6 months before 1998, 12 months now).

43 -32- Figure 3. Unemployment Rate (survey-based) 25 _ 20 %15. it ~~~~~~~~~~~~ Summer Fall 96 Winter Spring Summer Fall LFS unemployment rate - "Soft" unemployment rate 89. Regional differences in unemployment are large, with Tbilisi showing the most persistently high unemployment rate (Table 3). Initial conditions appear to be a crucial determinant of regional urban labor market performance. The prevalence of large industrial and mining enterprises in Imereti, and of heavy industry in Tbilisi and Shida Kartli can help explain the high unemployment levels observed there. Similarly, urban Guria's dependence on tea processing can heip explain its relatively poor performance. Table 3. Urban unemployment rate by regions (percent to labor force) -. >>. S Eunw. : W- s97 ;,_ S97 ig Snmner97 Fall 97 Kakheti Tbilisi Shidakartli Qvemo kartli Samtskhe-javakheti Adjara Guria Samegrelo Imereti Total urban areas Source: SDS household survey. As the additional cleaning of primary records has been undertaken to remove double-counting in some cases, the numbers may slightly differ from official publications by SDS. 90. Matching these regional unemployment trends with observed developments in employment, we find that most of the unemployment reduction in urban areas has been absorbed by t:he expansion of self-employment; primarily (and quite paradoxically) in agriculture, and only secondarily in trade and other activities. However, self-employment is by now so dominant, that it is difficult to expect it to grow equally rapidly in the future. As employment in Government and in SOEs are expected to continue to shrink, prospects for continued reductions in unemployment hinge critically on the expansion of formal and informal private employment. Efforts to attract private investment (mainly foreign), and to stimulate small business development are crucial in this regard. 4.3 What Has Happened to Earnings? 91. Following a drastic decline between 1990 and 1995, wages increased rapidly, and practically doubled between 1995 and 1997 in real terms. There are significant wage differentials between Government, SOEs and private firms, with much of the growth in average wages driven

44 -33- by developments in the private sector (Figure 4). Controlling for differences in workers' endowments, we estimate that private firms are paying a premium of 20 percent over the government sector, SOE and privatized sector wages. However, take-home wage incomes have been fairly volatile in all sectors, reflecting the changing incidence of arrears in the State sector, and episodic windfall rents in the private sector. 92. Rising average real wages have been accompanied by widening wage differentials and earnings dispersion. The Gini coefficient for monthly (non-zero) wages reported in SDS household survey was 0.5 in , and a bit higher (0.53) for hourly wages. This is comparable to what was observed in Russia in 19515, and in sharp contrast with the highly compressed wage structure that existed pre-transition. Similarly, the decile ratio of wages (equal to the 90th percentile divided by 10th percentile) for full time workers in Georgia increased from about 3.35 in 1989 to 10 in 1997-tripling in a veiy short time period. However, Georgia's experience in this regard is not completely out of line with that observed in other transition economies. For example, the decile ratio for full time workers in Hungary, Poland and Czech Republic in 1995 was in the range of 3-4; in Bulgaria around 6; in Russia between 7 and I I(official versus household survey data). And if we look at annual wage income, we find that dispersion in Georgia is very similar to that observed in much of Eastern Europe, and actually lower than in most countries of the FSU, including the Baltics. Figure 4. Real Monthly Wage, Lari, A ugust prices /96 2/97 6/97 10/97 8/96 12/96 4/97 8/97 12/97-4--State real wage r Private real wage - O---- SOE real wage Source: SDS household survey; last month toial pay including in-kind wages; full-time employees only. 93. A large part of observed wage dispersion can be attributed to differences in workers' endowments, sector of employment and location. About 55 percent of the observed inequality can be explained in this way. Private returns to education are strong and significant, especially for secondary education, which commands a premnium of about 30 percent over incomplete secondary. In contrast, the earnings differential between completed secondary and higher education is modest, barely 15 percent. Employrnent in the budgetary sector, agriculture or the social sectors are all associated with lower wages, as is living in rural areas. There are also significant differences in pay between men and women, with the latter earning percent less than comparable male workers (controlling for individual characteristics, location, industry but not occupation). This male-female differential may reflect the occupational segregation of women into lower paid jobs, outright discrimination in pay, or the existence of very large costs associated with women's more intermittent labor force attachment.

45 Rural Labor Markets: Low Productivity and Low Incomes 94. The rural economy has played a crucial role as a safety net during the crisis years. Building on the almost universal access of the rural population to land, it has been able to absorb a huge inflow of labor released from other sectors. There is practically no unemployment in rural areas and a high employment rate. 95. The down side of the rural economy's remarkable flexibility has been a decline in p:roductivity, and consequently weak growth in rural incomes. Real earnings in rural areas have remained practically stagnant during the recent high growth period ( ), despite a record grain harvest and favorable weather conditions. The poor performance of rural incomes can be attributed, to different degrees, to unequal access to inputs complementary to labor (fertilizer, tractors and capital equipment); barriers to land transactions and consolidation of holdings; lack of market access and information; scarcity of rural credit; and limited off-farm earnings opportunities. 96. Land market. Georgia's land reform created a fairly equal distribution of land. About 60 percent of all arable land is presently in private hands, with an average land holding of about 0.75 hectares. Only some 8 percent of agricultural households report that they do not own or cultivate land. There are signs, however, of increasing concentration of landholdings. As of mid 1996, the richest quintile of the rural population owned 23 percent of cultivated land, while the bottom quintile had 18 percent. By end-1997, the share of the richest had increased to 25 percent, while that of the poorest had eroded to 16 percent. Such noticeable changes in distribution of land in a very short period of time raise concerns over the role of power and local connections in securing access to land; a concern mirrored in the skepticism over the fairness of land reform expressed by small farmers during the qualitative assessment Although access to land per se is not a determining factor of poverty, there is a clear link between the degree or extent of poverty and the size of a household's land holdings (Chapters I and 3). Poorer rural households are those with smaller holdings and less diversified crops. Conversely, larger land holdings and more diversified crops are associated with higher money incomes, and higher living standards. In this context, the incipient trend towards increasing concentration could have an adverse effect on poverty. 98. Interestingly, although the richest rural households are those with larger land holdings, small farmers may be actually more efficient than larger ones. Plotting average agricultural income per hectare against land holding size reveals a clear inverse relationship (Figure 5). Hence, increasing the land under cultivation by smaller farmers, through sales, lease or rental, could bring about significant efficiency gains and increase rural incomes overall. To date, the biggest obstacle to this consolidation has been the absence of a clear legal framework regulating land transactions. The recent adoption of far-reaching legislation on private ownership, titling and registration, and leasing, should go some ways towards building such a legal basis. What is sti.ll needed, however, is the development of the accompanying institutions and mechanisms necessary for the efficient functioning of the land market, starting with complete cadasters for all rural districts in Georgia. 4See Dudwick, Technical Paper 4, Volume II of the study.

