* ~~31098 66 * ESP Discussion Paper Series Social Assistance and the Poor in Romania K. Subbarao Kalpana Mehra May 1995 . ~~~~~~Education and Social Policy Department Human Resources Development and Operations Policy The World Bank ESP Discussion Papers reflect work in progress. They are intended to make lessons emerging from the current work program available to operational staff quickly and easily, as well as to stimulate discussion and comment. They also serve as the building blocks for subsequent policy and best practice papers. The views expressed here are those of the authors and should not be attributed to the World Bank or its Board of Executive Directors or the countries they represent. ABSTRACT Following the transition to a market economy, the incidence and depth of poverty have been a source of concern in Romania. Social assistance programs have a critical role in mitigating the short run adverse impacts of transition on vulnerable groups. This paper analyses the distributive impact of public social assistance (transfer) programs, based on the data from the Integrated Household Survey for March 1994, and the available secondary information. The main questions addressed are: how have social assistance transfers fared in the recent period; how effective have the transfers been in reaching the poor; do the programs differ in their distributive impacts; what is the net impact of transfers on the poverty gap ratio, and how significant are the redistributive gins from a better targeting of programs. The paper concludes that though transfers are contributing to a reduction of the incidence and depth of poverty, many deserving poor (such as female-headed households) may not be receiving the needed transfers. Simulations suggest that significant gains can be realized by better targeting. The paper also notes that runal-urban disparities in transfer receipts are currently substantial. The paper also examines the pros and cons of the proposed minimum income guarantee, and argues that the scheme may be difficult to implement. Moreover, care needs to be taken to ensure that the proposed scheme does not trigger adverse incentives for work. SOCIAL ASSISTANCE AND THE POOR IN ROMANIA I. Introduction 1.1 Since the transition to a market economy in 1990, the incidence of poverty has risen dramatically in Romania (Rashid, 1994). According to the most recent (1994) Integrated Household Survey (IHS), for the country as a whole, the headcount index of poverty was 20 percent, and the poverty gap ratio was 5.7 percent.' Poverty ratio was higher in rural Romama (27 percent), than in urban Romania (14 percent). Poverty can be reduced in the long run only by sustained, labor-demanding growth and assured access of the poor to human resources. In the short run, given the increased incidence and depth of poverty in the recent period, it is important to protect the vulnerable households from a further worsening of their living conditions. In this respect, social assistance programs have a critical role to play. 1.2 This paper analyses the distributive impact of public social assistance (transfer) programs. Four transfers are considered: children's allowances, unemployment benefits, social assistance pensions, and other social assistance benefis (e.g., special aid for the handicapped, allowances for war widows, etc.). In addition, i-kind transfers are also analyzed. It uses the IHS survey data for March 1994, which contains socio-economic information on approximately 8200 individuals from 2650 households. The available secondary sources of data are also analyzed. 1.3 The main questions addressed are: (a) How have social assistance transfers fared (in real terms) in the recent period? (b) What are the household characteristics of the recipients of transfers? Are social assistance transfers effective in reaching the poor? Do the programs differ in their distributive impacts? What has been the net impact of transfers on the poverty gap ratio? Do the impacts differ by rural/urban location? (c) How significant are the redistributive gains by better targeting of programs? Overall, what specific policy changes are required to further close the poverty gap? 1.4 The paper is organized as follows. Section II describes the main social assistance transfer programs in the country, and provides an overview of transfers in the recent period. Section El[ examines the household characterics of the recipients of transfers. The distributional impacts of transfers in rural and urban areas are analyzed in Section IV. The impact of transfers on poverty, and the potential gains from better targeting of transfers is examined in Section V. Section VI offers some policy recommendations. E. Socal Assstance Transfers in Romania 2.1 During 1990-92, the system of social protection underwent many changes. At present, quantafively the most significant transfers are child allowances and unemployment benefits, which are briefly described below. 