70584 v1 Georgia Programmatic Poverty Technical Note #1: Poverty and Crisis Impact JUNE 28, 2010 EUROPE AND CENTRAL ASIA REGION 1 Georgia: Poverty and Crisis Impact Note1 Summary: This note is part of the World Bank’s programmatic poverty work in Georgia, which is composed of a series of periodic notes on a range of policy topics of relevance to Georgia’s poor. It examines the impact of the crises on poverty and welfare at the household level in Georgia. The note uses new data from UNICEF as well as simulations based on 2008 data to provide a first insight about transmission channels of the crisis and their impact on well-being. It also explores the extent to which Georgia’s system of social transfers reduces poverty. Key findings include (i) simulations indicate that the poverty rate may have increased by 1 to 2 percentage points as a result of the crisis; (ii) the main impact of the crisis on household welfare in Georgia has been through the credit and labor markets; and (iii) social transfer programs are playing an important role in reducing poverty and crisis mitigation, but offer more limited scope than the labor market policies deployed elsewhere in the region. The note also outlines a number of questions that will be addressed in subsequent work. I. Introduction 1. After four years of rapid economic growth and reform, Georgia’s economy faced a serious downturn in mid-2008. Armed conflict with Russia in 2008 was followed by the financial crisis and the subsequent global recession. The economy contracted by an estimated 3.9 percent in 2009, dampening the gains that had been achieved since the Rose Revolution in November 2003. The economy bottomed out during the second half of 2009, and returned to positive growth of 0.4 percent in the fourth quarter. 2. The main impact of the crisis on household welfare in Georgia has been through the credit and labor markets. Debt repayments impose a significant hardship on the majority of households. From 2005 to 2008 there was a rapid expansion in consumer credit. Following the onset of the financial crisis households faced a combination of increasing interest rates, adverse exchange rate movements on loans in foreign currencies, and reduced new lending. These impacts were compounded by lower real incomes for borrowers and non-borrowers alike through increases in unemployment and reduced hours and employment earnings. Both transmission channels have contributed to an increase in poverty. Georgia’s social transfer programs, which were considerably strengthened and expanded during the growth years of 2004–08, play an important role in keeping people out of poverty and providing some cushion for those already in poverty. 3. Simulations undertaken for this technical note suggest that poverty is likely to have increased by about 1 to 2 percentage points in 2009. Baseline poverty headcount estimates from recent surveys, while based on a different methodology, have been in the range of 23 to 26 percent. The poverty and welfare impact of the crisis is greater in Tbilisi and other urban areas than it is in rural areas. Similarly, the impact is more pronounced among those working in the service and manufacturing sectors, which experienced the largest contractions in the recession. 4. This note analyzes the impact of the crisis on poverty and well-being, government policies to mitigate the crisis, and possible future policies to better protect the poor and accelerate recovery. The impact of the financial crisis and global recession on poverty and welfare of Georgian households is analyzed using a 2009 UNICEF survey as well as simulations based on the 2008 Household Budget Survey (HBS). The note addresses four main questions. First, how have the crises and recession affected households in Georgia, and how have they tried to cope with the shock to their livelihoods? Second, 1 Prepared by Kenneth Simler, ECSP3, with inputs from Owen Smith, ECSH1. 2 how have the level and distribution of poverty changed as a result of the crisis? Third, to what extent have government policies, and particularly social protection policies, mitigated the impact of the crisis on the poor? And fourth, looking forward, what other policy options are available that may provide better protection for the poor? 5. The note is organized as follows. Section 2 provides an overview of the Georgian economy and discusses the evolution of poverty in Georgia. Section 3 assesses the impact of the economic crisis on poverty and welfare in Georgia, drawing from a range of household surveys and simulations. Section 4 reviews the government’s social policy response to the crisis, particularly the effectiveness of targeted social assistance and pensions in mitigating the poverty impact. Section 5 explores future policy options for crisis mitigation in Georgia, including expansion of existing social protection programs as well as adoption of policies that have proved successful elsewhere in the region. Section 6 presents a brief summary and concluding observations. A technical annex describes the details of the simulation methodologies employed. II. Overview of Recent Economic Developments and Poverty Levels 6. Georgia’s development path since 1990 can be characterized as decline followed by reform and recovery. During the 1990s Georgia was one of the poorest performing countries in the ECA region, with weak governance and public finances, accumulating unpaid wages and pensions, limited electricity service, and widespread corruption. In 2003 GDP was only at 40 percent of its 1989 level. From 2004 to 2008 Georgia implemented far-reaching reforms to strengthen public finances and administration, invest in public infrastructure, liberalize trade, and improve the business climate. Significant FDI inflows helped fuel annual GDP growth rates of more than 9 percent, led primarily by the services sector and to a lesser extent by manufacturing. Tax and anti-corruption reforms led to a sharp increase in government revenues, which supported public infrastructure investments and the clearing of wage and pension arrears, as well as the expansion of social services and transfers. 7. Georgia experienced a sharp downturn in economic growth as a result of the double shocks from the August 2008 conflict and the subsequent global economic crisis.2 The double shocks led to significant deterioration in investor and consumer confidence, contraction in foreign direct investment, exports, and remittances, a cutback in bank lending, and stress on public finances from revenue shortfalls and increased expenditure needs. As a result, the economy contracted by an estimated 3.9 percent in 2009 following growth of 2.3 percent in 2008, which represents a sharp slowdown from rapid growth in excess of 9 percent during the preceding four years. FDI inflows collapsed from $1.67 billion (16.4 percent of GDP) in 2007 to a projected $770 million (7.1 percent of GDP) in 2009, while exports fell from 31 percent of GDP in 2007 to an estimated 25.6 percent of GDP in 2009. Inflation (end-of-period) at 3 percent for 2009 and 5.6 percent for 2008, is down significantly from 11 percent in 2007. In general, however, the 2009 GDP contraction in Georgia was less severe than in the hardest hit countries in the region, notably the Baltic States and Ukraine. 