46 -35.- Figure 5. Land Holding and Productivity 99. Market access. As evidence by the high share of rural incomes that is, 9;%b«ijt consumed in-kind, the rural CL "Z;s economy is still E 8, I4lEitdl!l predominantly non-market. The breakdown of E previously-existing a 4 marketing and trading arrangements, physical isolation, difficult terrain and , '",,, severely deteriorated infrastructure are all Land holding and productivity i,nrural Georgia contributing factors. While most rural households appear to have some difficulty in accessing markets, this constraint appears to be particularly serious for the poor. Poor households are more likely to be physically isolated, and less likely to exchange produce or animals in the local market than non-poor ones. Among rural households with land that were interviewed four consecutive times during , some 73 percent reported at least one sale of their produce; 27 percent did not report a single sale (reflecting the general low level of market development in rural areas). Among poor rural households, the fraction that did not sell any of their agricultural produce on the market during the whole year was as high as 45 percent Access to rural markets also implies access to information, which helps farmers use their assets, both land and labor, more productively. This includes access to technical assistance services provided by private or public agencies (agricultural extension). One of the reasons for low productivity of rural employment is the lack of adequate technical information on advanced cultivation methods. For example, despite boom ing cereal production, some farmers did not enjoy a rise in income, since they were planting the wheat varieties deemed not of the standard required by the bakeries, which resulted in lower prices. When the technical advice comes handin-hand with extending the credit to farmers, it leads to sustainable improvements in productivity Rural credit. The recent success of an exte nsion and rural credit project to help farmers growing cereals (TACIS-RARP project) has proven the high potential of this kind of intervention to increase the efficiency of the rural sector while benefiting the poor. In 1997, the project channeled about 27 million lari to the farmers, traders and millers in the grain sector. The highest harvest of cereals in recent years was achieved manly through rising yields rather than increasing harvest areas. The encouraging trend is the high incidence of small-scale credits, particularly to farmers' credit unions. This effort is supported also by various international organizations, including the World Bank, which is implementing targeted intervention to help small farmers Limited off-farm eamings opportunities. Rural, off-farm employment plays only a limited role in supplementing agricultural incomes. The main off-farm activities in rural Georgia are not htghly lucrative. The main employers are education (with 25 percent of all off-farm wage employment in rural areas), trade, health care and public administration. But even Government employees in rural areas rely on cultivating their own land plots to provide for food Despite their low level, off-farm earnings seem to play an important role in lifting households out of poverty. While only 22 percent of poor rural households in Georgia receive

47 -36- any wages from off-farm activities, as much as 33 percent of non-poor households supplement their farming with wage earnings. A higher share, and higher incomes from self-employment on of-'f-farm activities in rural areas are observed precisely in the areas with the lowest rural poverty incidence. For example, in Samegrelo, while earnings from agriculture are well below the national average, earnings from non-agricultural activities are 50 percent higher than on average, contributing greatly to the fact that this region has the lowest incidence of rural poverty. On the contrary, Imereti has the lowest level of earnings per self-employed in off-farm activities, despite being close to the national average in earnings from agriculture; it is one of the regions most affected by rural poverty.

48 -37-5 The Safelty Net' The collapse offiscal revenues in Georgia all but destroyed the pre-existing formal safety net, which was reduced to a bare minimum. An extensive informal safety net has emerged in its place: migration, petty trade, family support networks atnd clans constitute its main pillars. However, the coverage of this informal safety net isfarfrom being universal, and many families slip through the cracks. An exception are IDPs from internal co iflicts, who have benefited much more than any other group from state programs and have received a large share of all humanitarian assistance provided to Georgia. The main challenge for the formal safety net for the near future will continue to be a lack offiscal resources. Despite these constraints, however, the formal safety net has an important role to play in poverty alleviation. Given the small overall poverty gap, even a small but well targeted poverty benefit can have a significant impact on poverty. 5.1 The Importance of the Informal Safety Net 104. Under central planning most Georgians were provided with a universal array of social benefits, comprising widespread subsidies on goods and services, nominally free health and educational services, quasi-permanent employment, and a number of other social entitlements. The collapse of the planned economy brought witli it an end to most subsidies and entitlements, and a drastic curtailment of state support for health and education services. Traditional formal forms of social protection were reduced to a bare minimum, and a myriad of informal mechanisms or coping strategies sprung up in their place. Today in Georgia, as in many neighboring FSU republics, informal coping mechanisms-migration, petty trade, subsistence gardening -are essential elements of the safety net. In sheer size, private (informal) monetary transfers are at least as large as formal transfers (Table 1). And this does not take into account the importance of non-monetary mechanisms such as family connections in obtaining jobs or access to services. Table 1. Georgia: Total Safety Net Expenditures in 1997 All formal safety net monetary transfers of which pensions* Informal (private) monetary safety net transfers of which remittancesfrom abroad All in-kind transfers** of which State*** Total estimated safety net expenditures (state and private) Note: data on formal safety nets are based on the household survey: official estimate of total budgetary monetary transfers is 155 mln. Iari. * Data on actual payment of benefits excluding the transfer for eleciricity payments, estimate based on survey data gives 103 mln. lari of actually received pensions in 1997 ** The sum of private in-kind transfers based on the survey (gifts in-kind received, including related to rituals) plus the electricity tariff rebate *** Estimate of electricity tariff rebate financed from various social budgets in The informalization of income sources and social protection mechanisms has affected all Georgian households, regardless of income level. For the poorest quintile of the distribution, income from subsistence agriculture, self-employment, sale of assets and private transfers jointly I This Chapter is based on background materials by Alexander Marc, and on the paper by Nora Dudwick, Georgia: A Qualitative Study of Impoverishment and Coping Strategies, Technical Paper 4, Volume 11 of this study.

