Protecting the Poor and Vulnerable 1

Transcription

1 Emil Tesliuc, Lucian Pop and Richard Florescu INDEX Protecting the Poor and Vulnerable 1 I. Overview of the Social Protection System II. The Impact of Social Protection Programs on Income Distribution III. The Impact of Social Protection Programs on Poverty Reduction A. Coverage of Social Protection Programs B. Distribution of Social Protection Benefits across groups and Benefit Adequacy C. Marginal Benefit Incidence Analysis D. Cost Benefit Analysis E. Overall Effectiveness of Social Protection Programs in Reducing Poverty F. Improving Social Assistance Administration IV. Summary of Key Issues and Policy Recommendations Bibliography Statistical Annex Annex 1. Main Social Assistance Cash Transfers during Annex 2. Unemployment Benefits Annex 3. Social Insurance Benefits The authors are indebted to Gordon Betcherman, Nicholas Burnett, Margaret Grosh, Cem Mete, Stefano Scarpetta and Quentin Wodon for their guidance and advice. We thank the Quality Team of the Social Protection Anchor, Randa el-rashidi in particular, for organizing a Quality Enhancement Review during the early stages of the paper, in our benefit. The institutional review of the main social protection programs, annexed to the paper, was prepared by team of experts from the Ministry of Labor and Social Solidarity, the National Pension Authority, and the National Employment Agency. We thank Mr. Petre Ciotlos, Mr. Dumitru Calinoiu and Mr. Ioan Cindrea for their help. The views expressed here should not be attributed to the World Bank, its Executive Directors, or the countries they represent. All errors are ours. 118

2 I. Overview of the Social Protection System This chapter describes the structure and reviews the effectiveness and efficiency of Romanian main social protection programs in fighting poverty and vulnerability. Social protection is a set of public interventions that assist individuals, households and communities to manage risk better and provide support for the critically poor (Holzmann and Jorgensen, 2001). In Romania, the protection of the poor and vulnerable was a constant priority of successive governments, as witnessed by the large share of resources mobilized through the system in the last eight year, and its anticyclical role. The Romanian state redistributes about 10% of GDP via social protection programs. During , the share of social protection transfers in GDP was remarkably stable, between 9% and 11% of GDP (Table 1). During the recession of , the Romanian social protection system acted like a social-shock absorber, smoothing the social costs of economic transformation. This anti-cyclical role is not common; in many countries and regions social protection programs tend to contract during time of crises, when they are most needed. This performance was achieved in paralle l with a process of decentralization of social assistance activities. The share of social assistance activities (all funds except for social insurance and unemployment) financed from the local budget increased from 16% in 1995 to 31% in Table 1. Share of Social Protection in GDP, and Social Protection Financing by Source Year Social Protection Source of Funds for Social Protection Programs Total Unemployment % GDP State Budget Local Budgets Social Insurance Fund Source: Consolidated Government Budget, Ministry of Finance, Bucharest The main social protection programs or policies can be grouped into: (i) social insurance; and (ii) social assistance. The social insurance systems consist mainly of (i) pensions for former employees or farmers (for old age and invalidity) and their dependents (survivors), and (ii) unemployment benefits. Besides these benefits, the social insurance system provides a wide range of benefits and services for contributors, such as: maternity and child-raise leave, sick leave, funeral benefits, as well as severance payments and active labor market measures. In 2002, these programs channeled 7.6% of the GDP. From social risk management perspective, the social insurance system mitigates a number of social risks, such as the risk of unemployment, accident, disability or lack of earning capacity during old-age. The public pensions system is a classical PAYG scheme, which in spite of the reforms introduced in 2001, continues to face a chronic deficit (close to 1% of the GDP). The deficit is the result of (i) a very low dependency rate caused by population aging and a shrinking number of employees, and (ii) past early retirement policies. To maintain the fiscal balance of the system, the administrators opted for low replacement rates for pensioners (the ratio of the average pension to the average wage is around 37%) that for many pensioners are too small to protect them against poverty. Currently, the Government implements a three year re-correlation plan to restore the equity among various cohorts of pensioners 119