8. The downturn in economic activity during 2008-2009 has been concentrated in those sectors which fueled strong growth during the preceding four years. The construction sector contracted by an estimated 13 percent in 2009, following contraction of 11 percent in 2008. Manufacturing contracted by 6 percent in 2009 and 2.4 percent in 2008. The services sector (including particularly retail and wholesale trade, transport, telecommunications, and financial intermediation), which 2 This discussion borrows from the analysis of the proposed second Development Policy Operation for Georgia. 3 constitutes more than 50 percent of the economy, contracted by 2.5 percent in 2009. The collapse in FDI inflows across sectors contributed to these contractions, with FDI falling by 65 percent in the transport, banking, and other services sectors, and by 50 percent in construction and industry. 9. The economy is projected to grow by 4 percent in 2010, although there is significant uncertainty regarding the pace of economic recovery. The improved economic outlook is based upon stronger signs of a pickup in real economic activity starting from the fourth quarter of 2009, in particular the export sector. Growth is projected to remain at 4 percent in 2011 and increase to 5 percent during 2012-2013. However, significant uncertainty remains regarding the timing and strength of these expected drivers of economic recovery. Table 1: Selected economic indicators, 2007-2010 2007 2008 2009p 2010z GDP growth rate 12.3 2.3 -3.9 4.0 Unemployment rate 13.3 16.5 16.9 n.a. CPI (end of period) 11.0 5.6 3.0 5.0 (% of GDP, except where noted) Overall fiscal balance -4.7 -6.4 -9.2 -6.8 Exports of goods and services 31.1 28.7 29.8 34.7 Imports of goods and services 57.9 58.3 49.0 55.3 FDI inflows 16.4 11.8 7.1 7.6 FDI inflows (million USD) 1,675 1,523 765 858 Note: p=preliminary; z=projections; Source: Georgian authorities and WB estimates 10. The Bank’s most recent poverty assessment estimated that in 2007, 23.6 percent of Georgia’s population lived below the poverty line of GEL 71.6. Poverty was higher in rural areas (29.7 percent) than in urban areas (18.3 percent). A Welfare Monitoring Survey undertaken by UNICEF in June–July 2009 estimates a poverty headcount rate of 25.7 percent, and an extreme poverty rate of 9.9 percent; however, these results are not comparable with the LSMS 2007 due to different survey methodologies. Furthermore, issues with the reliability and comparability of the annual Household Budget Survey (HBS) make it difficult to assess poverty trends in Georgia over the past decade (see Box 1). 11. Although data constraints pose a challenge to drawing reliable inferences about changes in poverty levels over time, the poverty profile is fairly consistent across the different surveys. For example, in the UNICEF 2009 survey the difference in the rural and urban poverty rates (31.5 and 20.1, respectively) is similar to that found in the Bank’s 2007 survey. There is wide variation in poverty rates by region in both surveys, with poverty rates ranging from 15 percent in Adjara to 40 percent in Mtskheta-Mtianeti in the UNICEF survey. Households with more children are more likely to be poor; according to the UNICEF survey households with three or more children are more than twice as likely to be poor than a household with no children. Also, households in which adults have more education are less likely to be poor. A remarkably high 38 percent of households in the UNICEF survey reported that their income was insufficient to meet their basic food needs, with an additional 39 percent saying that their income was enough only for food. This represents an increase over the corresponding figures of 28 and 34 percent, respectively in the 2007 survey, which used the same question and is considered comparable on this specific measure of well-being. 4 Box 1: Data sources and their limitations This technical note uses two micro-data sources to assess the impact of the crisis on household welfare . One source is a Welfare Monitoring Survey implemented by UNICEF during June and July 2009. The survey covered 4,808 households, and was designed to capture the impact of the crisis on household welfare. It is the first (and at present, the only) nationally-representative household survey data available that covers the period since the onset of the crisis. The questionnaire includes modules on demographics, employment, income, expenditure, housing and durable assets, access to services, coping strategies, and child care. The second major source is the national Household Budget Survey from 2008. The HBS is an annual survey that collects detailed information about demographics, income, expenditure, and employment. In 2008 the HBS covered 19,684 individuals. However, the analysis is limited by several data constraints. Most importantly, the UNICEF data from 2009 are not fully comparable with the HBS or LSMS surveys. Although some of the questions in the UNICEF survey are the same as previous surveys, most are not, so it is not possible to make before-and-after comparisons for most variables. For example, because of differences in survey implementation the unemployment rate from the UNICEF survey is nearly three times current estimates. Also, the UNICEF survey includes a consumption module, but it differs in many ways from the HBS, such as different coverage of food and nonfood items, different recall periods for nonfoods, and seasonal issues arising from the short two-month window in which the UNICEF survey was collected, compared to the 12 months of the HBS. This makes the UNICEF data very difficult to compare with the 3 HBS. The availability of the 2009 Household Budget Survey (expected in mid-2010) will make a more detailed comparison of welfare effects possible, especially since GeoStat instituted changes in the survey implementation methodology in 2008. These steps included stronger quality control and replacing a significant number of interviewers. It is anticipated that this will improve the quality of the 2009 HBS data, although careful analyses of the changing methodology in data collection will be necessary. Further, the available data sources do not allow for a full treatment of the impact of credit and borrowing. Ideally one would like to know what percentage of households were carrying debt at the onset of the crisis, as well as the amounts and terms of that borrowing. None of that information is available in the household surveys. Moreover, the data that are available do not always appear to be internally consistent. For example, while 62 percent of the households who reported a worsening economic situation (31 percent of the total sample), listed debt repayment as one of the top three reasons, only 22 percent of all households listed debt as one of their economic problems, and only 7 percent reported debt as their main economic problem. Lastly, separating the conflict shock from the economic shock is very difficult. The armed conflict with Russia in August 2008 undoubtedly adversely affected the well-being of a