49 -38- represent well over one half of all reported incomes, and probably a significant fraction of "unreported" incomes as well (see Chapter 3, Table 9). However, earnings from selfernployment activities and subsistence agriculture also constitute an important fraction of the incomes of the non-poor (at least 30-40% of reported incomes): in other words, these elements of the informal safety net are core subsistence strategies for poor and non-poor alike! Relative to other quintiles, the poorest quintile rely more on sale of assets (distress sales) and private transfers; less so on self-employment earnings and wages. The poor also rely more on formal transfers from the state; but the latter are presently too small to have a large impact on living standards. The activities covered by this informal safety net vary, but the following forms are the more prevalent: 106. Small trade and services is one of the most important, and successful, coping mechanisms. According to the qualitative study, many Georgians equate the market economy with the right to engage in petty trade. This trade, typically involving domestic or intemational buying and selling on a small scale, provides many people with incomes ranging from subsistence levels to substantial. Interestingly, women have come to play an important role in trade, even when it involves behavior once considered unseemly for women, such as traveling abroad by themselves and absenting themselves from their families. Cross border trade seems to be the most profitable, and in many cases is done illegally. In~~00ia I.orml sw.ai trade in Jgavakheti U-a> tiade pwe wit e 107. Although petty trade is very common, not all households are able to engage in it and make a living. In fact (as discussed in Chapter 4), poor families are less toiafie.vati&k b likely to be involved in such trade 3 - i se os':11 than non-poor ones. One problem b :d Ithi di6t t for poor families is the need for -,-y -e.i. start up capital. If a trader does not seflialthopoftentmwiiihnig have such capital (obtained usually,tol yt~ciila IfS~theY indicate that necessary through the selling of assets or by '*star v.erapes leas at U$$FOO t borrowing from family members), -hkdail~proflt~mnaycome -- -:S4... m he/she has to resort to borrowing on l, f A W S f^e and Coping the informal market, but often at very high interest rates. Another L. barrier to engaging in petty trade, which is particularly severe for the poor, are the many bribes and informal payments required "to stay in business" Sale of assets. Selling household assets and durables, including jewelry, clothing, household items, linens, and fumiture, continues to be used as a strategy to "tide" many households over during difficult periods. It seems to be particularly important for the lowest quintile of the distribution, where it represents a relatively large proportion of household income (nearly 10 percent). As in other FSU republics, sale of assets is also emerging as a strategy to smooth income over time, through which households purchase assets when they have money and sell them when they experience a drop in income, due to seasonal fluctuations, wage arrears, or other unforeseen events. The relative importance of asset sales in the informal safety net is a reflection of the malfunctioning of other (formal and informal) insurance mechanisms Labor migration. Relatively little information exists on migration and the ensuing important flow of remittances. However, many respondents in the qualitative study mentioned receiving gifts, money or clothing from relatives working abroad. And in the SDS survey,

50 -39- remittances account for nearly 2 percent of total reported household income for the average Georgian family, and for over 6 percent of the income of households in the lowest quintile. In 1997, total remittances form abroad (as inferred from the SDS survey) amounted to about 42 million lair or 0.6% of GDP. Estimates of people who have left Georgia to work abroad since 1991 vary from about 300,000 to 800,000. The bulk of permanent migration took place during the early years following independence. These migrants were mostly either ethnic minorities who left the country definitively or migrants with contacts/networks abroad, who kept family members in Georgia. Most migrants left for Russia, with Greece as a second important destination (especially for residents of Batumi and Tbilisi). Current migration flows are of a more temporary nature, and relate purely to work opportunities abroad. Labor migration has become extraordinarily important in South Ossetia, which experienced high unemployment even during the Soviet period Subsistence gardening/agriculture. During the years of crisis subsistence agriculture became the main receptacle for surplus labor, and the core of the informal safety net in both rural and urban areas. Still today, incomes from agricultural self-employment and subsistence gardening constitute over one-half of the incomes of all the population, and over 70 percent of incomes in rural areas. Even in urban centers, agriculture has become the largest (and fastest) growing sector of employment-much of it made tip of unpaid family workers laboring in small, family-held plots. In Imereti, for example, agricultural self-employment accounted for 60 percent of all urban employment in 1997; in Guria and Kakheti, it accounted for 55 percent! For urban dwellers, small plots in the urban periphery have become a core part of the safety net; as a result, a large part of intra-family (and intra-clan) transfers comprise agricultural products in kind. I 1. Family transfers and support. Intra-household transfers (within family or within clan) represent a sizable proportion of household incomes, especially in urban areas. Monetary family transfers account for as much as 10 percent of total reported incomes of the urban poor. This figure is even larger if one includes transfers in-kind. But family support goes beyond transfers: family or clan connections (see below) are key in securing access to jobs or services. According to the qualitative survey, the most important strategy for finding either private or public employment is to use one's "connections", or to pay a bribe. In the words of a young Georgian from Batumi, "to find good work, without connections-this is fantasy". Family has also played a role through changing household composition: the regrouping of previously nuclear households under a single roof has served as a natural coping mechanism, that takes advantage of the existence of significant household economies of scale. As a result, most Georgians presently live in some form of extended multigenerational household. As unemployment is particularly severe for first-time job seekers, families are also providing support to their children much longer than they used to Connections and clan support. Connections, through families or clans, have become extremely important, particularly in terms of securing employment in either the private or public sectors. Often, the jobs secured may not provide high pay in themselves (as is the case, for example, with many public sector jobs), but they do provide access to all sorts of advantages, privileges, and again, further connections. The importance of these other "benefits" is illustrated by the fact that, despite the extremely low level of public sector wages, state employee face a nauch lower risk of poverty (especially chronic poverty) than other population groups Exclusion from informal safety net. Because of the importance of social ties and connections, the informal safety net can easily exclude groups that are outside the network-based support system-a high risk in a country as fragmiented and ethnically diverse as Georgia. The importance of connections and "exchange of gifts" mentality have created a vicious circle, in which the poor, who have little to offer either in terms of gifts or connections, find themselves