3 which retired with significantly different pension levels for similar contribution terms. Table 2 presents the structure of contributory pension s benefits, number of beneficiaries and the associated costs for Annex 3 reviews the legislative and institutional aspects of the system. Table 2. Pension Benefits in 2002 Number of beneficiaries (thou) GDP share (%) Total Social Insurance benefits 7.09 Social Insurance Pensions, of which 6, State social insurance pensions 4, Farmers pensions 1, Social Aid Pension War veterans pensions Source: Government of Romania, Bucharest The unemployment benefits provided by the unemployment insurance system were rationalized beginning with 2002, when a new legislation was enacted. The benefit level is set for 75% of the minimum gross wage and is granted for a period of 6 to 12 months depending on the length of service. In addition, severance payments are granted for collectively dismissed workers, their level being linked to the previous average wage and the duration to their length of service. Table 3 presents the structure of the unemployment benefits, number of beneficiaries and the associated costs for Beside the cash benefits, the unemployment fund finances a wide range of active labor market measures, including job counseling, public works and micro-credit programs. Annex 2 reviews the legislative and institutional aspects of the system. Table 3. Unemployment Benefits in 2002 Number of beneficiaries GDP share (%) Unemployment benefit 226,949/ month 0.29 Support allowance 217,379/ month 0.07 Graduates allowance for vocational integration 52,275/ month 0.03 Severance payments 19,000/ year 0.08 Source: Government of Romania, ANOFP, Bucharest The social assistance system includes a number of cash benefits (universal and targeted), subsidies and services. There are two main targeted cash benefit programs: the Minimum Income Guaranteed (MIG) benefits and heating subsidies, which are granted to the households having an aggregate income below certain threshold, established and periodically updated by the Government. These two programs were initially (2002) regulated by the same law, but after the first year of implementation they were split into two separate programs. Together, the two programs channeled about 0.35 % of the GDP in 2002 (compared with a budgetary allocation of 0.40%). Annex 1 reviews the legislative and institutional aspects of the system. The MIG program (means-tested), enacted in 2002, replaced the Social Aid Program implemented from 1995 to 2001, which due to its poor financing, design and implementation became ineffective. Eligibility for the MIG is gained for applicants that pass an income and asset test. The income threshold is function 120

4 of family income and size. The MIG benefit covers the gap between the program threshold and the actual family income. For able-bodied family members, benefits are conditioned by a workfare requirement, an attempt to self-target program benefits to those in need. In 2002, the program covered almost 619 thousand families, for a total cost of 0.28% of the GDP. At the end of 2002 the number of families benefiting of MIG was about 380 thousands, or about 5.4% of the country s population. Program beneficiaries are entitled to two other tied-benefits: health insurance and heating subsidies (see below). The Heating Subsidy Program provides lump-sum benefits for low income families during the cold season (November to March), the size of the benefit depending on the aggregate income level of the family and the source / type of fuel used for heating (district heating, gas or wood/coal). For households not connected at the heating grid, the benefits that are paid as a lump-sum or in monthly installments. For those households connected to the heating grid, the benefits are deposited in escrow accounts, from where they are accessed by the district heating suppliers. In 2002, almost 756,000 families (3,023,048 persons) benefited of this program, covering 13.5% of the country s population, for a total cost of 0.1% of the GDP (included in the MIG budget). Initially, the heating subsidies where provided only for MIG beneficiaries. In January and September 2002 the GOR issued ordinances modifying the MIG law, and raised the heating subsidy eligibility threshold above the MIG threshold in an attempt to cover a larger share of the population. By far, the biggest share of the social assistance transfers, are represented by the State Child Allowance and the Supplementary Allowance for Families with More Children. These benefits were granted to 4,835,606 children (state allowance) and 1,022,900 families (supplementary allowance), at a cost of 0.68% of GDP in The State Child Allowance is an universal benefit, granted monthly for each child up to the age of 16 (18 if enrolled in regular secondary education system), provided those over the age of seven are attending the school classes on a regular basis. Begin ning with January 2003, the level of the benefit is set for 210,000 ROL/month. The level of the benefit was periodically indexed, to protect its purchasing power against inflation. At the same time, families with two or more children are entitled to a Supplementary Child Benefit. The level of the benefit was set in 1997 at 40,000 ROL/month for a family with two children, 80,000 ROL/month for a family with three children, and 100,000 ROL/month for a family with four or more children. The benefits were not indexed after The supplementary allowance was introduced in an attempt to improve the targeting of the program to the poor, knowing that families with more children face higher risk of poverty. However, there were two inconsistencies between this obje ctive of the program and the way it was designed and implemented, that worked against its targeting performance. First, the program had a lower marginal benefit rate for families with 4 children (20,000 ROL/month) and provided no extra benefits for children rank five or higher. Thus, the program failed to cover the marginal income gap for families at higher risk of poverty, despite the low cost of expanding the program coverage for this group. Second, only the weakly targeted child allowance was indexed against inflation, while the better-targeted supplementary allowances lost its purchasing power through time. Table 4 presents the structure of the social assistance benefits, number of beneficiaries and the associated costs for

5 Table 4. Main Social Assistance Benefits in 2002 Number of Beneficiaries GDP Share (% ) FAMILY ALLOWANCES State child allowance 4,835, Supplementary allowance for families with children 1,022, Newborn grant 150, SOCIAL ASSISTANCE Social assistance benefit (MIG) 1,098, Residential heating allowances and facilities provided to households to 3,023, defray heating costs Allowance for foster care and family care 43, Emergency benefit Subsidies to associations and foundations 5, Grants 2, BENEFITS TO PEOPLE WITH DISABILITIES Social assistance allowance to people with severe or significant sight NA 0.08 impairment Monthly compensation NA 0.08 Personal assistant NA NA Source: Government of Romania, Ministry of Labor and Social Solidarity, Bucharest The coverage of the social protection system is extensive. Overall, 87% of the population is covered by at least one social protection transfer, directly or indirectly (as household members, through income and consumption sharing) (see Table 5, based on ABF ). The program with largest coverage are the child allowances (56%), followed by pensions (especially old-age [29%] and farmers [12%]), unemployment benefits (7%) and MIG (4%). The ABF confirms the ranking of social protection transfers, the largest programs being pensions (82% of reported social protection spending), followed by the child allowances (8.5%), the unemployment benefits (4%) and the MIG (2%). In general, the social protection benefits captured by households tend to fall in a narrow range, with low coefficients of variation. For comparability, Table 5 presents similar statistics for private transfers remittances received from other households, in cash or in kind. Overall, the benefits collected from private transfers reach 41% of the population and represent one quarter of total social protection benefits they are quite important. These remittances tend to occur between members of extended family relatives from the country side providing food to the urban relatives, with urban households providing cash support at times to their rural relatives. Most private remittances are in-kind (both as number of transactions and value). 2 ABF is the Romanian Acronym for the Household Budget Survey. See background paper 1 for a description of the survey. 122