significant part of the population, but it is nearly impossible to separate the impact of this shock from the financial crisis that hit shortly thereafter. This is not strictly a data limitation, but represents a significant analytical limitation. III. Impact of the Crisis on Poverty and Welfare 12. There are several transmission channels through which the economic crisis may increase poverty. The recent ECA regional report The Crisis Hits Home outlined three major channels through which the multiple shocks of the past two years have affected household welfare. The first is the disruption of credit markets, following the rapid growth of consumer credit in the region. Much of that credit is denominated in foreign currencies, so that households are vulnerable not only to interest rate shocks, but also exchange rate movements. The second channel is household employment and incomes 3 In particular, the method of Lanjouw and Lanjouw (2001) relies heavily on stable consumption patterns across the two surveys. In the context of the food and fuel price shock in 2008, the conflict with Russia in August 2008, and the subsequent global economic upheaval starting in late 2008, this assumption would be difficult to defend. 5 via the labor market. Reduced demand for exports as well as non-tradables such as the formerly- booming construction sector, combined with tight credit for new investments, increases unemployment and exerts downward pressure on wages. Third, changes in relative prices, especially for food and fuel, have affected real incomes. Although food prices have subsided from their record levels of mid-2008, they remain above their pre-2007 levels. The regional report concluded that the crisis led to a significant increase in poverty, with considerable heterogeneity across countries in the region. 13. In Georgia, the main transmission channels appear to be through credit and labor markets. In the UNICEF survey 31 percent of households reported that their economic situation had worsened between mid-2008 and mid-2009, with an additional 19 percent reporting that their situation had significantly worsened. Forty-six percent said their situation had not changed, and only 2 percent said their economic situation had improved. Among those whose situation had worsened or significantly worsened, 62 percent listed debt repayment as one of the top three reasons for their economic difficulties (Figure 1). Combining the “Lower total income,� “Loss of job� and “Decreased remittances� responses, 51 percent of households reported employment-related events as a cause of their worsened economic situation.4 These are discussed in greater detail below. Figure 1: Main reasons given for worsening of household’s economic situation 70 Households (up to 3 reasons per HH) 60 Rural Urban Total 50 40 30 20 10 0 Debt Lower total Loss of job Decreased Serious Reduced Loss of repayments income remittances illness agric output breadwinner 14. Poor households were disproportionately affected by the worsening economic situation. In response to the question “How has the economic situation of your household changed during the last 12 months?�, 27 percent of households in the poorest consumption quintile reported that their situation had significantly worsened, compared to 15 percent in the richest quintile (Figure 2). Urban households were also more likely than their rural counterparts to report that their economic situation had worsened or significantly worsened. 4 The next most common response was increased expenses from a family member’s illness, reported by 29 percent of households. Presumably the illness is not a result of the economic shock, and is therefore not investigated here. 6 Employment 15. Reduced employment is one of the main transmission channels through which the economic crisis has affected the welfare of Georgians. Open unemployment is only one manifestation of the employment shocks. The crisis has also affected household welfare through reductions in the number of hours worked, as well as some workers shifting to less productive and remunerative positions— including informal self-employment—after losing their primary job. The reduced demand for labor is most pronounced in Georgia’s export sectors. Figure 2: Self-reported changes in economic situation between mid-2008 and mid-2009 100 90 80 Percent of Households 70 Don’t know or refused 60 Significantly improved 50 Improved 40 No change Worsened 30 Significantly worsened 20 10 0 Quintile Quintile 1 2 Quintile Quintile 3 4 Quintile 5 Urban Rural Total Source: World Bank staff calculations from UNICEF 2009 survey 16. Households consider unemployment as their biggest problem by a wide margin. The 2009 UNICEF survey asked households to indicate the main problems that caused them hardships, and which of these problems was the main problem. A majority of households named four or more problems, with the most common problems cited being unemployment, buying medicines, and medical services (Table 2). When asked to name their most pressing problem, 42 percent listed “unemployment of household members,� while “buying medicines and “medical services� were a distant second and third at 13 and 12 percent, respectively, at the national level. The unemployment result is almost identical in rural and urban areas, while other problems such as utility charges and hunger appear to be more prevalent in urban and rural areas, respectively. Of those who said that unemployment was their main problem, approximately one-half responded that the household’s economic situation had worsened (33 percent) or significantly worsened (17 percent) over the previous 12 months, which aligns closely with the overall results shown above in Figure 2. 17. Unemployment has increased from 13.3 percent in 2007 to 16.5 percent in 2008 and further to 16.9 percent in 2009. According to the UNICEF survey, those who lost their jobs between mid-2008 and mid-2009 tended to be from higher income households, were better educated than the average worker, and were more likely to have worked in the formal sector (private or public) in urban areas, especially Tbilisi. In addition, men are slightly more likely than women to be among the newly unemployed. 