51 -40- increasingly excluded and isolated from job opportunities and access to services. Weakened community ties, increasing distrust towards officials and the state, and sharpened social stratification have all contributed to weakening social solidarity, and to growing social disintegration. In its most extreme form, this is revealed in cases of abandonment of children, child labor, prostitution, or others directly linked to the limitations of the informal safety net The State Safety Net does not Protect the Poor 114. The formal safety net currently comprises three main cash benefits (family allowances, pensions and unemployment benefits), and some additional programs such as tariff discounts on electricity and transport, direct payments for electricity, and student stipends. In addition, the State provides significant benefits to IDPs displaced by the conflict in Abkhazia, both in cash and as consumer subsidies. Because of their importance and magnitude, IDP benefits are reviewed separately in section 5.3 below. Table 2. Main (Budgetized) Social Safety Net Programs in Georgia, 1997 Pensions* 962, Family allowances 350,000" Refugee benefits** 282, Unemployment assistance 20, Student stipends 23, Including direct payments to Sakenergo for electricity consumed by pensioners. " As originally planned. Actually paid out to only about 240,000 in * Includes all budgeted refugee programs including direct payments for electricity, but not other in-kind subsidies..15. The combination of macroeconomic crisis, hyperinflation and complete fiscal collapse eroded cash social benefits to a bare minimum. Starting in early 1995, the Government tried to reverse this trend in the wake of (slowly) improving fiscal revenues. Since then benefits have rnore than tripled in real terms, although they still remain well below subsistence levels. With a view to concentrating its meager resources on the most vulnerable, the Government has takcen a number of measures to improve the targeting of social benefits. In February 1996, the retirement age was increased by five years (in a single step - something unprecedented in the FSU). And in early 1997, the soviet-era system of universal child allowances was replaced by a system of family allowances targeted to three categories of poor families. Nonetheless, benefit levels remain extremely low (representing only 20-30% of the cost of the minimum survival food basket). Moreover, actual budget expenditures have often lagged the already modest planned allocations: in 1997, for example, actual spending on family allowances amounted only to 7.8 million lari, less than one-half of planned expenditures of 15.9 million. Similarly, at different times during 1998, the Government accumulated arrears on pension payments; and actual spending on family allowances during the first 10 months of the year was only 3.2 million, as compared to a planned annual allocation of 14.4 million. As of end- 1997, actual total social safety net expenditures, including pensions but not refugee benefits, amounted to around 2 percent of GDP (or 13.5% of total spending). Refugee benefits represented an additional 0.8% of CGDP Social Assistance Benefits (Family Allowances). The family benefit program was introduced in early 1997, to replace the previous system of universal child allowances. It was intended to cover some 350,000 individuals (in 175,000 households), of which 90,000 were single pensioners (defined as pensioners with no legal relatives of working age); 110,000 were 2 As documented by Dudwick in her social assessment.

52 -41- disabled individuals, and 150,000 were unemployed in three narrowly-defined categories. Due to lack of resources, this last category was reduced during the year to families with no employed members and at least one registered unemployed. The benefit level was set to equal the standard deduction for the income tax, which was 9 lari per month. One-person households received 100 percent of this standard benefit; two-person households received 80 percent per person (7.2 lari per person per month); while households with three persons or more received 60 percent per person per month (5.4 per person per month). Plarnned funding for the family benefit in 1997 was 15.9 million lari, but actual spending amounted only to 7.8 million, less than one-half of planned expenditures In 1998, the family benefit program was reduced in scope. Eligible categories were reduced to one: (legally) single poor pensioners. ]Benefit levels were kept the same. Some 40,000 households with about 55,000 members received the benefit during the first half of the year. During the second half, the benefit was rarely paid out for lack of resources. Planned funding for 1998 budget was 14.4 million lari, but actual spending for the first 10 months of the year amounted to only 3.2 million lari. The difference was due in part to a shortfall in revenues, but also to a reallocation of spending towards other (non-social) categories, by both central and local budgets In addition to a severe lack of funds, the family benefit system faces the problem that it is funded out of the central budget, but delivered through local budgets with little accountability. Transferred funds are theoretically earmarked, but in practice the central government has little capacity to monitor or enforce payments, and the funds can be easily reallocated to other uses. The situation is expected to worsen in 1999, as the transfer will no longer be earmarked: in other words, local governments will have a nominal responsibility to pay out the family benefit, but there will be no direct link between the central transfer and actual payments. On a more positive note, the Government has expressed its commitment to keep the allocation for the poverty benefit in the 1999 budget at the same level as was originally planned for 1998 (14.4 million). These resources will continue to be targeted only to single poor pensioners, but the benefit level will be increased to 18 lari for a single person household (29 for a two-person household)-sufficient to bring the average consumption of poor single pensioners close to the level of the poverty line As discussed in Chapter 1, single pensioners face a high risk of extreme poverty, especially in urban areas. The average monthly consumption of poor single pensioners is very low (only 38 lari per month). And this group typically lacks the connections (to family or clan) that are so important in Georgia to ensure survival. Hence, lhe benefit appears to be well targeted. Unfortunately, the budget envelope is so small that many other families at risk of poverty are necessarily excluded: most importantly, families with disabled members, families with no working members; families with more than 3 children; and single adults with dependents Given Georgia's existing resource constraints, very little can be done in the short term to extend social assistance coverage to other families at risk. Several possible measures, however, could be implemented to improve targeting and alleviate the resource constraints. first, transferring the social pension budget (which pays pensions to individuals with no working history) to the family allowance program, which focuses on poor pensioners; second, improving the eligibility for disability benefits to focus on t]he truly disabled; and third, targeting special pensions for war veterans and others to the groups more at risk (for instance, war veterans with dependent members). These simple measures could allow for an increase in the level of benefits received by eligible beneficiaries, or alternatively, resources could be used to extend the existing limited family allowance program to other groups at high risk.