6 Table 5. Receipt of Social Protection Benefits: Household-level Descriptive Statistics, HBS 2002 Program # of cases % covered Mean benefit Coef. Variation Share of SP benefits Social Protection, o.w.: % % Social Insurance pension - old-age % % pension - disability % % pension - survivorship % % pension - farmers % % pension - war veterans 145 0% % pension - social assistance 71 0% % unemployment benefit % % redundancy payments 27 0% % Social Assistance child allowance % % scholarship 123 1% % support for people with disabilities 614 2% % allowance for war veterans % % allowance for victims of political persecution 121 0% % social assistance provided by mayor's office (MIG in 2002) 808 4% % other social assistance benefits 171 1% % Remittances (Private Transfers) % % cash remittances % % in-kind remittances % Source: WB Staff Estimations based on ABF % During , the coverage of the social insurance system expanded slightly, due to the increase in the number of households benefiting from pensions (Table 6). The coverage of the unemployment benefit program was highly anti-cyclical, with the program expanding during the recession, and contacting thereafter. Most social protection programs maintained their coverage, with the notable exception of the MIG the program which inherited the Social Aid in 2002, where the coverage went up for 0.5% to 3.9% of the population. This expansion was mainly driven by increased funding for the program. In 1997, the share of other social assistance benefits rose from 1.2% to 14%, during the implementation of a weakly targeted bread allowance that accompanied the liberalization of the sector for six months. The bread allowance temporarily increased the coverage of the social assistance programs by 8-9 percentage points. 123

7 Table 6. Changes in the Coverage with Social Protection Programs, social insurance /contributory benefits 48.9% 46.5% 46.1% 49.4% 51.6% 51.4% 52.6% 51.3% pension - length-of-service 23.9% 24.4% 25.0% 25.7% 26.5% 27.6% 29.0% 29.2% pension - disability 5.6% 5.7% 6.0% 6.8% 6.8% 7.1% 8.6% 9.0% pension - survivor 5.2% 5.4% 5.4% 5.6% 5.9% 6.0% 6.4% 6.4% pension - farmer 12.1% 11.7% 11.5% 11.8% 11.6% 11.8% 12.1% 11.9% pension - war veterans 0.5% 0.4% 0.5% 0.4% 0.3% 0.3% 0.3% 0.3% pension - social assistance 1.0% 1.1% 0.6% 0.2% 0.2% 0.2% 0.2% 0.2% unemployment benefit 11.8% 7.5% 7.4% 10.0% 12.3% 10.9% 9.0% 7.0% redundancy payments % 0.1% and child care leave 1.6% 1.1% 1.1% 1.4% 1.3% 1.0% 1.0% 0.8% social assistance /noncontributory benefits 54.1% 52.6% 62.5% 56.5% 56.4% 56.5% 56.8% 57.9% child allowance 52.4% 50.9% 53.6% 54.8% 55.0% 54.9% 55.1% 55.5% scholarship 1.2% 0.9% 0.8% 0.7% 0.5% 0.8% 0.7% 0.6% support for people with disabilities 1.6% 1.8% 2.2% 2.1% 1.9% 2.0% 2.2% 2.2% Social Aid / MIG in % 0.5% 0.4% 0.5% 0.5% 3.9% other social assistance benefits (includes MIG in 95-96) 0.9% 1.2% 14.0% 1.0% 0.6% 0.6% 0.8% 0.7% other noncontributory benefits 3.6% 3.4% 3.0% 3.2% 2.8% 2.5% 2.6% 2.4% allowance for war veterans 3.4% 3.3% 2.8% 3.1% 2.7% 2.4% 2.3% 2.2% allowance for victims of political persecution 0.2% 0.2% 0.2% 0.2% 0.2% 0.1% 0.3% 0.2% Total SP 83.3% 81.5% 85.2% 85.5% 86.1% 85.9% 87.0% Source: WB Staff Estimations based on AIG and ABF % About 84% of households benefit from at least one social protection program. The gap in coverage with social protection program is low: about 7% of the poor and 24% of the non poor do not receive at least one type of transfer (Table 7). In contrast, one in five households receives more than one social protection benefit (23% for the poor, versus 15% for the non-poor in 2002). Table 7. Gaps and Duplications in the Coverage with Social Protection Programs by Poverty Status # Social Protection benefits / HHs Total Population No benefit One benefit Multiple benefits Non-poor No benefit One benefit Multiple benefits Poor No benefit One benefit Multiple benefits Total Source: WB Staff Estimations based on AIG and ABF II. The Impact of Social Protection Programs on Income Distribution The previous section highlighted the large redistributive effort occurring in Romania. Both socia l insurance and social assistance spending are designed to be redistributive: to transfer resources from current to past workers in the case of pensions, and from taxpayers to the poorest families in the case of targeted social assistance. How successful is this redistribution in reducing inequality? How are program benefits being 124