7 18. Microsimulation results reinforce the finding that the negative welfare impact of higher unemployment is most pronounced in Tbilisi and other urban areas. Microsimulations using the 2008 HBS (see Annex A for methodological details) show that an increase in the unemployment rate by 3.6 percentage points (i.e., equal to that experienced in Georgia from 2007 to 2009) would reduce mean consumption per capita by 2.1 percent overall, by 3.3 percent in Tbilisi, and by 2.9 percent in other urban areas (Table 3). The estimated impact on the poverty headcount is an increase by 2.0 percentage points in Tbilisi and 1.2 percentage points overall, while the impact on the poverty gap is an increase by 2.5 percentage points in Tbilisi and 1.7 percentage points overall. Because the simulations only estimate the impact of complete loss of employment, it is likely that the impact on consumption through reduced working hours and substitution into lower paying jobs is even greater. Table 2: Main problems faced by households Households that considered this a Households that considered this the problem in their household (percent main problem of the household of population) (percent of population) Rural Urban Total Rural Urban Total Unemployment of household 71.6 63.1 67.3 42.85 41.52 42.2 members Buying medicines 73.0 55.6 64.2 14.26 12.26 13.3 Medical services 64.4 48.0 56.1 15.92 9.05 12.5 Housing conditions 44.1 36.5 40.2 8.4 10.24 9.3 Hunger / malnutrition 41.8 22.1 31.8 10.4 4.25 7.3 Paying debt / bank loans 19.2 24.6 21.9 4.62 9.81 7.2 Paying utility charges 19.6 47.4 33.7 0.82 9.19 5.0 Leisure, entertainment 19.7 32.0 25.9 0.77 2.73 1.8 Buying clothes 37.0 23.8 30.3 1.25 0.06 0.7 Buying school items 21.8 14.1 17.9 0.45 0.28 0.4 Furniture 22.9 18.8 20.8 0.25 0.59 0.4 None of those listed 2.2 4.7 3.5 Source: World Bank staff calculations from UNICEF 2009 survey Table 3: Simulations of the impact of increasing unemployment on consumption and poverty Tbilisi Other Urban Rural Total Mean consumption (% change from baseline) 1 percentage point increase in unemployment -0.9 -0.8 -0.2 -0.6 3.6 percentage point increase in unemployment -3.3 -2.9 -0.9 -2.1 5 percentage point increase in unemployment -4.5 -4.1 -1.2 -3.0 Poverty headcount (pct point change from baseline) 1 percentage point increase in unemployment 0.6 0.5 0.2 0.3 3.6 percentage point increase in unemployment 2.0 1.9 0.6 1.2 5 percentage point increase in unemployment 2.8 2.6 0.8 1.7 Poverty gap (pct point change from baseline) 1 percentage point increase in unemployment 0.7 0.7 0.2 0.5 3.6 percentage point increase in unemployment 2.5 2.6 0.8 1.7 5 percentage point increase in unemployment 3.6 3.7 1.1 2.4 Source: Staff calculations from Georgia HBS 2008. 8 19. Simulations based on macro sectoral growth rates suggest that reduced employment and income have increased the poverty headcount by 2 percentage points. For a different perspective on the poverty impact from income losses a second set of simulations applied the sector-specific 2008–09 growth rates to the 2008 HBS data to estimate household income gains or losses depending on the sector of employment. Table 4 presents the results of these simulations, showing a 3.6 percent reduction in mean consumption and a 2.1 percentage point increase in the poverty headcount, which are somewhat larger than the estimated impact associated with the 3.6 percentage point increase in unemployment as reported in Table 3. Again, Tbilisi is the worst affected area, followed closely by other urban areas, with much smaller impacts in rural areas. The relatively low growth-poverty elasticity from this simulation, as well as that presented earlier, is due in large part to the relatively small share of employment earnings in household income for lower-income households. According to the 2008 HBS, the mean share of employment earnings is 35 percent of household income, with 43 percent coming from pensions and transfers, and the balance as income in kind. Another reason for the low growth- poverty elasticity is that the impact of the recession was spread more or less proportionately across the income distribution, rather than being concentrated at the lower end. Table 4: Simulations of the impact of GDP contraction on consumption and poverty (2008–09) Poverty headcount Poverty gap Mean consumption (percentage point (percentage point (percent change change from change from from baseline) baseline) baseline) Total -3.6 2.1 1.0 Tbilisi -5.6 2.0 1.6 Other Urban -5.0 1.9 1.7 Rural -1.4 0.8 0.4 Agriculture sector 0.2 -0.3 -0.1 Manufacturing sector -5.1 4.2 1.4 Services sector -7.2 4.9 2.5 Source: Staff calculations from Georgia HBS 2008 20. The poverty impact of the recession is greatest on those employed in the services sector, followed closely by the manufacturing sector. The sharp contraction in the services sector—especially retail and wholesale trade, transport, telecommunications, and financial intermediation—reduced mean consumption of households employed in that sector by 7.2 percent, increasing the poverty headcount in the sector by 4.9 percentage points. The manufacturing sector saw smaller losses, with estimated mean consumption decreasing by 5.1 percent and the poverty headcount increasing by 4.2 percentage points. 21. Remittances are only a small source of income in Georgia, and therefore do not appear to have a large impact on poverty. The 2009 UNICEF questionnaire does not capture remittances as a distinct source of income, so it is not yet possible to determine the proportion of households in that survey that receive remittances or the levels received following the financial crisis. Using HBS and LSMS data, the 2008 Poverty Assessment indicates that in 2007 remittances accounted for only 3.6 percent of household monetary income. This is a significant decline from the 8 percent registered in the 2003 HBS, and the poverty assessment attributes the reduction in large part to various restrictions imposed by Russia, including deportation of Georgian workers, reduced flights, and restrictions on seasonal workers 9 from Georgia.5 The 2007 LSMS data also show that remittance receipts are skewed toward higher income households, as remittances account for 5 percent of total income in the richest income quintile, but only 1 percent of total income in the poorest quintile. Remittance income is also slightly more common among urban households than rural households. 22. Loss of remittance income during the crisis has affected a small proportion of households, and has not disproportionately affected poorer households. As noted in Figure 1, of the 49 percent of households who reported that their economic situation had worsened between mid-2008 and mid- 2009, only 9 percent reported loss of remittance income as one of the top three causes. Richer households were slightly more likely to list loss of remittances as a source of hardship, with the percentage rising from 7 percent in the poorest quintile to 11 percent in the richest quintile. Credit and debt 23. Increased borrowing during the rapid growth years left Georgian households vulnerable to the financial shock in 2008. As seen in Table 5, the gross loan portfolio increased by 42 percent between 2007 and 2008. Household loans—including mortgages, credit card debt, and other consumer credit— increased even more rapidly as credit became more widely available. A large share of the new borrowing was denominated in foreign currencies and at higher interest rates. Households were thus exposed on three fronts when the financial crisis hit, facing higher interest rates, higher foreign exchange rates, and lower incomes. Table 5: Household loans and total lending in Georgia, 2007–2010 Dec 31, 2007 Dec 31, 2008 Dec 31, 2009 March 31, 2010 Georgian Lari (GEL thousands) 594,519 860,637 576,784 601,862 Foreign currency (GEL thousands) 970,018 1,607,795 1,429,179 1,424,806 Total (GEL thousands) 1,564,537 2,468,432 2,005,963 2,026,668 Gross loan portfolio (GEL thousands) 4,589,340 6,530,298 5,185,336 5,343,553 Share of household loans 34.1% 37.8% 38.7% 37.9% Average lending rate – Georgian Lari 21 23 24 23.3 Average lending rate – Foreign currency 17 18 19 17.9 Exchange rate 1.5916 1.667 1.6858 1.7494 Source: National Bank of Georgia 24. A large number of households assumed new debt during the crisis. The UNICEF survey shows that a total of 35 percent of households took out new loans during the June/July 2008 and June/July 2009 period, with almost equal frequency in rural and urban areas (37 and 34 percent, respectively). Better off households were slightly more likely to have taken out loans, with the percentage of households ranging from 30 percent in the poorest quintile to 41 percent in the richest. Overall, 41 percent of the loans were borrowed from banks or pawn shops, 31 percent from relatives and friends and 20 percent as store credit from shops. Eleven percent of households tried to borrow but were not successful. 