53 In the longer term, cash benefits should stay limited to individuals not able to uwork and not supported by families with members able to work-i.e. mainly pensioners living alone and the disabled. The poor able to work should be supported through self targeted employment schlemes organized to provide the minimum level of subsistence (see section 5.5 below) Supplementary cash and non cash benefits are important in some urban municipalities. These were nominally inherited from the Former Soviet Union, but in practice are relevant only in the few municipalities with meaningful fiscal resources. For example, the city of Tbilisi gives an annual transfer of 50 lari to some 4,800 families that are perceived to be needy. Similarly, in 1996, the city of Poti allocated some 125,000 lari in the form of free or subsidized firewood to 3650 families, and distributed some 12 tones of kerosene and 10 tones of coal for free. However. in general the extent and nature of these programs is not well documented; nor are they reflected in the SDS survey. 1:23. Tariffdiscounts and other consumer subsidies. In many municipalities, certain groups (often the elderly, defense, medical and education personnel) are eligible to receive certain services at discounted rates or free of charge (this applies primarily to urban transport, electricity, water and other urban services). In some cases, these costs are subsidized by local budgets; more olten, they are de facto subsidized by the providers. Unfortunately, the extent and real costs of these measures are not well documented. Only for the electricity sector is it possible to make some inferences According to the Tbilisi electricity distribution company (Telasi), in Tbilisi alone there are 44,814 customers who are eligible for a 50% discount on electricity tariffs, an additional 13,118 customers are eligible for a 100% discount (about 17% of all consumers). According to Szakernergo, some 123,000 customers in all of Georgia are eligible for discounts (approximately 20% of all consumers). Outside of Tbilisi, however, it is unclear how widely these discounts are applied. In Rustavi, for example, they were recently eliminated with minimum complaints From the population Given Georgia's rationed supply of electricity, and widespread non-payment by the population, it is hard to infer the true cost of these discounts. According to the SDS survey, the average Georgian household spends 14.7 lari per month on energy and fuel (about 5% of total household monthly expenditures). Approximately 60% of this figure is spending on relatively expensive heating fuels (kerosene and wood), with only 24% of the energy bill representing outlays on electricity. Hence, the average family is spending some 3.5 lari per month on electricity, or the equivalent to 60 kwh at current retail tariffs. One way to "cost' the transfer is by applying the discount to that base (i.e. 50% of the price of 6OkWh or about 1.75 lari per month per household). However, given significant non-payment by the population and the proliferation of illegal connections, actual electricity usage for those receiving supply is likely to be higher than that. Assuming that average household usage is around 150 kwh per month,' the expected value of the tariff discount (for those receiving the discount and electricity supply) would range from 4.5 lari per month (for those eligible for a 50% tariff reduction) to 9 lari per month for those eligible for the full 100%. In Tbilisi alone, this would amount to some 320,000 lari per month, or 3.8 million lari for the year. This is comparable to what was actually spent o01 the whole farmily ajlowance program in 1998! It Is impossible to determine, with any degree of reliability. actual average monthly consumption of electricity by households in Georgia. However, the comparison with Armenia, which has a similar climate and housing stock. suggests that an average of 150 kwh per month is a reasonable estimate.

54 There is some debate about the exact list of categories eligible for discounts: the most reliable information suggests that some 18 categories of households are eligible for the 50% discount; in addition, disabled veterans of WWII are eligible for a 100% discount. These discounts were enacted under 13 different legislative acts, of which 9 are old soviet laws. The categories can be broadly grouped into: (a) employees of the Ministry of Internal Affairs and State Security; (b) veterans of all wars; (c) distinguished pensioners; (d) people affected by Chernobyl; and (e) others. From the poverty profile, we know that among these categories, only the disabled face a high risk of poverty. Clearly, employees of the Ministry of Internal Affairs and State Security are not particularly likely to be poor. Similarly, veterans of WWII, who receive a pension of 45 lari per month rather than th- regular 11.8 lari per month pension, are not likely to be poorer than other groups. According the SDS survey, "special" pensioners have incomes that are at least 50% higher than the basic pension benefit. Or in other words, some of the pensioner households who are eligible for the discount are less likely to be poor than pensioners who are not eligible. In sum, the tariff dscounts are not targeted to the poorest groups. not only do they represent an inefficient instrument for social assistance (since the cost is born by the provider), but they are also most likely inequitable. These discounts should be phased out, and the only truly vulnerable category (1:he disabled) made eligible for the family allowance program Pensions. The current system of pension benefits was introduced in 1991, with the creation of the Social Insurance Fund (SIF). Initially, benefits were linked to pre-retirement wages and the length of covered employment. In early 1995, with the system on the verge of collapse, differentiated pensions were abandoned and replaced with a flat-rate pension. In February 1996, the retirement age was increased to 60 for women and 65 for men. At the same time, an attempt was also made to improve collections of the payroll tax which finances the system. As a result of these reforms, the dependency ratio decreased from 67 % in 1994 to 55 % in 1996; the pension replacement ratio increased from 14.5% to 27.6%; and the payroll tax collection increased from 0.7% of GDP to 1.2 % of GDP. Today, the SIF pays pension benefits to old-age pensioners, disabled persons, and war veterans as well as their survivors. Individuals with no working history receive a social pension, for which women become eligible at 65 and men at 70. The cash benefit is currently 11.8 lari per month for old-age pensioners and the disabled; 45 lari per month for war veterans. In addition, for each pensioner a monthly amount of 1.8 per pensioner is transferred directly from the budget to Sakernergo to cover the cost of their electricity consumption. 4 The Ministry of Defense administers a system of pensions for officers, which provides eamings-related pensions of, on average, 60 lari per month. Pension benefits currently amount to about 1.8 % of GDP The pension system is financed mainly through a payroll tax. Since 1997, the payroll tax for employers in the private sector is 27 percent, and the employee contribution is I percent. Despite the high rate, payroll taxes cover only about 60 percent of pension outlays. This revenue shortfall is typically compensated by the central govemment, and often translates into significant arrears and irregularities in the payment of pensions. As of May, 1998, for example, the system had accumulated arrears of about 4 months worth cf pensions. In its present form, the pension system is financially unsustainable. It is meant to deliver pensions to a large number of eligible beneficiaries (some 962,000 pensioners or 18 percent of the official population), but collects its rvvenues on an extremely small base. Compliance with payroll tax laws is estimated to be only about 12 percent. In 1996, for example, wages over which contributions were paid amounted to only about 4 percent of GDP. The total wage bill, including the informal sector, was estimated at 4 At current tariffs this amount corresponds to about 30 kwh of electricity. But it is transferred to Sakernergo irrespective of whether the pensioner actually receives electricity or not In this sense, it acts more as forced consumption of electricity than as a payment to the pensioner per se.