8 distributed? Who are more likely to capture the benefits of the social protection programs, the poorest or the richest? To determine if social protection programs are reducing inequality, or if the benefits are distributed to the poorest strata of the population, one should identify the welfare status of a households in the absence of the government intervention, or the counterfactual per adult equivalent consumption. Many benefit incidence studies subtract the entire amount of the transfer from household income or consumption to approximate preintervention welfare, and then rank the population into quintiles. Such approximation assumes, in fact, that there is no replacement of the lost income source through savings, increase labor effort, remittances or other behavioral responses. This assumption is implausible. Other studies take the post-transfer or observed consumption, as the welfare indicator. This alternative approximation assumes that any change in social protection transfers would be fully replaced from other income sources, and it is equally implausible. The correct counterfactual consumption will subtract the program benefits, but add the replacement income households would generate through their behavioral responses had they not benefited from the intervention. A series of estimates seems to suggest that the share of the replacement income would be around 50% of the value of the transfer (Ravallion (2000), van de Walle (2001; 2002)). In order to estimate the inequalityreducing (and, in the next section, the poverty-reduction) impact of social protection transfers, we net out 50% of the transfers from household consumption, and rank households based on this counterfactual consumption. While this estimate is not precise, the results presented in the paper are robust to the choice of a different share for the replacement income, especially for transfers that are small compared to household income. Thus, the assessment of the inequality- or poverty-reducing potential of the MIG and other social assistance programs would be robust to the choice of this parameter, while will tend to differ in the case of pensions. We tested for the sensitivity of the findings to the choice of this parameter, and we found very little divergence in the qualitative findings. Three related concepts are used to identify the inequality-reducing or redistributive impact of social protection transfers. A transfer is: Regressive: if the poorest groups receive a smaller share of program benefits than the share of the group in total consumption. If the share of the poorest quintile in total consumption is 4%, and the group captures only 3% of the program benefit, the program is regressive. Mildly Progressive, if the poorest groups receive a larger share of program benefits than the share of the group in total consumption, but less than the share of the group in total population. If the share of the poorest quintile in total consumption is 4% (and in total population is 20%), and the group captures 10% of the program benefit, the program is mildly progressive. Highly Progressive, if the poorest groups receive a larger share of program benefits than the share of the group in total population. If the share of the poorest quintile in total consumption is 4% (and in total population is 20%), and the group captures 21% or 60% of the program benefits, the program is highly progressive. Progressive transfers contribute to inequality reduction, and may contribute to poverty reduction as well. A transfer can be characterized as progressive or regressive given the distribution of program benefit across the income distribution (groups), i.e the incidence of its benefits. There are two ways in which one can assess the benefit incidence of a program: Statically: What share of program benefits accrue to a particular income group or quintile? and Marginally: Given a small increase (decrease) in the size of the program, which income groups will capture (loose) the extra benefits? 125

9 III. The Impact of Social Protection Programs on Poverty Reduction The previous section assessed the impact of the social protection programs in reducing inequality. This section takes a narrower look, at the poverty reduction impact of these programs. Using three related concepts of coverage, absolute targeting incidence and adequacy share of program benefits in the consumption of the beneficiary the last part of this section highlight that the good targeting performance of the MIG is accompanied by weak coverage and benefit adequacy. All other programs are weaker povertyreduction instruments. A. Coverage of Social Protection Programs Social protection programs reach directly (program participants) or indirectly (the household members of program participants) about 87% of the population, including 96% of the poorest quintile (Table 8). The MIG, program that represents a social safety net of last resort in fighting poverty, while very well targeted, covers only 11.7% of the poorest 20% of the population. Table 8. Program Participation Rate by Quintile, 2002 Program Poorest Richest National Social Protection, o.w.: 96% 93% 89% 85% 71% 87% Social Insurance pension - old-age 38% 36% 32% 25% 16% 29% pension - disability 12% 11% 10% 9% 5% 9% pension - survivorship 9% 9% 7% 5% 3% 6% pension - farmers 18% 17% 13% 8% 4% 12% pension - war veterans 0.4% 0.3% 0.1% 0.4% 0.1% 0.3% pension - social assistance 0.2% 0.2% 0.1% 0.1% 0.1% 0.2% unemployment benefit 8.9% 9.2% 7.5% 6.1% 3.1% 7.0% redundancy payments 0.2% 0.2% 0.2% 0.1% 0.0% 0.1% Social Assistance child allowance 54% 54% 56% 59% 55% 56% scholarship 0.6% 0.5% 0.5% 0.7% 0.6% 0.6% support for people with disabilities 4.6% 3.0% 1.8% 1.0% 0.7% 2.2% allowance for war veterans 3.6% 2.6% 2.1% 1.8% 0.8% 2.2% allowance for victims of political persecution 0.2% 0.2% 0.3% 0.3% 0.2% 0.2% social assistance provided by mayor's office (MIG in 2002) 11.7% 4.4% 1.9% 1.0% 0.3% 3.9% other social assistance benefits 1.1% 0.9% 0.7% 0.4% 0.4% Source: WB Staff Estimations based on ABF % B. Distribution of Social Protection Benefits across groups and Benefit Adequacy While the previous section examined the coverage of programs (in terms of people or beneficiaries), this section examines the distribution of benefits (the target incidence ) of social protection programs using data from the ABF Two concepts are used: (a) absolute target incidence, which measures average benefits received by any particular group as a share of total benefits (or the targeting outcomes of a program); and (b) relative incidence, which measures the average benefits received by any particular group as a share of average total consumption for that group (i.e., the relative importance of a program, or the adequacy of the benefit in protecting the recipient against poverty). Overall, in Romania most social protection transfers are highly progressive, with the exception of scholarships. Specifically, the bottom quintile receives close to one quarter (25%) of all public social protection spending, as compared with the top (richest) quintile, which receives 12% (Table 9). In other words, the richest receive a smaller absolute public transfers than the poorest. Furthermore, these transfers 126