2.2 Czild Alowances. The government provides a monthly allowance to all children under the age of 16 (18 if handicapped). The amount of benefit per child increases by number of children and decreases by family I Due to incompatibility in methodology and the survey design, the headcount index and the poverty gap ratio for the previous years are not comparable with our numbers. 1 income. For example, as of November 1992, the first child in the family with monthly earning exceeding 3,350 lei gets 1,365 lei per month and the first child in the family with monthly earning of 2,500 lei gets 1,465 lei per month. The allowance for the fourth child in these respective families are 1,515 and 1,720 lei per month (Figure 1). The amount is doubled if the child is handicapped. Since May 1993, the amount of child allowance per child is the same, irespective of the number of children in a family and of the household's income. The allowance is indexed, but adjusts only partially to inflation. Child allowance is by far the single most significant social assistance benefit, accouting for 92 percent of all social transfers in 1993/94 exclusive of unemployment benefits. 2.3 Unemployment Alowances. The support for the unemployed is provided for the first nine months at the level of the gross minimum wage (16,730 lei in 1993). After the unmployment benefit runs out, a support allowance-a amount equivalent to 40 percent of the minimum wage-is available for the next 18 months. The benefit is provided monthly and is indexed. Beneficiares of unemployment allowance can also simultanesly receive the Children's Allowance and free medical care. However, during the 18 month period, the individual is required to attend a trainig or retnaining course and to look for employment. Failure to participate in training or engage in active job search can rsult in termination of the benefit. Flgure 1: Chfld Allowances by Family Income, Prior to May, 1993. 1 ,800 1 1, 20 1,720 1,700 s 1 1,600 1 90 1,509.; 65 1 ,E5 A' 1 X400 | JW 1,300.. - 2500 2501-3350 Family Income/Month S1 st Child - 2nd Child 3rd Child - 4th+ Child Note: The cut-off points and allowances have changed since 1992, but the numbers still hold good at 1992 prices. 2 2.4 Figure 2 presents the data on total social assistance (excluding unemployment benefits). The dominant tansfer, viz., child allowance, is also shown alongside. Both are shown in 1991 prices, and represent national (tomal) expendiures. As can be seen, there has been a dramatic contraction of both child allowance as well as other social assistance benefits. In 1993, the total social assistance transfer was only 47 percent of the level in 1991. The level of child allowance in 1993 was 52 percent of the level in 1991. Figure 2: Total Expeadituris on Transfers, 1991-1993 (in 1991 prices) 25300 ChManowa 2-o -.fl Cash Trnsizs (eCbhding w3A~Oyminubmefes) a 2.5 Table 1 presents the mnmber of benefciaries and spending levels (in 1991 price) for thereang social assistane programs. Funding levels for some programs (e.g., benefits to motbers with thre children or more) have contracted to such an extent as tD become negligible. However, the number of recipients as well as funding levels have risen for both social cantens and wives of conscripts. At thrse canteens, free meals are Provided to children in famiflieS whose monthy income Per person is less tha 30 percent of the minimum wage. This implies that this in-kind Wransfer is means-tested, though it is not possible to assess the tageting efficiency. T'hese are mean to prevent destituton. The number of recipients of free meals has increased by nearlY two-and-a-half times over 199 1-93, clearly reflecting the deteriorating living conditions in thle country. 2.6 Child allowance, and other forms of support to bmfaiie with chilre, accounted for 2.9 percent of GNP in 1989. By 1993, these benefits accounted for only 0.8 percent of GNP (UNICEF, 1993). 3 Table 1: Social Assistance in Romania: Programs and Spending 1991 1992 1993 A. Children's Allowances Recipients 4,050,000 3,893,000 3,705,658 Total Expenditure (millions) 26,190.2 20,922 13,604 (83.0) (89.7) (92.1) B. Benefits to mothers with three children or more Recipients 1,025,104 947,715 1,041,774 Total Expenditure 4979.0 2,035 494 (15.8) (8.7) (3.3) C. Birth Indemnity Recipients 118,152 91,314 54,324 Total Expenditure 179.9 54.8 8.6 (0.6) (0.2) (0.1) D. Benefits for wives of conscripts Recipients 3,302 2,309 3,949 Total Expenditure 22.8 17.4 260.0 (0.1 (0.1) (1.8) E. Social Canteen Recipients 9,509 14,229 22,012 Total Expenditure 190.9 291.5 396.8 (0.6) (1.2) (2.7) Total Social Assistance 31,562.8 23,320.7 14,763.4 Note: All expenditure are given in 1991 prices. Numbers in parnhesis are percentages to total social assistance. Source: Ministy of Labor and Social Protection. 