55 -44- :33 percent of GDP. Clearly there is a need to extend the tax base to agricultural workers, the private sector and small businesses Although the real value of pensions have increased significantly in real terms since their low point in 1994, the pension benefit falls far short of the costs of even the minimum "survival" food basket, which is estimated at 40 lari per month. The pension benefit is also low relative to the average wage: the ratio is currently around 9 percent. In sum, the pension system delivers very low benefits at a relatively high cost In the short-term, efforts to strengthen the system must focus on extending the base for the payroll tax, and on improving efforts to track and register reczpients. These tracking mechanisms alre currently very weak, and it is possible that the system has a high level of ghost recipients. Other possible steps which could be taken include: a further rise in the retirement age for women; tightening eligibility criteria for disabled pensioners; making the benefit formula more restrictive; and improving the transparency and accountability of the SIP Also, the system for lump-sum, direct payments of electricity charges for pensioners should be phased out, as the electricity companies become privatized. These direct payments take away control from pensioners over part of their benefits. To the extent that they are not linked to actual consumption of electricity, they represent a redistribution of income from pensioners in areas with severely deficient or no electricity supply to retirees who do receive electricity. And they hamper the emergence of a culture of paying for electricity. Moreover, any rationale for maintaining such a system disappears once distribution companies have been sold to new private owners, who will face their own incentives for enforcing payment discipline If relatively little can be expected from the pension system in the short term to protect the poor, urgent reforms are needed in the medium term to create a system that is sustainable and that pays benefits at levels above the poverty line. Georgia faces a particular challenge because of its demographic profile: its population is older than in any other ECA country of similar income levels. But Georgia also faces an unprecedented window of opportunity, as the present crisis has de facto reduced both the political and economic costs of undertaking a drastic reform of the PAYG system. The steps necessary to establish a new system are: (i) to strengthen the PAYG in the short and medium term, and transform it into a system that pays a minimum flat benefit to the elderly, with no link to contributions; (ii) to introduce (over the medium term) a second pillar-i.e. a mandatory and fully funded system, or in other words, a real old age insurance scheme. This system could be complemented by a voluntary, private pension third ptillar, for which preliminary legislation has already been adopted. The first pillar (the PAYG) could disappear in the long term once the old age insurance system is fully functional. 1:32. Unemployment Benefits. Georgia pays unemployment benefits for up to six months to eligible persons who have been laid off. Payments are made at local State Employment OfiFices, and financed out of the Employment Fund (EF), which collects a 1 percent payroll tax from the non budgetary sphere. In 1996, benefits were paid at the rate of lari 9.8 for the first two months, lari 7.8 for the next two months, and lari 6.8 for the final two months. This amounted to just under 30 percent of the average wage in 1996; but by end- 1997, it represented only about 1 8 percent of the average wage. In addition, for each registered unemployed, the EF transferred 1.3 lari directly to the electricity company to cover the cost of electricity consumed by the urnemployed. In 1998, the unemployment benefit was raised to 13 lari, the direct payment for electricity was discontinued, and benefit coverage extended to a 12 month period. Still, the replacement rate remains very low (about 20 percent). 5Eka Vashakmadze, The Reform of the Pension System in Georgia, Tbilisi, 1998.

56 -455- Figu re : Georgia: registered unemployed, ^ =R Registered jobseekers with status of unemployed \X 5 \ 000 e of which: receiving om benefits 0- ol , a, '0 'o t 0 5, ' E E E E 5 E Source: Ministry of Labor and Social Protection Only a small fraction of the unemployed actually receive benefits. Because of the limited duration of benefits, many recipients have exhausted their eligibility, and less than 20 % of the registered unemployed actually receive a benefit (Figure 1). Moreover, because of the low level of benefits, and the irrelevance of local employment offices to the job search process, incentives for the unemployed to register are low. Trends in the number of registered unemployed seem to bear no relation to overall economic developments, but rather lag changes in the regime of granting unemployment benefits and family allowances. For example, the number of registered job seekers almost tripled between June and September of 1997 in response only to the announcement of eligibility rules for family allowances, which included families with registered unemployed. Registered unemployment started to fall as soon as the list of beneficiaries for family allowances was narrowed to exclude the unemployed Unemployment benefits are generally set up to insure against the loss of income from short-term unemployment, and to give persons who become unemployed the opportunity to look for appropriate alternative employment. The Georgian system of unemployment benefits, however, does not function as an insurance scheme; rather it has the flavor of a social safety net scheme, albeit an inefficient one. Replacement rati.os are very low; coverage is negligible; and as a safety net, the scheme is not very accurately targeted Despite the EF's small financing base and low tax rate, it supports several activities besides income maintenance, such as retraining, and temporary job creation. Given the lack of resources, and widespread international evidence on the high cost and low effectiveness of these types of measures, these activities should be discontinued. Moreover, given that Georgia at present lacks the resources to operate a true Unemployment insurance scheme, it vwould make more sense to eliminate the EF altogether. This would save resources currently spent on the administrative costs of running the Fund, and would allow for a lowering of the payroll tax. Responsibility for maintaining some sort of last-resort safety net for the unemployed could be transferred to the central budget, and take the form of a modest increase in thefamily allowance scheme to include fimilies with at least one unemployed member and no working members. AIternatively, the unemployed could be reached through self targeted employment or public works schemes organized to provide the minimum level of subsistence to those who are able to work (see section 5.5 below). Ir. the medium term, as the economic situation irproves and the economy becomes more formalized, the introduction ofa real unemployment insurance scheme could be considered; but its advantages and potential cost should be carefully assessed beforehand.

57 IDPs Receive a Disproportionate Share of State and Humanitarian Assistance 136. Some half-million people were displaced from,their homes by the civil conflicts in the regions of Abkhazia and South Ossetia. Nearly 290,000 have been dispersed through Georgia, with the largest concentration in Tbilisi. Others have remained in Samegrelo, the region bordering Abkhazia, and in Lower-Swaneti. In 1996 a UNHCR-sponsored survey registered 282,000 people as IDPs. Contrary to commonly held perceptions, the analysis of the SDS household survey suggests that IDPs that have resettled on their own or have integrated themselves into local communities (the refugee population that does not remain institutionalized) face a lower risk of poverty than the average Georgian household (4 percent of them are poor as opposed to 10 percent of the total population). And they face the lowest risk of extreme poverty of almost any population group. The surveys show, moreover, that one out of four refugees iamilies have left the country. 6 :137. The fact that IDPs face a relatively lower risk of being poor contrasts sharply with tie fact that they receive a large share of State and humanitarian assistance. The IDP program is one of Georgia's largest safety net programs, ranking second only to old age and invalidity pensions. In :997, Republican budget expenditures for this program totaled some 53 million lari; the planned allocation for 1998 was 62 million lari. The program comprises a diversity of cash and non-cash benefits (Table 3). The major component is the monthly stipend of 11 lari per refugee for those IDPs living in hotels or Government institutions; and of 12 lari pre refugee for families living in their own accommodations or with other families. In addition, for each of the 160,000 IDPs that have settled on their own, the budget transfers 1.8 lari per month directly to Sakernergo to pay for their electricity; another 3.8 lari per refugee are transferred for each of the 124,000 IDPs estimated to be living in hotels or refugee centers. An additional 7 lari per person per month is paid to those families that have resettled on their own to cover housing costs and other utilities. IDPs also receive public transit subsidies, including free subway tickets for those living in Tbilisi, and passes for ground transports such as buses. The Ministry of Refugees and Accommodation nmaintains a special assistance fund for the destitute. In addition to benefits paid out through the Ministry of Refugees, IDPs who are also old-age or invalidity pensioners receive a 20 percent pension supplement. IDPs also receive free education and are exempt of fees. And the Basic Etenefit Package of health care services provided free to the population has integrated in its calculation free health services for 140,000 IDPs in The costs of these free or subsidized education and health services are born by the Ministries of Education and Health respectively. Not counted in the budget is the value of the rent of the buildings occupied for free by the IDPs, but that would be probably high, especially in the case of hotels in city centers IDPs have also benefited from a large proportion of the humanitarian assistance in food, medical, and other aid provided by governmental and non- governmental organizations to CGeorgia. Although the magnitude of humanitarian assistance has decreased since 1994, it remains substantial. Total humanitarian assistance received between April, 1994 and March, 1995 was estimated at US$62.7 million; between April, 1995 and May, 1996 at US$60.4 million; and between June, 1996 and May, 1997 at US$30.6 million. In 1996, some 800,000 beneficiaries received humanitarian assistance; in 1997, beneficiaries amounted to 340,000 people (a large fraction were IDPs). 6The survey does not sample IDPs living in state provided shelters (hotels or other structure). It is therefore difficult to compare the poverty status of this specific group of IDPs with the rest of the population. These IDPs are probably more vulnerable than the ones that have been integrated in local communities.