10 are relatively more important to the poor than the non-poor (Table 10). Specifically, public transfers represent 80% of total consumption for the poorest quintile (net of 50% of social protection transfers), as compared with 6.8% for the top quintile. Table 9. Distribution of Social Protection Benefits, by Quintile 2002 Program // Consumption Quintile Poorest Richest National Social Protection, o.w.: 25% 24% 21% 18% 12% 100% Social Insurance pension - old-age 24% 24% 22% 18% 12% 100% pension - disability 21% 23% 22% 22% 12% 100% pension - survivorship 26% 28% 22% 15% 9% 100% pension - farmers 30% 30% 22% 13% 6% 100% pension - war veterans 32% 24% 13% 18% 13% 100% pension - social assistance 25% 37% 17% 13% 9% 100% unemployment benefit 25% 26% 21% 18% 11% 100% redundancy payments 29% 27% 26% 9% 9% 100% Social Assistance child allowance 20% 20% 20% 21% 19% 100% scholarship 16% 18% 17% 21% 28% 100% support for people with disabilities 42% 22% 20% 10% 6% 100% allowance for war veterans 37% 23% 21% 14% 6% 100% allowance for victims of political persecution 18% 14% 26% 31% 12% 100% social assistance provided by mayor's office (MIG in 2002) 62% 21% 12% 4% 2% 100% other social assistance benefits 33% 24% 19% 11% 13% 100% Memorandum Items Share in total population 20% 20% 20% 20% 20% 100% Share in total consumption 7% 12% 17% 23% 40% Source: WB Staff Estimations based on ABF % A number of social assistance programs, such as the MIG, support for the disabled, and allowances for war veterans are the best targeted programs to the poorest. Notably, the MIG succeeds in transferring 62% of program benefits to the poorest quintile, a performance that overshadows other similar programs in the region (Serbia, Kyrgyz Republic, or Estonia). 127

11 Table 10. Benefit Adequacy by Quintile 2002 Ratio of program benefits in the consumption of the group Program // Consumption Quintile Poorest Richest Social Protection, o.w.: 80.7% 45.3% 29.0% 17.7% 6.8% Social Insurance pension - old-age 48.5% 28.1% 18.4% 11.3% 4.3% pension - disability 6.3% 4.0% 2.8% 2.0% 0.7% pension - survivorship 4.2% 2.7% 1.5% 0.8% 0.3% pension - farmers 4.7% 2.8% 1.5% 0.6% 0.2% pension - war veterans 0.2% 0.1% 0.0% 0.0% 0.0% pension - social assistance 0.1% 0.1% 0.0% 0.0% 0.0% unemployment benefit 3.4% 2.1% 1.2% 0.7% 0.3% redundancy payments 0.2% 0.1% 0.1% 0.0% 0.0% Social Assistance child allowance 5.5% 3.2% 2.3% 1.7% 0.9% scholarship 0.2% 0.1% 0.1% 0.1% 0.0% support for people with disabilities 1.6% 0.5% 0.3% 0.1% 0.0% allowance for war veterans 1.1% 0.4% 0.3% 0.1% 0.0% allowance for victims of political persecution 0.2% 0.1% 0.1% 0.1% 0.0% social assistance provided by mayor's office (MIG in 2002) 4.1% 0.8% 0.3% 0.1% 0.0% other social assistance benefits 0.4% 0.2% 0.1% 0.0% Source: WB Staff Estimations based on ABF % C. Marginal Benefit Incidence Analysis The current distribution of program participants, however, is a poor indicator of how marginal benefits will be distributed if the size of the program will change, as discussed in Lanjouw and Ravallion (1999). If the socioeconomic composition of a program changes as programs expands or contracts, the distribution of the new program participants will be different than the current ones. One can distinguish the case of early capture, when the poor benefit from the program in the initial stages, but as program expands, the non poor will capture most of the gains from increased participation. In contrast, some programs may exhibit late capture by the poor. In the initial stages of program implementation, the non-poor get most benefits. However, as the program expands, the poor get most of the gains from increased participation. When reforming existing social protection programs, by rationalizing some and expanding others, it is marginal distribution of beneficiaries that coveys the right information to reallocate funds among competing programs and focalize better the resources toward the poor. Within the set of poverty-reduction programs implemented by the Government, such analysis is most relevant for the MIG. Compared to the earlier program (Social Aid), the coverage of the MIG increased substantially in its first year of application, thanks to more generous funding and adequate eligibility criteria the income threshold for the program was substantially increased. However, the program coverage of the extreme poor is still deficient: only three out of ten extreme poor benefit from the program. It is important to assess if, under the current implementation arrangements, an increase in the MIG will entail a reduction or an increase in the targeting performance of the program. An increase in the size of the MIG can be implemented without changing the targeting performance of the program. Table 11 presents estimates of average and marginal program participation rate, and the average and marginal distribution of program beneficiaries by quintiles. The average share of program beneficiaries illustrates what proportion of the MIG beneficiaries are from a given quintile. The marginal share of program beneficiaries presents, for a given increase of decrease in program size, what proportion of the new beneficiaries will come from a given quintile. The MIG is a case of early capture program. From 128