2.7 In the absence of data on the number of the unemployed by duration of unemployment, it is difficult to estimate the total spending on unemployment benefits. The total number of the recipients of unemployment benefits has been rising, wheras the number of recipients of child allowances has been declinmg (see Tables 1 and 2). Assuming one-third of the unemployed in 1994 were long-term unemployed (i.e., unemployed beyond a period of 9 months), the total amount spent in 1994 would be about 16 billion lei, compared with 14 billion lei on child allowances. 2.8 Three important characteristics of the unemployed are worth noting. First, nearly 40 percent of the unemployed were under 25 years of age. Approxunately another 40 percent belonged to the age group 25-39. Second, 56 percent of the unemployed were women. Third, over 85 percent of the unemployed were persons with less than lyceum (upper secondary school) degree (see Table 2). 2.9 Given the overall contraction (in real terms) of social assistance transfers in general and child allowances in particular, it is important to ensure that the (reduced) resources are distnbuted disproportionately to the poorest groups. The extent to which the vulnerable groups are being served by social assistance can be gleaned from (a) the household characteristics of the recipients of social assistance transfers, and (b) the impact of transfer receipts on headcount and poverty gap ratios. These aspects are examined in the next two sections. 4 Table 2: Characteristics of the Unemployed Unemployed 1992 1993 1994 Total 601,479 1,065,844 1,182,583 Under 25 246,606 429,668 452,893 (41.0) (40.3) (38.3) 25-39 226,156 401, 147 457,747 (37.6) (37.6) (38.7) Females 365,641 623,171 657,156 (60.8) (58.5) (55.6) Degree 525,425 911,919 1,014,121 Less than Lyceum (87.4) (85.6) (85.8) Note: The mimbers in parenthesis are percentages to total number unemployed. Sowrce: Ministry of Labor and Social Protection mII. Household Characteristics of Transfer Recipients and Targeting Efficiency 3.1 Table 3 defines some of the characteristics of the household receiving transfers as compared to those not receiving the transfers. Households receiving children allowance, unemployment benefits or social assistance have lower mean per capita expenditures (both total and food). These households also have higher number of children than the households that do not receive the transfers. A higher proportion of male-headed households than the female-headed households receive children allowance and unemployment benefits in both rural and urban areas. A higher percentage of female-headed households in rural areas are claiming social assistance but headship is not a factor for social assistance recipients in the urban areas. The age of the head of household is significantly lower for the recipients of unemployment benefits implying that the problem of unemployment is more serious among young adults. Educational attainment of the head of household is higher for those claiming children's allowance in both rural and urban areas but the difference in education of household head is not as significant in the urban areas as it is in the rural areas. 3.2 To assess the extent to which public transfers are meeting a country's distributional objectives, it is useful to address the question: who is receiving public transfers? To answer this question, a probit analysis is attempted. In general, one should expect the probability of receiving a public transfer to be inversely related to current income, net of transfers. In the probit equation, gross income per capita net of transfers is used as one of the explanatory variables. 2 Tobit analysis would allow us to capture the differences in the level of transfers. Since the level of child allowance per child, the dominant transfer in Romania, does not vary across households, probit is as good a specification as Tobit. 5 Table 3: Household Characteristics for Transfer Recipients and Others, Rural/Urban Other Social Assistance Children's Allowance Unemployment Social Assistance Benefits In-kid Trasfers Mean Levels Not Not Not Not Not receiving Receiving receiving Receiving receiving Receiving receiving Receiving receiving Receiving Rural Per capita gross income net of transfers 262594 279076 269830 250642 267935 258042 266797 281918 266150 296106 Per capita expenditure 55818 45158 54006 43234 53294 40464 53107 43154 52190 63275 Percapita food expend. 33163 25542 31739 25075 31259 24136 30984 29388 30716 35266 Children 0-16 yrs. 0.20 1.94 0.65 1.19 0.69 1.20 0.73 0.33 0.70 0.98 Elderly 1.04 0.34 0.89 0.47 0.84 0.77 0.79 1.75 0.85 0.42 Female beaded (%) 25.7 10.0 22.2 12.8 20.9 24.6 21.8 5.4 21.3 16.4 Age of head 60.5 45.4 57.0 50.0 56.1 56.1 55.3 72.5 56.4 49.4 Education of head 3.8 4.9 4.1 4.3 4.1 3.9 4.1 3.3 4.1 5.0 Eamning members 0.95 1.71 1.16 1.26 1.17 1.28 1.20 0.58 1.17 1.38 Unemployed 0.13 0.27 0.04 1.12 0.17 0.15 0.17 0.11 0.17 0.22 Pensioners 0.91 0.33 0.78 0.44 0.73 0.77 0.70 1.42 0.75 0.45 Number of households 889 372 1105 156 1200 61 1206 55 1206 55 Urban Per capita gnss income net of transfers 193039 233145 215992 147002 212240 132045 . 