58 -47- Table 3. IDP program 1997 and 1998 (In million lari) Monthly benefit 34.9 Utilities for people housed by the state 12.4 Free metro fare (Tbilisi) 3.0 Ground transport 0.6 Special repairs of dwellings 1.5 Special assistance fund for destitute families and children 0.2 Interest paid for banking service 0.3 Total 52.9 Not included: value of free education; BBP IDP benefits are high in comparison with family allowances, unemployment benefits and even pensions. Not including benefits in kind, in 1998 an IDP family of four would receive a monthly cash benefit of 38.8 lari per month, above the average budgetary monthly wage of 35 lari. In contrast, a non-idp family of four would normally not be eligible to receive any state support since eligibility for the family allowance has been limited to legally single pensioners. If the non-idp family were to comprise a pensioner or an unemployed person, they would be eligible for a pension of 11.8 lari or an unemployment benefit of 13 lari-still well below the level of IDP benefits. Even ifidp benefits are not by themselves sufficient to lift a family completely out of poverty, they still provide non negligible support in comparison to that received by other Georgians. Moreover, IDPs are eligible for benefits regardless of other income or asset ownership. And as discussed above, in many cases, IDPs are better off than other groups of the population The contrast between benefits received by IDPs and those received by other, often more needy, families highlights the need to improve the targeting of assistance to IDPs, to limit eligibility only to those who truly need it. This is irnportant from a social equity point of view. It also is pressing because of the open-ended character of the IDP program. In the absence of a solution to the conflicts in Abkhazia and South Ossetia, there is no clear horizon for the phasing out of the IDP program; the burden on the state's meager resources could continue indefinitely. Mechanisms of self-targeting could be usefully employed to select the most needy. Linking payment of the monthly stipend to participation in Ipublic works, for example, could serve to weed out those with alternative, more lucrative sources of income from work. This could function as a self-contained program or as part of a broader public works scheme designed to target resources to those able to work. Alternatively, registration requirements could be increased significantly, so as to impose a sufficiently high cost on recipients, to again weed out those who do not truly need the benefit In addition to improving the targeting of cash benefits to IDPs, in-kind benefits should be phased out. Most of the burden of financing these benefits is born by providers of the services, or by other users. The government should assess carefully the cost of maintaining them, and if these in-kind benefits to IDPs are to be kept, the budget should fully cover the costs The authorities should continue their efforts to integrate remaining IDPs housed in iustitutions into local communities, as the evidence from the household survey suggests that once this happens, IDPs are able on their own to dramatically improve their living conditions. This integration is made difficult by the awkward legal status of many IDPs. IDPs can not acquire land in Georgia but may only lease it. They cannot purchase apartments, and their movements are restricted because of registration of residence requirements. This legal situation does not support their integration in the Georgian economy, and should be evaluated.

59 The collapse of transfers to poor groups of the population is correlated with the fact that the Government has given priority to protecting the level of t:ransfers to IDPs. In the current context, this may ultirnately create tensions between IDPs and other members of Georgian society. And it does raise questions of equity and allocation of resources. Somile of these choices may need to be reevaluated. It is also urgent that the status of idp be clarified, and that they benefit from the same ownership rights than any other citizen ofl Georgia. IF.4 The Voluntary Sector and Civil Society. Still in its Infancy The world of NGOs and associations dealing with vulnerable groups in Georgia is small but vibrant. The Government has adopted an open and liberal policy towairds hngos, and it is less cumbersome to obtain the status of NGO in Georgia than in many neighboring countries. Many rnunicipalities and regional Governments have shown interest in working with the NGO sector Anumberofcivil > z _. ~society la~of organizations hlave Di -t, iiemerged since indepenidence. InSome, like the Georgian., Organization of invalidis, have undertaken im-pressive advocacy work and implemented successfuli programs to support their members. The Georgian Organization of Invalid~s, for instance, lobbied Parliament to pass a Law on the Handicapped; and obtained ex.istin social chall, Gthat all taxes paid by the organization be reinvested in L ~~~~~~~~~~~~~~~~~activities benefiting in-valids. Other associations and NGOs are dealing with issues related to IDPs, street children, and the elderly. A number of Intemational NGOs have set-up programs to train and support the elergence of local NGOs. For example Save the Children and the Eurasia Foundation have setup training and support centers, as well as small grants programs. Considering the magnitude of existing social challenges, Georgian NGOs still remain underdeveloped and their activities linited in scope, but they are off to a promising start As in the case of other FSU countries, lack of local financing represents a major limitation to the sustainable development of local NGOs. At the same time, the availability of relatively lar-ge pools of donor funds sets up some perverse incentives: many groups and individuals are trying to obtain the NGO status solely in order to attract funds from donors, often with very weak prngramn proposals and sometimes even hiding a commercial activity. 5.5 Strengthening and Improving the Safety Net 147. Many of the comprehensive reforms needed 'to support the poor and vualnerable groups will be possible only -when the fiscal situation of the State has improved and that more mon-ey will be available to finance any prograrns. However, as discussed in Chapter 2, even a small, but welltargeted poverty benefit, can play a crucial role in dampening poverty. If the family allowance were to be kept even at the very low level observed in the 1997 budget (relative to GDP) but were accurately targeted, it would still be sufficient to reduce the incidence of poverty by several