12 100 new program participants, 58 will be from the poorest quintile, and only 2 will come from the richest one. Compared to the current distribution of program participants, the distribution of new participants will be equally focalized to the poor. Even in the absence of changes in program design, an expansion of the MIG program will be strongly pro-poor 3. We suggest, in section III.F, that beneficial changes to program administration may be implemented to further improve its targeting performance. Table 11. Benefit Adequacy by Quintile 2002 Program Participation Rate Share of Program Beneficiaries Quintiles Average Marginal Average Marginal % 10.5% 60.4% 58.1% 2 4.5% 4.2% 23.0% 23.3% 3 1.9% 1.9% 9.8% 10.8% 4 1.0% 1.1% 5.1% 5.9% 5 0.3% 0.3% 1.7% 1.9% National 3.9% 3.6% 100.0% 100.0% Source: WB Staff Estimations based on ABF 2002 D. Cost Benefit Analysis Different programs contribute toward the poverty reduction goal proportionally with the amount of resources channeled through the program (transferred to the beneficiaries) and the degree of targeting. One way to focus on the targeting performance of various programs is to estimate the amount of resources spent in order to reduce the poverty gap of the program beneficiaries by 1 ROL. In this section, we use a simplified costbenefit analysis (CBA) framework to assess the contributions of various social protection programs in reducing poverty. For each program, we computed a cost/benefit ratio: (i) The benefits of the social protection (SP) programs were judged against their impact on reducing poverty, that is how much they contribute to the reduction in the poverty gap? and (ii) Program costs include the benefits transferred to program participants (without taking into account administrative costs, or other intake-cost for participants). The estimation proceeds in three simple steps: 1. First, we compute estimates of the current and counterfactual poverty gap 4. To estimate the counterfactual poverty gap, we use the results reported in section 3, that the level of consumption in the absence of a welfare program equals current consumption minus half of the welfare benefit. 2. Second, we estimate the contribution of the program to reducing the poverty gap, as the difference between the observed and counterfactual poverty gap Third, we estimate the cost-benefit ratio by dividing the reduction in the poverty gap due to the program by total volume of benefits derived from that program. 3 The validity of our inferences is limited to marginal changes in program participation, say, an increase by 1% to 10%. These estimators will be a poor guide for a major reform of the program. 4 The counterfactual poverty gap takes into account important incentive effects, such as the change in their behavior if public transfers were discontinued (supplying more work, tapping private transfers, etc.). 5 The implicit social welfare function used in the CBA penalizes all spending in excess of the household poverty gap. Such criterion fails to take into account the fact that (i) some social programs have multiple objectives; (ii) it is not feasible to "fill the poverty gap" by passive transfer policies (i.e., the administrative costs for such fine targeting may outweigh the marginal benefits from the transfers, and such programs may have adverse incentive effects: poverty traps for those close to the poverty line, who will be discouraged to seek work as they may loose program benefits). 129