210833 143678 212147 186070 Per capita expenditure 78125 58673 71077 50639 69572 56316 69218 63240 68401 74453 Per capita food expend. 40180 31214 37272 29982 36756 31492 36493 39246 36301 38606 Children 0-16 yrs. 0.26 1.56 0.76 1.06 0.77 1.13 0.80 0.12 0.79 0.75 Elderly 0.68 0.17 0.51 0.13 0.47 0.63 0.44 1.76 0.46 0.62 Female beaded(%) 28.7 10.5 22.3 12.5 21.4 21.1 21.7 9.1 21.0 24.5 Age of head 54.3 42.2 50.1 42.6 49.4 51.0 48.9 72.0 49.0 52.5 Education of head 5.6 6.1 5.9 5.4 5.8 5.6 5.8 5.0 5.8 5.8 Eamning member 0.97 1.60 1.24 1.09 1.23 1.10 1.25 0.27 1.25 1.07 Unemployed 0.15 0.22 0.08 1.10 0.18 0.23 0.18 0.03 0.19 0.10 Pensioners 0.69 0.22 0.53 0.20 0.49 0.79 0.48 1.45 0.48 0.67 Number of househokds 828 561 1253 136 1337 52 1356 33 1234 155 Source: Romania IHS, 1994. 3.3 Education increases pemanent income, which should lower the household's need for a public transfer. At the same time, education imcreases the awareness of public tansfers and, to that extent, mcreases the probability of a household capturing transfers. To gauge the effect of education, we entred educational attaunnents of heads of households. Vulnerability is typically higher both at the beginn g of the life cycle, and at the close of the life cycle. So we included in the probit equation both the number of children under age 16, and the number of adults (excluding red persons) and the elderly (males 60+ and females 55+) in each household. In all economies in transition, female-headed households are a particularly vulnerable group; so we entered in the probit equation a dummy for the sex of the head of the household. Finally, to capture regional differences, we ented five dummies for different regions in the country. We estimated probit equations for the whole sample, and for nrral and urban samples separately. The results are shown in Tables 4, 5 and 6. 3.4 For the whole sample, and for the nual and urban samples, per capita income net of transfers has the expected inverse sign, implying that the probability of a household receiving public transfers declines as the per capita income increases. The marginal effect of per capita income (at mean value of income) is very low (at 0.000002) implying that while transfers decline with income, their margnal effect is negligible. 3.5 The higher the educational atainment of a household head, the higher the probability of the household receiving a public transfer; the marginal effect of education is much stronger in nrual than in urban areas. 6 Also, the effect is significant for total cash transfers and for children' s allowances, but not for social assistance (includes social assistance pensions, special aid for the handicapped and aid from social assistance fund). 3.6 An interesting finding relates to the sex of the head of the household. In general, the probability of receiving a public transfer is lower if the household is headed by a female than if it is headed by a male. This is true not only for cash transfers as whole, but also for children's allowance in urban areas. The marginal effect of female-headship is higher in urban than in rural areas. The coefficients of both education and sex of the household head are highly significant-whether analyzed for the total sample (Table 4), or separately for the nrual and urban samples (Tables 5 and 6). 3.7 In general, households have to spend (invest) their time for receiving social assistance transfers (trips to offices, standing in queues, etc.). Availability of adults of working age with a family is, therefore, an advantage to receive transfers. Results confirm that the higher the number of adults of working age, the higher the probability of receiving transfers.3 However, the higher the number of elderly persons in a household, the lower the probability of receiving a public transfer. Elderly persons contribute to family incomes (via pension receipts); so their presence in a household reduces the probability of receiving a public cash transfer, especially if some of these are means-tested. The exception is social assistance, where the number of elderly in a household increases the probability of receiving it. As for regional differences, only the households in the (poorest) Northeast region have a higher probability of receiving transfers than others. 3 The appropriate explanatory variable is adults of working age minus actual workers. We could not tease out this information from the data set; so we opted for the second-best specification of all adults of working age in a household. 