60 -49- percentage points. Needless to say, if the social assistance budget were to be increased more rapidly (to 1% of GDP by 2007), the impact on poverty would be large (Figure 2). Figure 2. Poverty in Georgia: future paths under different levels of poverty benefit (New poverty line) 12%, ~-9% 2 3% - 0%- Ei ()5 200) Percent poor - initial level --o-poverty incidence- base case and no targeting of actual (1997) poverty benefit % Poverty incidence - base case, increase in poverty benefit to the planned 1998 levels and targeting l L-Poverty incidence - base case, targeted poverty bene:fit gradually increasing to I% of GDP by How can targeting of existing benefits be impiroved in the short term? And if the Government were to allocate more resources to the s3ocial safety net, how should it do so? We distinguish here between some immediate (very short term) responses, and those that require a slightly longer (one or two year) horizon Very short term responses (1999): Protect the family allowance system. Ensure that the allocation in the 1999 budget is at least the same level as in the 1998 budget, and that unlike in 1998 it is fully paid out. Eliminate tariff discounts to special population groups and other in-kind benef its. Replace with a direct payment (through a modest increase in the fami ly allowance system) for those groups of recipients who are truly needy, mainly the disabled (categories I and II). * Eliminate direct payments ftom the budget for electricity. This can be phased in geographically, starting with Tblisi, as electricity distribution companies are privatized. * Improve the targeting of assistance to IDPs. T];lirough self-targeted public works, proxy means testing or frequent registration requirements. Redirect released resources to the family allowance program or towards establishing public works programs as discussed below. Short-medium term responses. Establishing a more comprehensive social safety net ( ): The design of a comprehensive safety net i n Georgia has take into account two key characteristics of Georgian poverty. First, that poverty is closely linked to labor market status. Second, that poverty is mainly transient and that families move in and out of poverty frequently over time, so that accurate targeting is difficult. Moreover, this safety net has to be designed within a context of limited resources-a constraint which will remain in place for a long time.

61 -50- How can a safety net be designed within this context and still achieve its key objective of protecting (insuring) the poor against unexpected drops in income? The specific strategy proposed here focuses on a very simple principle: limit administered transfers and cash benefits to identifiable vulnerable groups who are unable to work. and assist the poor who are able to work through self-targeted public works and other programs that support employment. * Limit transfers (family allowances) to those who cannot work: poor elderly, the disabled with no other support, single adults with dependents. By focusing available transfer resources on these specific groups, benefit levels can be increased to more significant levels and truly help bring recipients out of poverty. This can be done without imposing an unrealistic burden on the State budget. * Support the poor able to work through self-targeted workfare programs (see Box). Such programs would essentially combine the best features of an Employment Guarantee Scheme with those of Social Funds aimed at promoting labor intensive projects in poor areas. To ensure that the money goes only to those who need it, without undermining the incentives to seek a normal job, the wage offered has to be low (at the level of local wage rate for unskilled manual labor in "normal" times). The government guarantees the financing of the wage costs of the program, and possibly part of the non-wage costs, with local communities coming up with the financing of remainder of non-wage costs. Such a workfare scheme has several advantages over administrative systems or other forms of cash benefits. First, it relies on self-targeting, so it is not necessary to identify a priori what categories of individuals or families will need support. Second, by linking benefits to work, it avoids creating a culture of dependence. Third, it can function effectively as a permanent insurance scheme for the poor that protects them against any unexpected downturn in their incomes. As such, it is particularly well-suited to respond quickly in crisis situations. It is in essence a permanent program or institution, with little administrative discretion, which lays dormant in good times, but is triggered automatically by increased demand for workfare when there is a shock to incomes (whether because of seasonal fluctuations, droughts, or mnacroeconomic crisis). E itsof - : a Good Workfar Program A ted be belovdl* mas tgicture '' ;irketae ullar or the ibiuityf< p*iticiga ion shouldte 'avoided; idealy, the only restriction should be the *...Fis unavo e (b+ecs ofbdet consainbts) he th program should be targeted to poor --. y(s of wa. bill in total ost) shouldbe higher than normal for similar. projects in b: j~eii6 on the remtia given to inmediateincome hould tbe targete towardpoor areas to easure*& the asets creaed ae of maidmun Siibr~jcts argte'to non-poor a-reas shoul require Tinancing fromi thec sihei~none sbo04cg backito the wv9or : gegt.. abrw a~en,m~aiwra~si~ig Wotare WorldBankPesearch Obswr y One of the challenges of establishing such as scheme is how to finance it. Demand for workfare is likely to be countercyclical: i.e. it goes up in bad times, precisely when the Government has less resources. Hence, it is important to separate its financing from the annual budget cycle. One way to do this is by capitalizing a Fund (perhaps using privatization receipts), which will then finance the implementation of public works programs. On a small scale, it could

62 -51- be piloted through the Georgia Social Investment Fund, which is financing a program of community based infrastructure rehabilitation. The GSIF program could be increased in size and additional labor intensive projects be carried out in urban centers. Alternatively, it could be piloted through a completely separate Fund. Two approaches can be used for the implementation of public works programs. One option is to implement the work through small private contractors. In this case labor intensity is ensured through the design of selection criteria for the infrastructure to be rehabilitated, and by limiting access to work contracts to small-size contractors from the private sector. In this case no specific conditions are set on hiring of labor other than a low wage rate. Another possible approach is to have the work implemented by local governments directly, with the wage level set at a level slightly under the market wage for unskilled manual workers. The local government would ensure that works are carried out using labor intensive methods. This method has been used in some large scale programs in Latin America and in Asia. However, this requires having relatively strong and accountable local governments. And public wages have to be high enough so that workers from local government and public sector are not tempted to supplement their jobs with participation in the public work scheme. Using this second approach in Georgia at present could risk a high misuse of funds. s _ 5b.. W. 2. Fr W; X~~~~~~~~~~~~~~~~

63 Other measures to support this new safety net. * Eliminate the EF and shifting responsibility for protecting the unemployed to the new public works program. Eliminate the EF and reallocate the resources saved to the family allowance program (or reduce the payroll tax to stimulate employment). * Continue to improve the efficiency of the PAYG pension system. * Increase incentives for good and efficient NGOs to expand their activities. Focus efforts on attracting local financing, improving institutional capabilities, and ensuring good coordination with national policies. * Introduce self targeting mechanisms to orient what is left of humanitarian assistance to the poorest. Ensure that remaining flows of humanitarian assistance go to the poorest groups. Replace distribution of humanitarian aid to socio-economic categories with self-targeted mechanisms (soup kitchens, food to mother and children with special needs at local health centers etc).