13 Due to limitations in the ABF data, the program cost includes only the cost of the benefits provided through the program as reported or assessed by the beneficiary. This means that neither administrative costs nor other hidden costs (time and monetary costs associated with the means-test, supported by the beneficiary) effects are taken into account. The resulting ratio gives the numbers of ROL spent per 1 ROL reduction in the poverty gap, and the lower the number (preferably close to one), the better the outcome. To test how sensitive the results are to the choice of the poverty line, we estimate cost-benefit ratios for both total and extreme poverty lines. Table 12. Cost Benefit Ratio for Social Protection Programs, Extreme Poverty Total Poverty social insurance /contributory benefits pension - length-of-service pension - disability pension - survivor pension - farmer pension - war veterans pension - social assistance unemployment benefit redundancy payments social security benefits for maternity, child care leave social assistance /noncontributory benefits child allowance scholarship support for people with disabilities MIG in 2002, MISS in other social assistance benefits other noncontributory benefits allowance for war veterans allowance for victims of political persecution Total SP Source: WB Staff Estimations based on AIG and ABF Comparing the costs 6 and benefits of social protection programs, the most efficient program in reducing (total or extreme) poverty is the MIG. The MIG program costs 3.1 (2) ROL to reduce the total (extreme) poverty gap by 1 ROL. The next most efficient program, from the list of programs for which we can estimate these indicators with sufficient precision in ABF is support for people with disabilities and the child allowances. E. Overall Effectiveness of Social Protection Programs in Reducing Poverty The indicators of coverage, absolute target incidence, and relative target incidence (importance/adequacy), all reveal important information about the effectiveness of social protection programs. This section seeks to combine those multiple indicators for a more comprehensive review of these programs, in particular with respect to their effectiveness in reducing poverty. Figure 1 plots in a single graph the three related concepts of coverage, absolute target incidence, and adequacy for various social protection programs based on a simulated model that classifies the poor based on a counterfactual of consumption (in the absence the transfers, see section II). The x-axis presents the coverage of the poor. The share of total benefits received by the poor is plotted on the y-axis (absolute target incidence). Adequacy (relative incidence) is captured by the size of the bubbles in the graphs (and mentioned on the graph, next to the bubble ). A perfectly-targeted program would be located on the upper right-hand side of these graphs, with a large bubble (equal to the size of the poverty gap before the transfer). 6 Due to data limitations, the costs used for this analysis include only the value of the benefits provided by the programs as reported by the beneficiaries in the AIG or ABF. They do not include administrative costs or potential incentive effects. 130

14 The effectiveness of programs and policies varies significantly (Figure 1). Three types of programs are observed. First, the MIG has the best targeting performance, although low coverage and benefit adequacy. Second, the child allowances are less well targeted but have the largest coverage of the poor and extreme poor. Finally, the rest of the programs have high leakage, low coverage of the poor. Some programs within this group may be good candidates for rationalization, with the resulting savings to be added to the MIG. 131

15 Figure 1. Effectiveness of Social Protection Programs in Reducing Poverty Total Poverty Extreme Poverty targeting coverage pension - social assistance social security benefits for temporary disability, maternity and child care leave unemployment benefit/labour market-entry benefit/income support child allowance redundancy payments support for people with disabilities other social assistance benefits MIG 9 targeting coverage pension - social assistance social security benefits for temporary disability, maternity and child care leave unemployment benefit/labour market-entry benefit/income support child allowance redundancy payments support for people with disabilities other social assistance benefits MIG Source: WB Staff Estimations based on ABF 2002

16 Figure 2 illustrate the improvement in coverage that occurs with the implementation of the MIG in Compared to the previous Social Aid program, the MIG has better benefit adequacy and substantially higher coverage. However, the program does not serve a large fraction of the poorest 20% of the population (89% of them). Figure 2. Change in the Effectiveness of the Social Aid / MIG Program in Reducing Poverty Total Poverty Extreme Poverty Social Aid , MIG 2002 Social Aid , MIG targeting targeting coverage coverage Source: WB Staff Estimations based on ABF 2002 The Government has a good tool to reduce extreme poverty in Romania: the MIG program. Marginal benefit incidence analysis had shown that an expansion of the program can be implemented without a loss in its targeting performance. Increased funding, to the target level allocated for 2002 of 0.4% of GDP or even more, will cover an increasing proportion of the extreme poor. Some of the extreme poor, however, are likely to be missed by the program, as illustrated in Table A1 and A2 annexed. In 2002, the program was successful in covering larger households, those where the couple was living together or unmarried (pooling resources, while one of the adults claimed the benefits only for him and all dependents), the Rroma (compared to an average coverage of the poorest decile of 17%, the coverage of extremely poor Rromas was 36%), those headed by adults with no or few formal schooling, by farmers, unemployed and housewives. In general, MIG recipients are households familiar with the social assistance system; most of them (87% of total recipients) receive a number of other social assistance benefits, especially child allowances. The extreme poor households who failed to be covered by MIG benefits are especially urban households, households affected by industrial restructuring which do not take up the program due to stigma costs, and households with two or more able -bodied individuals. These households are equally poor, and need to be covered by the program. Changing the MIG program administration rules to bring them into the safety net, and help them to climb out of poverty, is the main challenge faced by the current administration. The next section suggests a number of changes in program implementation to facilitate such an outcome. F. Improving Social Assistance Administration The previous sections showed that, among social assistance programs, the MIG program has the best targeting performance, but a relatively low coverage of the poor. This section will take a closer look at the