7 Table 4: Probit Estimates of Types of T ransfers, Total Total Cash Transfers Children's Allowances Social Assistance Asymptotic Marginal Asymptotic Marginal Asymptotic Marginal Co-efficetnt t-value effect at Co-efficeint t-value effect at Co-efficeint t-value effect at mean of x mean of x -mean of x Constant -0.615 -5.19 -0.165 -0.936 -7.52 -0.163 -2.007 -10.08 -0.044 Gross Income without wransfers -0.0000005 -3.30 -0.0000002 -0.0000001 -1.01 -0.0000001 -0.0000007 -2.69 -0.0000003 Head's schooling 0.025 1.95 0.013 0.045 3.35 0.026 -0.006 -0.28 -0.003 Female Head -0.413 -5.74 -0.192 -0.382 -4.88 -0.178 0.150 1.31 0.077 Urban -0.024 -0.40 -0.011 0.098 1.57 0.051 -0.125 -1.23 -0.059 South-West 0.062 0.80 0.031 0.071 0.86 0.036 -0.028 -0.21 -0.014 North-West -0.070 -0.89 -0.034 0.062 0.75 0.031 -0.105 -0.75 -0.051 Nonth-East 0.131 1.68 0.067 0.156 1.89 0.080 0.104 0.82 0.053 Bucharest -0.041 -0.16 -0.020 -0.008 -0.03 -0.004 -0.016 -0.03 -0.008 Unclassified -0.112 -1.05 -0.055 -0.064 -0.57 -0.032 0.074 0.40 0.037 Working adults 0.357 12.81 0.260 0.284 10.25 0.194 0.197 4.94 0.124 Elderiv -0.131 -3.01 -0.061 -0.428 -9.04 -0.167 0.212 3.02 0.117 N 2646 2646 2646 Log-likelihood -1554.9 -1396.9 -449.6 Note: The dependent variable is transfer receipt: 1 if transfer is received, 0 otherwise. Source: Romania IHS, 1994 Table 5: Probit Estimates of Types of Transfers, Rural Total Cash Transfers Children's Allowances Social Assistance Asmpttic Marginal Asymptotic Maginal Asymptotic Marginal Co-efficeint t-value effect at Co-efficeint t-value effect at Co-efficeint t-value effect at mean of x mean of x mean of x Constant -0.953 -5.16 -0.162 -1.327 -6.74 -0.122 -1.817 -6.16 -0.062 Gross Income without transfers -0.0000002 -1.14 -0.0000001 -0.0000002 -0.77 -0.0000001 -0.0000002 -0.53 -0.0000001 Head's schoolng 0.077 3.15 0.048 0.127 5.01 0.089 -0.039 -0.92 -0.017 Female Head -0.241 -2.23 -0.116 -0.196 -1.60 -0.095 0.233 1.43 0.121 South-West -0.116 -1.06 -0.056 -0.144 -1.19 -0.069 -0.153 -0.90 -0.073 North-West -0.132 -1.14 -0.064 0.002 0.02 0.001 -0.147 -0.82 -0.071 North-East -0.068 -0.63 -0.033 -0.011 -0.10 -0.005 -0.212 -1.22 -0.101 Bucharest 0.00 0.00 0.00 0.00 0.00 0.00 Unclassified -5.075 -0.00 -2.521 -4.555 -0.00 -2;264 -4.580 -0.00 -2.276 Working adults 0.403 10.53 0.298 0.316 8.19 0.219 0.193 3.74 0.120 Elderly -0.056 -0.92 -0.027 -0.338 -5.09 40.131 0.108 1.14 0.058 N 1257 1257 1257 Log-likelihood -705.6 -596.9 -235.1 Source: Romania IHS, 1994 Table 6: Probit Estimates of Types of Transfers, Urban |__________ Total Cash Transfers Children's Allowances Social Assistance Asymptotic Marginal Asymptotic Marginl Asymptotic Marginal Co-efficeint t-value effect at Co-efficeint t-value effect at Co-efficeint t-value effect at mean of x mean of x ean of x Costant -0.424 -2.65 40.142 -0.655 -3.96 -0.167 -2.269 -7.77 -0.026 Gross income without transfers -0.0000009 -3.60 -0.0000004 -0.0000006 -0.36 -0.000000 -0.0000025 -3.92 -0.0000008 Head's schooling 0.005 0.36 0.002 0.010 0.62 0.005 0.020 0.69 0.011 Female Head -0.565 -5.74 -0.26 -0.482 -4.67 -0.221 40.053 -0.31 -0.026 South-West 0.237 2.13 0.123 0.271 2.35 0.141 0.102 0.45 0.052 North-West -0.023 -0.21 -0.011 0.103 0.92 0.052 -0.026 -0.12 -0.013 North-East 0.363 3.18 0.191 0.342 2.93 0.179 0.472 2.31 0.252 Bucharest 0.061 0.23 0.030 0.143 0.51 0.071 0.034 0.07 0.017 Unclassified 0.0004 0.00 0.00 0.043 0.35 0.021 0.243 1.06 0.125 Working adults 0.317 7.61 0.227 0.255 6.25 0.173 0.239 3.64 0.159 Elderlv -0.177 -2.79 -0.082 -0.480 -6.92 -0.197 0.356 3.28 0.202 N 1389 1389 1389 Log-likelhood -835.9 -787.2 -201.1 Source: Romania IHS, 1994. 8 3.8 One important attnbute of vulnerability in Romania is ethnicity. It is widely believed that the gypsies are among the poorest and least literate in the country. In the data set, the only source of information for identifying ethnicity is "language spoken". On this criterion, only 0.5 percent of households in the sample can be deemed to belong to the gypsy community. It is probable that more households belonged to the gypsy community, but did not (for whatever reasons) identify themselves by their language. It is also probable that the sample itself underrepresents the gypsy community. Owing to small sample size, it has not been possible to explore the relationship of transfer receipts to ethnicity. 3.9 The main finding is that public transfers are only loosely targeted, are biased against heads of households headed by women, and by persons with generally lower educational attnainments. The marginal effect of the bias against persons with lower educational attainments is higher in rural than in urban areas, while the opposite is mte with respect to the marginal effect of the bias against female-headed households. 