5 The 2008 Poverty Assessment noted that the percentage of households reporting remittance income was much higher in the 2003–06 HBS data than in the 2007 LSMS, but could not explain the discrepancy. However, the discrepancy between the 2007 LSMS and the 2008 HBS is much smaller. According to the 2008 HBS, 7 percent of households received remittances, and these constituted 5 percent of total household income. The 2008 HBS also corroborates the 2007 LSMS results showing remittances are more common among richer and urban households. 10 25. Assessing the poverty and distributional impacts of the shock to credit markets would require further analyses. The available survey data (HBS, LSMS, UNICEF) and relevant documents such as the poverty assessment include little information about credit transactions such as the percentage of households carrying debt at the onset of the crisis as well as the amounts and terms of that borrowing. A deeper analysis of longer-term indebtedness of households would be warranted but suitable data for this task may be elusive. Coping strategies 26. A majority of the households whose economic situation had worsened reported no source of additional assistance to offset the shock. Of the one-half of households whose situation deteriorated, 59 percent said that they did not receive additional public or private assistance,6 nor were they able to borrow or draw down assets. This problem is more acute in rural areas, where 64 percent of households have no additional means of support, compared to 54 percent in urban areas. For households that had some form of fallback support, the most common source was assistance from private individuals (usually relatives or friends) (20 percent), followed by borrowing (10 percent), dissaving or selling assets (7 percent), and public transfers (5 percent).7 As shown in Figure 3, urban households that were facing economic hardship had better access to private assistance (transfers) from individuals, as well as borrowing or dissaving. Rural households, in contrast, were more likely to cite public social assistance or pensions as their source of support. Based on post-crisis consumption quintiles, poorer households are less likely to have additional means of support albeit we cannot assess whether these households dropped to the lower quintiles because of the lack of coping mechanisms, or if they were already in the lower quintiles before the economic downturn. Figure 3: Main sources of support for households experiencing deteriorating economic conditions 120 Assistance from private organizations 100 Public social assistance Percent of Households 80 Dissaving, selling or renting assets 60 Borrowing 40 Assistance from private individuals 20 0 None Rural Urban Total 6 Among those households whose situation had worsened and stated that they did not receive additional assistance, 55 percent received pensions and 5 percent received targeted social assistance. The survey asked those whose situation had deteriorated about “additional sources of livelihood.� It is likely that respondents did not consider increased benefit levels (see section 4) as an additional source of assistance. 7 Sources sum to more than 100 percent because of possible multiple sources of support and exclusion of minor sources of support. 11 27. Whether or not they had additional assistance, most households altered their consumption patterns in response to the economic downturn. A majority of households reported that they stopped buying certain food and nonfood items, or turned to less expensive alternatives, or both. Expenditures on services were also reduced, including expenditures on medical treatment. Some rural households increased their subsistence food production. IV. Georgia’s Social Policy Response During the Crisis 28. Georgia’s social assistance programs have played a significant role in reducing poverty. Like most countries in the ECA region, Georgia has several forms of social assistance. The major programs include the pension system, targeted social assistance (TSA) for the poorest households, targeted medical assistance for the poor, and aid to internally displaced persons (IDPs). As of 2007, prior to the crisis, social transfers accounted for less than 5 percent of GDP. Although this was low compared to EU countries, in which social transfers represent approximately one fifth of GDP, Georgia’s social transfer programs have wide coverage, reaching about two-thirds of the population. Georgia does not have an unemployment insurance program. 29. Georgia has undertaken and continues to pursue significant reforms to strengthen and expand its social assistance programs. Pensions are a benefit that accounts for almost 75 percent of social transfers, and are received by slightly over half of all Georgian households. Since 2004 large arrears in the pension system have been cleared, minimum contributions have been abolished, categories were streamlined, and benefit levels significantly increased. TSA was introduced in July 2006 and by 2009 represented 14 percent of social protection spending. TSA uses a proxy means test to target poor households, using a database of 1.7 million registered (self-identified) individuals, or nearly 40 percent of the population. 30. In response to the economic crisis, the main policy response from Government was to increase social transfers. Social transfers’ share of GDP increased from 4.6 percent of GDP in 2007 to 6 percent in 2008 and an estimated 7 percent in 2009; it is projected to remain close to 7 percent in 2010. In 2009 the monthly pension benefit was increased from GEL 70 to 80. Also in 2009, the TSA monthly top-up per additional family member was doubled from GEL 12 to GEL 24, which is in addition to a base benefit per household of GEL 30. In the wake of the crisis TSA coverage expanded from about 131,000 households in December 2008 to about 155,000 households in April 2009, and to over 164,000 by April 2010. Figure 4 shows this expansion in terms of number of beneficiaries – an increase from 400,000 in July 2008 to nearly 470,000 in April 2010. Already having a targeted safety net program in place clearly facilitated this policy response, and gave Georgia a head start in addressing the impact of the crisis on vulnerable households. Ongoing efforts to improve the targeting performance are also notable. 