17 institutional arrangements of MIG, trying to highlight two of the most important weaknesses of the program: (i) funding arrangements; and (ii) implementation. 1. Funding arrangements Unclear distribution of the financing burden between the central and local administration According to the law, the financing of the MIG program may be provided from two sources: (i) earmarked transfers from the central budget, distributed by the judet councils to localities; and (ii) local budget revenues and/or equilibration grants 7. Theoretically, the combination of central and local financing provides a good mix of incentives for the local administrations. A decentralization of the program financing, as it was tried after 1996 with the former Social Aid program, would have been detrimental, as the poorest localities, with the largest share of potential MIG beneficiaries, are also the ones with lowest capacity to raise revenues. In earlier years, various studies (Tesliuc et all (2001);World Bank (2002)) documented that decentralization of program financing resulted poor targeting, with larger programs in relatively richer areas and smaller programs in poorer areas, determined by the financing capacity of the local administrations alone. The optimal solution to this problem is pooling the poverty risk nationally. The central budget should cover (finance) the MIG benefits, i.e. guarantee this minimum income level, while local administrations should only be responsible for program implementation. However, there is a risk that local administrations will abuse the program (allowing substantial leakage of funds) if the capacity of the central administration to control how the program was implementation is weak, and the local budget incurs no costs (penalties) for such action. Introducing the obligation of co-financing by local authorities provides the right incentives for them to use the funds in a responsible manner. While these arrangements may work, the implementation of the program was hindered by the lack of clarity with respect to the share of co-financing of the local versus central government. The MIG law leaves unspecified the share of the state vs. local co-financing. Some local administrations reacted to these unclear legislative provisions by not providing any co-financing from local resources. Qualitative evidence suggest that local councils were and are afraid that MIG co-financing rules may change again, and they will end up paying the full cost of the program as was the case with the Social Aid program in 1996 (after the first year of implementation, when financing was provided by the central budget). Given the guaranteed feature of the MIG, the program may end up crowding out other local expenditures, considered more important by local constituencies. Unpredictable funding of the MIG Local authorities (Local Councils) cannot plan their budget without a clear provision specifying their level of co-financing, and face similar difficulties in planning their cash-flow. In 2002, this impossibility to predict the volume of resources provided by the central government for MIG pushed many Local Councils to limit the demand for the program. Some examples of limiting demand practices are: imputing by default revenues from informal work during the summer and consequently payin g only 40% -60% of the nominal benefit amount, and/or setting the number of community work days at 9 days/month independently of the benefit amount. MIG funding is characterized by a relatively complex institutional arrangement, which involves actors at central, judet and local level. An important role is played by the Judet Council, which is in charge to distribute the funds among localities. The lack of any regulation regarding the co-financing requirement is reproduced at this level, affecting thus especially poor communities, which cannot afford to co-finance MIG benefits. Lack of enforcement with regard to benefit payments 7 The equilibration grants are resources transferred to the poorest localities, on the basis of the principle of social solidarity. 134

18 The main difference between the MIG and the former means-tested Social Aid program is that MIG is guaranteed. However, at the moment there is no authority responsible or able to enforce this provision. The Ministry of Labor and Social Solidarity has the mandate to monitor the program, but does not have any power to apply sanctions to the local authorities which paid only partially the benefits or not at all, since the distribution of funds towards local budgets is entirely under the control of the autonomous Judet Councils, and the eligibility criteria can be relaxed or tightened by the local authorities (Local Councils). MIG resources from the central budget are not clearly earmarked The central budget earmarks in block the resources for the MIG and heating allowances, and from 2003 the salaries of the personal assistants for disabled persons. This arrangement does not allow to separate MIG benefits funding from heating benefits or salaries for the personal assistants of disabled, making thus impossible to control the local budget share allocated for MIG co-financing. This arrangement has a direct impact on spending priorities of local public administration, which has incentives to set the payment of salaries for personal assistants as the highest priority (because the personal assistants have a similar status with the employees of the city hall, and are included in their payroll). From the funds left, the local council has to prioritize the spending of MIG benefits and heating subsidies. Combined with the lack of enforcement regarding the obligation to cover entirely the amounts due to MIG beneficiaries, the actual arrangement leads to overdue or partial payments of MIG benefits. 2. Implementation Discretionary power of local public administration with regard to the criteria for assets evaluation and eligibility. A large share of the income of the poor is derived from informal activities, notably subsistence agriculture. In order to take into account this type of income, the MIG law require that program administrators will estimate an imputed value for the productive assets owned by the households, based on norms established by the local administration. The procedures to determine the imputed value of productive assets (such as livestock, land or agricultural equipment) vary from one Local Council to another. In theory, the local administration is best placed to use local information to better target the poor. Along these lines, the MIG law grants the power to determine the imputed value of productive assets to be taken into account in the means testing procedure to the Local Councils. This provision generated large discrepancies between the imputed values of similar assets over relative homogeneous areas 8. Evidence from earlier surveys confirms the existence of price differences across localities within judets, but of considerably smaller magnitude. To counteract extensive local discretion, some Judet Councils produced legislation that established the maximum and minimum values for these assets. The Government should consider the harmonization of this good practice nationwide. The procedures to determine the eligibility for the MIG are far from being uniform. There are significant differences between localities with respect to the kind of documents that should be provided by the potentia l beneficiaries. Many Local Councils ask for documents legalized by notaries, although the law does not require such type of evidence, thus raising the monetary and time cost for the applicants. The Government and the Ministry of Labor and Social Solidarity may consider following with detailed implementation norms to clarify and limit the amount of paperwork and documents requested from the applicants. The Government may consider strengthening the design, financing and implementation of the MIG program: 1. A more equitable co-financing scheme, based on an algorithm/formula that takes into account the financial capacity of localities together with the expected number of poor. As a rule, poorer localities should be subject to smaller co-financing requirements. To capture the poverty level at locality level, the 8 e.g. in the same judet the use value of a horse could vary between 150 thou ROL and 1 million ROL, for a swine between 42 thou and 420 thou, etc 135