3.10 Ineffective targeting implies that many eligible households are not receiving transfers (exclusion errors), and many ineligible households are receiving transfers (inclusion errors). The demographic characteriscs of recipients are shown in Table 7. The average number of children per household, and the proportion of households with children, are higher in urban than in mral areas, particularly in the low income deciles. Rural-urban differences in demographic characteristics get narrowed for the highest income decils. 3.11 Table 8 presents the percent of eligible households not receiving child allowance, and the proportion of ineligible households receiving child allowances, separately for rural and urban areas. About 20 percent of eligible households are not receiving children's allowance. The proportion of "exclusion errors" is high for the poorest deciles in both rural and urban areas. The high proportion of exchlded eligible urban households in the high income deciles may be volntary; the allowance is so small that most rich households may not take it up. 3.12 Inclusion errr-the proportion of ineligible households receiving child allowances-is low in rural areas, but high in urban Romania, especially among middle income deciles. 3.13 The characteristics of eligible households who did not receive a child allowance are shown in Table 9. The most obvious and worrying finding is that a high proportion of eligible households not receiving child allowance were below the food poverty threshold. 9 Table 7: Family Size, Number of Children and Households Receiving Children's Allowances by Per Capita Consumption Decile and Rural/Urban Rural Urban Per Capita Households Households Consumption Number of Eligible to Households Number of Eligible to Households Decile Eligible Receive Receiving Eligible Receive Receiving Family Size Children Child Child Family Size Children Child Child Allowance* Allowance Allowance* Allowance 1 4.1 1.3 51.0 33.8 4.5 1.8 67.4 57.1 2 3.7 1.2 54.6 42.8 4.0 1.4 70.3 50.0 3 3.6 1.0 44.2 39.1 3.9 1.2 62.7 50.6 4 3.1 0.7 38.9 30.5 3.6 1.1 66.4 55.2 5 2.9 0.6 32.6 30.2 3.3 0.9 54.1 49.6 6 2.8 0.5 32.8 26.6 3.3 0.8 55.7 50.7 7 2.8 0.5 26.8 18.8 3.1 0.7 49.7 46.0 8 2.6 0.4 24.3 23.3 2.9 0.7 46.9 35.9 9 2.6 0.4 28.8 24.3 2.6 0.5 31.9 27.9 10 2.3 0.2 15.2 17.0 2.3 0.4 26.8 23.0 Total 3.1 0.7 36.2 29.5 3.1 0.8 47.4 40.4 N 1261 1389 * As percent of total number of households. Note: Eligible children are defined as those 0-16 years of age. Source: Romania IHS, 1994. 10 Table 8: Child Allowances: Exclusion and Inclusion Errors in Targeting Rural Urban Households not Receiving Households Receiving Households not Receiving Households Receiving Child Per Capita Child Allowance as Child Allowance as Child Allowance as Allowance as Percent of Consumption on Percent of Eligible Percent of Ineligible Percent of Eligible Ineligible Households Decile Households Households Households 1 37.8 4.2 18.2 1.7 2 22.9 1.5 28.9 0.0 3 11.5 0.0 23.1 6.5 4 25.6 2.5 23.4 12.8 5 9.5 1.2 19.2 12.9 6 21.4 1.2 19.2 12.9 7 30.0 0.0 21.0 13.4 8 12.0 2.6 31.1 6.9 9 21.9 2.5 21.5 4.3 10 11.8 4.2 27.7 5.1 Total 22.1 2.0 23.5 7.8 Note: Eligible households are defined as those with one or more children 0-16 years of age. Source: Romania, IHS, 1994. Table 9: Characteristics of Eligible Households Not Receiving Children's Allowances Eligible Households Not Percent All Households Receiving Child Allowances Poor 15.4 25.8 Ultrapoor 8.6 16.4 Female Headed 21.2 9.0 Rural 47.6 39.5 Head has less than primary education 26.3 16.4 Households with 3 or more children 7.0 11.7 Households with unemployed 14.8 19.5 members N 2650 256 Source: Romania IHS, 1994 IV. Distnrbutional incidence of Transfers. 4.1 Total cash and in-kind transfers as a percent of gross income, net income and total expenditures of each decile are shown in Table 10. In general, transfers are a more important source of income for the poorest deciles than for the richer deciles; this is particularly so when transfers are considered as a percent of expenditures for each decile group. For example, the poorest decile group receives 18 percent of its expendiue from transfers, compared with 2.7 percent for the richest decile group. The transfers as percent of expenditures also varied by headship. Female headed households received ondy 5.6 percent of expenditures as transfers versus 7.8 percent for the male headed households. I11 Table 10: Total Transfers* as a Share of Gross Income Net Income Expenditure Decile 1 10.1 12.1 18.1 2 18.2 20.4 12.7 3 10.1 11.0 9.9 4 9.4 10.8 9.3 5 6.3 7.2 7.1 6 6.2 7.0 6.9 7 5.7 6.4 4.9 8 5.0 5.4 5.1 9 5.2 5.9 3.6 10 3.8 4.7 2.7 Residence Urban 7.8 9.4 6.9 Rural 7.1 7.4 7.8 Poverty Ultrapoor' 11.1 12.9 17.5 Nearpoor 18.3 21.0 13.4 Nonpoor3 6.2 7.0 5.8 Household Head Male 7.9 8.9 7.8 Female 5.9 6.9 5.6 Sample 7.4 8.5 8.3 Note: Transfers include: Unemployment benefits, allowances for children, student financial aid, social assistance pension, special aid for handicapped, allowances for veterans and widows, and persons persecuted for political reasons and in-kind transfers. ' Households below the food poverty line (Rural = 22475 lei; Urban = 24287 lei). 