12 Figure 4: Trends in TSA program applicants and beneficiaries through the crisis 2,000,000 1,800,000 1,600,000 1,400,000 1,200,000 1,000,000 800,000 600,000 400,000 200,000 - Mar-08 Mar-09 Mar-10 May-08 May-09 Dec-08 Dec-09 Oct-08 Oct-09 Jan-08 Jul-08 Jan-09 Apr-08 Aug-08 Apr-09 Jul-09 Jan-10 Aug-09 Apr-10 Feb-08 Feb-09 Feb-10 Sep-08 Sep-09 Jun-08 Nov-08 Jun-09 Nov-09 Applicants for TSA in database # of TSA beneficiaries Source: SSA 31. During the crisis, Targeted Social Assistance (TSA) and pensions both contributed significantly to protect lower income groups. According to the UNICEF household survey data available, 8.1 percent of the population lives in households that receive TSA. This is a slight increase over the 6.7 percent in 2007 that was reported in the Bank’s most recent poverty assessment (World Bank 2008). The UNICEF survey shows that among households that receive TSA, the average amount received per household is GEL 78 per month. In contrast to TSA, pensions are not targeted, and are received by more than one-half of the households in Georgia. The standard monthly pension benefit is GEL 80 per man over 65 and woman over 60, funded through general taxation. The UNICEF survey reports that 52 percent of the population lives in households that receive pension income, which is comparable to the 53 percent reported in the latest poverty assessment. Among those that receive pensions, the average amount received per household is GEL 114 per month (UNICEF 2009). 32. Today, TSA is better targeted to poorer households, but pensions are larger than TSA across the income distribution. Table 6 shows the mean consumption per capita by consumption decile, using total consumption, total consumption minus TSA receipts, and total consumption minus pension receipts. Although there is some minor leakage of TSA to the upper deciles, the TSA benefits are clearly concentrated in the poorer deciles. However, because the coverage of TSA is low, on average TSA adds only GEL 3–4 per person per month even in the poorer deciles. In contrast, pensions add GEL 11 –15 per person per month even in the poorest deciles, rising to GEL 19–23 per month in richer deciles. 13 Table 6: Mean consumption by decile (2009 GEL per capita per month) Consumption decile 1 2 3 4 5 6 7 8 9 10 Consumption 28.7 48.0 59.7 70.8 83.0 97.5 118.8 146.6 193.6 372.9 Consumption minus TSA 25.1 44.0 56.8 68.4 81.3 96.4 116.9 145.7 193.1 372.6 Consumption minus Pensions 17.4 33.6 44.5 52.8 66.9 78.9 99.9 127.7 170.5 356.0 Source: World Bank staff calculations from UNICEF data (2009) 33. Because TSA targets the poorest households, most of TSA’s poverty reduction impact is seen in the poverty gap, not the poverty headcount. This depends, of course, on the poverty line that is chosen. Figure 5 shows the cumulative distribution of consumption, or poverty incidence curve, for Georgia using the 2009 UNICEF data. The vertical line is the Georgia national poverty line of 89.7 GEL per adult equivalent per month. At that poverty line, the poverty headcount without TSA is almost two percentage points higher than the headcount with TSA (27.5 versus 25.7).8 The difference in the poverty gap is slightly greater than two percentage points (9.6 versus 7.5). As pensions are larger in value and reach a much wider population, poverty levels would be much higher without pensions, with the headcount and gap reaching 39.1 and 18.6, respectively. Figure 5: Cumulative distribution of consumption with and without pensions and TSA Consumption minus pensions 60 Consumption minus TSA Consumption 50 % of Population 40 30 20 10 0 -25 0 25 50 75 100 125 150 Monthly consumption per adult equivalent (GEL 2009) Source: Staff calculations from 2009 UNICEF survey data 8 Implicit assumptions are that (a) TSA and pension all go to consumption, and (b) if TSA or pension income is lost then it isn’t replaced by other income such as private transfers or employment. Note that because of differences in methodology of the UNICEF survey, these poverty rates are not directly comparable to the poverty rates estimated previously using the HBS or the LSMS data. 14 34. Coverage of both pensions and TSA is higher in rural areas, with a correspondingly larger effect on poverty reduction. As shown in Table 7, rural households are more likely than urban households to receive TSA and pensions. The average amount of TSA received is higher in rural areas, while average pensions received is slightly higher in urban areas. Because of the higher rural coverage, pensions and TSA both help keep a higher percentage of rural households from falling below the poverty line, as well as reducing the poverty gap. Pensions and TSA thus also reduce rural-urban inequality. Table 7: Incidence of pensions and TSA by area (rural/urban) Rural Urban Total Households receiving TSA (%) 12.5 3.8 8.1 Average amount received per recipient household (GEL/month) 81 71 78 Households receiving pensions (%) 58.7 44.7 51.6 Average amount received per recipient household (GEL/month) 113 116 114 Poverty headcount (using 89.7 GEL poverty line) Consumption 31.5 20.1 25.7 Consumption minus TSA 34.3 20.9 27.5 Consumption minus pensions 48.5 30.0 39.1 Poverty gap (using 89.7 GEL poverty line) Consumption 8.3 6.8 7.5 Consumption minus TSA 11.4 7.9 9.6 Consumption minus pensions 21.7 15.5 18.6 Source: World Bank staff calculations from UNICEF 2009 data V. Future policy options for crisis mitigation: Georgia and the ECA region 35. A range of policy options is available for mitigating the crisis impact on households. The first imperative is to ensure that overall economic policies are aligned with a resumption of growth, by pursuing stable macroeconomic policies while promoting a healthy business environment. On this front Georgia’s performance has been strong (as indicated, for example, by the Doing Business indicators). However, there are also more specific policy options that can make an important contribution to lessening the impact of the crisis on poor and vulnerable households. 36. While the Georgian response to the crisis through increased social assistance and pensions has been broadly successful, these programs lack the flexibility of other policy options. Neither the proxy- means tested TSA, nor age-based pensions, offers the benefit of ‘self-adjustment’ as in the case of certain safety net measures applied elsewhere in the region. The experience of others with such self- adjusting safety net programs and risk mitigation measures offer possible lessons for Georgia going forward. Such alternatives tend to create fewer fiscal contingencies for the future than TSA or pension increases. However in Georgia, the near-absence of specific labor market policies and programs meant that the response had to focus largely on expanding these transfers. 37. As in Georgia, some countries also opted for higher transfers (e.g., Latvia, Macedonia), but most have decided to pursue a wider range of policy responses, including self-adjusting measures. These include a combination of unemployment insurance (as done in Estonia and Russia); conditional 15 income transfers for the unemployed (e.g., paid training programs for the unemployed as in Turkey); public works programs (e.g., Kazakhstan, Latvia); steps to improve access to health and education by the poor; and active labor market programs (Box 2). 