2Households below the poverty line (includes a minimli allowance for non-foods) and above the food povery line (Rural= 27377 lei; Urban = 31126 lei). 3Households above the poverty line. Source: Romania IHS, 1994 4.2 From a distributional perspective, it is important to know whether or not the poorest deciles are receiving a share of ransfers higher than their shares in income/expenditre. The decile-wise shares in inme and expendituer are shown in Table 11 for the whole sample, and in Tables 12 and 13 for Rural and Urban samples, respectively. 4.3 When the whole sample is nsidered, it is inteestng that cash transfers are about equally distribud across deciles-i.e., each decile is receiving a share approximating to 10 percent.4 However, the poorest deciles received cash taners highr than proportionate to their share in expenditure, whereas the richest deciles received cash tansfers less than proportionate to their share in expenditure. By contrast, in-kind transfers are acc disp onely to the richer expenditure deciles. In-kind transfers include largely subsidized electricity and housing. These transfers together constitute less than 5 percent of total transfers. Though the in-kind transfers are currenly highly unequally distributed, a more equal distribution is unlikely to augment the poor's income substantially because of their relatively insignificant share in total transfers (particularly in electricity). Nonetheless it would be desirable to exclude the upper income deciles from in-kind 4 This is a studard featr of unsmion economies. See B. Milanovic (1994). 12 Table 11: Distribution of Income, Expenditures and Transfers by Per Capita Consumption Decile, Total Transfers Decile Number of Percent Gross Net Income Total Households Income Expenditure Cash In-kind Total 1 185 7.1 6.3 6.7 2.8 8.7 2.1 7.7 2 208 8.0 7.4 7.7 4.6 11.5 2.1 10.7 3 213 8.2 7.0 7.3 5.8 11.0 1.6 10.1 4 243 9.3 8.6 8.8 6.8 11.8 5.2 11.3 5 260 10.0 9.9 10.1 7.9 9.9 5.5 9.6 6 265 10.2 10.2 10.2 9.2 11.1 9.5 11.1 7 272 10.5 10.6 10.4 10.6 8.9 10.5 9.1 8 294 11.3 11.5 11.4 12.3 9.6 15.7 10.2 9 310 11.9 12.1 11.6 15.1 8.0 18.4 8.9 10 351 13.5 16.5 15.9 24.8 9.6 29.3 11.3 2601 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Source: Romania, IHS, 1994. Table 12: Distribution of Income, Expenditure and Transfers by Per Capita Consumption Decile, Rural Transfers Decile Number of Gross Total Households Percent Income Net Income Expenditure Cash In-kind Total 1 137 11.2 9.9 10.2 4.7 12.4 5.1 12.1 2 146 11.9 11.4 11.5 7.7 18.0 8.6 17.6 3 130 10.6 9.2 9.3 8.5 12.8 1.3 12.3 4 128 10.5 9.9 9.9 8.2 10.6 5.1 10.3 5 126 10.3 11.2 11.3 8.7 9.6 13.5 9.8 6 125 10.2 10.1 10.1 9.7 8.9 19s3 9.3 7 111 9.1 9.6 9.6 9.9 7.4 7.5 7.4 8 102 8.3 9.3 9.3 9.9 9.8 8.2 9.8 9 108 8.8 9.0 8.7 13.1 6.5 18.9 7.0 10 110 9.0 10.3 10.1 19.7 4.1 12.5 4.4 1223 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Source: Romania, IHS, 1994. 13 Table 13: Distribution of Income, Expenditure and Transfers by Per Capita Consumption Decile, Urban Transfers Decile Number Percent Gross Net Income Total Cash In-kind Total of Income Expenditure House- holds 1 48 3.5 2.1 2.2 1.5 4.7 1.4 4.3 2 62 4.5 2.8 2.8 2.4 6.1 0.5 5.4 3 83 6.0 4.4 4.6 3.8 9.5 1.7 8.5 4 115 8.3 7.1 7.3 5.9 13.0 5.2 12.0 5 134 9.7 8.5 8.5 7.3 10.3 3.5 9.4 6 140 10.2 10.2 10.4 8.9 13.3 7.0 12.4 7 161 11.7 11.6 11.6 11.0 10.3 11.3 10.4 8 192 13.9 14.1 14.0 14.1 9.5 17.6 10.6 9 202 14.7 15.6 15.2 16.5 9.2 18.2 10.4 10 241 17.5 23.6 23.4 28.4 14.1 33.6 16.7 1378 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Source: Romania, IHS, 1994. subsidies not so much for redistributive effects as for promotig fiscal balance. For example, the higest expenditure decile received 29.3 percent of all in-kind transfers in the sample. In-kind transfers are not contributing to a reduction of incquality in expenditure disribution. 4.4 Interestng nual/urban contrasts are worth noting. Cash transfers have accrued disproportionately to the poorest deciles in rural areas, but that was not the case in urban areas, where the middle- income/expenditure deciles (and the nchest decile) received a disproportionate share of cash transfers. As a consequence, cash transfers in urban areas are not playing the same "equalizing role" as in rural areas. Moreover, as much as 81 percent of all in-kind transfers were captured by the top (nchest) four icome/expendih decile groups im urban areas. Clearly, in-kind tansfers may have in fact contributed to an accentuation of inequality in expenditure distribution in urban areas. 