38. From the extensive list of responses deployed across the region, unemployment insurance (UI) programs are notable for their broadly successful performance. In many countries they were the first and most significant government response to help households impacted by an income shock, particularly the newly unemployed. Thus they appear to have fulfilled their role as ‘automatic stabilizers’ (and as ‘faucets’ that are easy to turn on and off). However, in some countries low coverage rates of UI programs dampened their overall effect.9 Box 2: Active labor market programs in ECA Elsewhere in the ECA region, labor market policy measures – embracing four broad types of interventions – have been among the most common responses to the crisis. The first type of intervention includes measures to create new or temporary jobs, such as public works, wage subsidies for new entrants, reductions of non-wage labor costs (e.g., social security taxes), or providing start-up grants. The second type includes measures to preserve and protect existing jobs, such as wage subsidies, tax concessions, or re-training programs. The third category includes measures to enhance employability in the event of job loss, such as assistance with job search, training and re-training, and mobility allowances. The fourth are direct income support programs, such as an increase in unemployment benefits or extension of payment duration.10 All have been used in ECA during the crisis. However, it should be noted that some of these measures (e.g., wage subsidies) come with their own significant fiscal costs, and evidence on the effectiveness of others (e.g., training programs) is mixed. 39. Looking forward, in addition to rejuvenating growth but in the absence of a broader array of public policy instruments, a reliance on expanded social expenditures such as TSA and pensions would require additional (and scarce) fiscal resources. For example, an increase in monthly TSA benefits by 25 percent (or, equally, an increase in coverage by 25 percent) would imply a fiscal cost of about GEL 38m annually, or 3.3 percent of 2010 social protection spending (0.6 percent of the overall budget). By way of comparison, increasing the monthly old-age pension benefit from GEL 80 to GEL 100 would cost about GEL 175m annually, which is likely beyond the means of government in the current fiscal environment. Clearly, the better targeting of the TSA program would make it a preferred channel for reaching the poor with the limited budget funds available. The extent to which additional social spending is still required to address the crisis will need to balance the pace of growth acceleration with the reality that unemployment tends to be sticky even as growth resumes. 40. Over the medium-term, a significant agenda of reforms related to amplifying pro-employment policies and social assistance policies that support the poor during crisis situations could be considered. While Georgia has opted against pursuing an active labor market agenda in recent years, as its social policy evolves and lessons of the crisis emerge (both at home and in the region), a full range of policy options may be revisited. As a concrete first step in an agenda of bringing more people into work, it may be possible to capitalize on the strong achievements in establishing local SSA centers by using these to link the poor (TSA beneficiaries) with job openings and training possibilities in the post-crisis period. Further, amplifying the existing social assistance program with self-adjusting safety nets, such as 9 “Social impacts and responses to the 2009 recession in ECA� (2010), World Bank, forthcoming. 10 This framework draws on Kuddo, Arvo (2009), “Labor market monitoring in ECA: Recent trends�. World Bank mimeo. 16 through public works, might be an option to consider. Many ECA countries are also exploring alternative ways to encourage the employed to save for crisis periods, such as through mandatory unemployment savings accounts (these are now under consideration in Turkey, and have been implemented in countries such as Chile). Also, enhancing skills through education reforms and vocational training (possibly in the form of conditional unemployment assistance) is another potential avenue for medium- term reform. 41. As a final note on policy lessons from the crisis, one area where Georgia’s performance has been significantly lagging (for example, compared to Armenia) has been the implementation of household surveys to monitor socioeconomic trends. Long-standing concerns about the reliability of data produced by the household budget surveys of the Statistics Department (now GeoStat) have posed serious challenges to monitoring the impact of the crisis on households. This is a potentially important tool for informing policies that has been neglected in recent years. The Armenian experience, for example, is instructive of what is possible: a regular quarterly survey with strong field implementation that included an additional module on coping strategies during the crisis yielded a rich database on the impact of the crisis on households that offered just-in-time insights for policy-makers. This would be a valuable area for future investment in Georgia. VI. Conclusions 42. Following a strong record of economic growth and improved governance from 2004 –08, Georgia has been set back by the global economic crisis. The broad economic downturn came on the heels of the food and fuel price shocks in 2008, and the armed conflict with Russia in August 2008. While Georgia’s progress in recent years provided a more stable base to weather the economic crisis, it remained vulnerable to the multiple shocks. This is true particularly among the poorer members of Georgian society. 43. Georgia has been particularly hard hit through the credit and labor markets. In the only household survey that is available since the onset of the crisis, debt repayment stands out clearly as the largest problem faced by households. However, data on household credit and debt are extremely limited, and collecting more information on the nature of the debt, including repayment terms, interest rates, currency of the loan, and scope for rescheduling should be a priority. 44. Unemployment has increased to 16.9 percent in 2009, up from 13.3 percent in 2007. In addition to the increase in official unemployment, many employed workers are suffering welfare losses through reduced hours, downward wage pressure, and taking jobs that are below their skill level. Simulations indicate that increased unemployment and a decline in labor earnings more generally because of the economic crisis have increased the poverty headcount in Georgia by 1 to 2 percentage points. The negative poverty impact of reduced labor earnings is greatest in Tbilisi and other urban areas. Unlike many other countries in the region, Georgia is not heavily dependent on remittance income, so this has not been an important factor behind the increase in poverty. 45. Georgia’s system of social transfers, especially pensions and Targeted Social Assistance (TSA) are vital for maintaining basic levels of well-being. Pre-crisis social policy reforms, such as introducing TSA and streamlining pensions, have proven to be valuable investments. As noted, pension benefits have wide coverage, and without them 39 percent of the population would be below the poverty line instead of 26 percent. TSA is a much smaller program, but as it is targeted to the poorest part of the 17 population, it is crucial for raising the minimum living standard of the extreme poor in Georgia. However, TSA still does not cover a large segment of the poor in Georgia, so continued expansion, while maintaining or improving targeting efficiency, should be considered, although this needs to be done in a manner consistent with aggregate fiscal constraints. Given the economic uncertainties, the government may also consider alternative counter-cyclical mechanisms, including labor market policies (such as unemployment insurance, public works schemes, or conditional unemployment assistance benefits) that provided an effective response in many other countries in the region. Potential policy initiatives in the labor market arena should be given consideration in the post-crisis environment. 46. These findings provide an initial assessment of the impact of the crisis on well-being and poverty, and also lay out another layer of policy-relevant questions that require further analysis. A few areas deserve to be highlighted here. First, it is necessary to gain a better understanding of the impact on the labor market. What are the characteristics of the newly unemployed, such as age, sex, experience level, sector of employment, and geographic location? To what extent have those who are still unemployed been required to accept fewer working hours, reduced compensation, or employment in lower paying jobs? Second, what is the distribution of the credit shocks? Who has been most affected by tightening credit and higher borrowing costs, and who has defaulted on their loans? What, if any, avenues are open for rescheduling of household debt? Third, how has poverty actually evolved during the crisis (as opposed to simulated evidence as necessitated by incomplete data)? These and other pertinent topics will be taken up in subsequent work, particularly as new data such as the 2009 HBS become available. 18 Annex A: Simulation Methodology The 2009 UNICEF survey includes only limited employment data, and the 2009 HBS data are not yet available. Therefore micro-simulation techniques were used with the 2008 HBS data to gauge the impact of unemployment and reduced earnings on household consumption and poverty levels. Two different simulation approaches were used. The first approach uses Monte Carlo techniques to estimate the impact of increased unemployment, and the second uses macro sectoral growth rates to estimate the change in employment income following the onset of the crisis. This annex describes the simulation techniques. Unemployment simulations In the unemployment simulations, the negative impact of unemployment is borne by those households in which one or more members loses his or her employment, and other households are unaffected. Employed persons are defined as those who reported having worked for compensation during the 7 days immediately preceding the survey interview. A given percentage of employed persons (1 percent and 5 percent in the simulations presented in this note) were randomly selected to become unemployed. The employment income from each person made unemployed by the simulation was then subtracted from total household consumption, creating a new distribution of consumption. The mean consumption of this new distribution was calculated, as was the corresponding poverty headcount and poverty gap. This process was repeated 1,000 times, each time drawing a new random subset of “simulated unemployed,� and new simulated consumption vectors. At the end of the 1,000 simulations, the mean consumption and poverty estimates were averaged over the 1,000 simulations. These averages are the figures reported in the note. The unemployment simulations include at least three simplifying assumptions. One is that the income loss only comes through open unemployment, i.e., income loss from reduced working hours or substitution into lower paying jobs is not taken into account. A second assumption is that all workers have an equal chance of becoming unemployed. A third is that losses in employment earnings translate one-for-one into reductions in household consumption. Macro sectoral growth rate simulations The second simulation approach uses sector-specific GDP growth rates for the period immediately following the onset of the financial crisis, and applies that change in income uniformly to all workers in that sector. Whereas the unemployment simulations allow for changes both within and between sectors of employment, the sectoral growth simulations only allow for changes between-sectors. Thus any distributional impact arises entirely from the way in which workers in different sectors are spread across the income (consumption) distribution. The sector-specific growth rates are for the period from the end of the second quarter in 2008 to the end of the second quarter in 2009, which covers the 12-month period when the Georgian economy contracted the most. The rates are shown in Table A1. As in the unemployment simulations, the 2008 HBS was used to identify employed workers based on employment status in the week preceding the interview. The survey collected information about sector of employment using the same classification as Table A1, so matching workers to sector was straightforward. 19 After matching workers, employment income was increased or decreased by the sector-specific proportion shown in Table A1. Household consumption was increased or decreased by a corresponding amount, and a new consumption distribution was generated. The mean of this simulated distribution is reported in the note, as well as the poverty headcount and poverty gap. Table A1: Real GDP growth rates by sector, end-June 2008 to end-June 2009 (end-June 2008=100) Agriculture, hunting and forestry; fishing 93.1 Mining and quarrying 118.3 Manufacturing 80.3 Electricity, gas and water supply 109.7 Construction 99.1 Wholesale and retail trade; repair of motor vehicles, motorcycles and personal and household 68.7 goods Hotels and restaurants 87.0 Transport and communication 94.9 Financial intermediation 84.4 Real estate, renting and business activities 95.7 Public administration 104.1 Education 103.3 Health and social work 108.2 Other community, social and personal service activities 74.7 Private households employing domestic staff and undifferentiated production activities of 100.9 households for own use This simulation also involves simplifying assumptions. The strongest is that the impact is felt uniformly across the sector. A second assumption is that employment incomes in a sector follow real sectoral GDP perfectly. Third, the simulation does not allow for movements of workers from one sector to another. Fourth, the simulation assumes that changes in employment income map one-for-one to changes in household consumption. 20