4.5 The distnrbutional impacts of all transfers, cash transfers, and in-kind transfers can also be gleaned from Lorenz distributions shown Figures 3 (total sample), 4 (mrual sample) and 5 (urban sample). Tbe expenditure distributions are also shown alongside. In sum, cash transfers as a whole are progessive in nual areas, whereas in-kind transfers are inequitable in both rural and urban areas-urban to a much larger exent than in rurl. For the couty as a whole, cash transfers are only mildly progressive, whereas in-kind transfers are regressive. 14 Rgw a DWk' of Tinof C.IUI*f ud T,UUU T--~~~- 0 20 40 o 3 oo CaiiI uIw et popuaw Rps4. OabaU o Telal CmiapUf d~ Taumbie 40 C 40 60 30 00 Ciansid" we m of popuMdun tm. 4 DObrbu of Tel Cn_u_Um und Tausfs - U,Au 10 0- - I~~~~~~~~~~~~~~~~~~~~~~Tu E*wW I ln 20-- lbi ,, / , 1X1.0 - 2 - - ___Tea gi 20 O D0 40 s0 so 100 C o -~~~~~~~~~~~~5 Ssabue: of TodCnmdSw mk 4.6 Table 14 presents the relative shares of each type of transfer for the ultra-poor, near-poor, and nonpoor households. These are shown separately for rural and urban samples. For all households, children's allowances are the most significant of all transfers. In urban areas, the share of children's allowances in total transfers is the same for the nonpoor as for the ultra-poor; this is not surprising since the new law mandates that every household receives the same level of child allowances regardless of the level of expenditure/income. Unemployment benefits constitute a slightly higher share of total transfers for the ultra-poor and near-poor. Table 14: Relative Shares of Transfers by Region and Poverty Ultrapoor' Nearpoor Nonpoor3 Urban Rural Urban Rural Urban Rural Unemployment 16.1 23.6 22.0 24.0 14.6 23.1 Children's Allowances 70.6 54.2 63.9 54.9 60.9 53.4 Social Assistance 7.8 11.4 7.7 9.6 4.0 6.4 Other Social Assistance Benefits 2.2 7.0 2.2 8.7 4.1 9.1 In-kind 46 86 46 65 648 412 N 3.3 3.8 4.2 2.8 16.3 8.0 lHouscholds below the food poverty line (Rural = 22475 lei; Urban = 24287 lei). 2Households below the poverty line (includes a minimal allowance for non-foods) and above the food poverty line (Rural = 27377 lei; Urban = 31126 lei). 3Households above the poverty line. Source: Romania IHS, 1994. 4.7 To know the distriuional impacts of each transfer, dis g data are presented in Tables 15, 16, and 17. For convenience, the data are also presented in Figures 6 (rural) and 7 (urban). In nrral areas, the sbares of all except "other transfers" are much higher than the shares in total expenditures for the lowest five deciles, contributing to distributional equity. In urban areas, though the bottom two deciles are receiving shares in children's allowances and unemployment allowance equal to or higher than their shares in expenditures. The middle expende groups are capturing relatively higher shares of all transfers (including child allowances). Moreover, the richest decile captured the highest share of "social assistance" tranfer-an unquestionably inequitable outcome.5 'The shares of the top five deciles in "other transfers" are also much higher than would be warranted on grounds of equity. In many transition counties, it is not uncommon that social assistance transfers for the disabled comprising hearing aids, wheel chairs, etc. are cornered by the rich among the disabled. 16 Table 15: Distribution by Expenditures and Cash Transfers by Per Capita Consumption Decile, Total Transfers Total Children's Unemploy- Social Decile N Expenditure Cash Allowance ment Assistance Other 1 185 2.8 8.7 9.8 7.6 8.8 6.6 2 208 4.6 11.S 11.8 13.1 8.0 8.3 3 213 5.8 11.0 11.6 12.4 8.6 3.1 4 243 6.8 11.8 10.0 15.9 7.9 6.6 5 260 7.9 9.9 11.4 8.2 9.3 10.9 6 265 9.2 11.1 9.9 14.7 5.2 11.2 7 272 10.6 8.9 10.7 8.0 5.4 9.1 8 294 12.3 9.6 8.9 7.1 8.0 35.0 9 318 15.1 8.0 9.3 7.9 5.5 6.0 10 351 24.8 9.6 6.6 5.1 33.3 3.1 Total 2601 100.0 100.0 100.0 100.0 100.0 100.0 Source: Romania IHS, 1994. Table 16: Distribution by Expenditures and Cash Transfers by Per Capita Consumption Decile, Rural Transfers Total Children's Unemploy- Social Decile N Expenditn Cash Allowance ment Assistance Other 1 137 4.7 12.4 12.9 12.5 21.1 8.5 2 146 7.7 18.0 18.6 18.7 14.4 12.1 3 130 8.5 12.8 17.3 12.6 5.0 2.5 4 128 8.2 10.6 10.3 11.6 9.3 6.1 5 126 8.7 9.6 10.3 6.7 15.5 12.4 6 125 9.7 8.9 7.9 9.3 6.5 11.7 7 111 9.9 7.4 4.9 8.7 9.0 8.2 8 102 9.9 9.8 5.7 7.7 10.5 37.0 9 108 13.1 6.5 7.9 6.8 3.0 1.4 10 110 19.7 4.1 4.1 5.4 5.7 0.0 Total 1223 100.0 100.0 100.0 100.0 100.0 100.0 Source: Romania IHS, 1994. 17 Table 17: Distribution by Expenditures and Cash Transfers by Per Capita Consumption Decile, Urban Transfers ToW Children's Unemploy- Social Da:ile N Expenditgre Cash Allowance ment Assistance Other 1 48 1.5 4.7 7.5 2.2 2.7 2.5 2 62 2.4 6.1 6.9 6.7 4.9 0.0 3 83 3.8 9.5 7.5 12.1 10.4 4.4 4 115 5.9 13.0 9.8 20.8 7.3 7.6 5 134 7.3 10.3 12.2 9.9 6.2 7.7 6 140 . 8.9 13.3 11.3 20.7 4.5 10.0 7 161 11.0 10.3 14.9 7.2 3.6 11.2 8 192 14.1 9.5 11.2 6.4 6.8 30.7 9 202 16.5 9.2 10.3 9.2 6.8 16.1 10 241 28.4 14.1 8.4 4.9 46.8 9.7 Total 1378 100.0 100.0 100.0 100.0 100.0 100.0 Source: Romania, IHS, 1994. ng.Obabudm d C _ d_ To pe TUUi., bud 1W - ~ ~ ~ ~ ~ ~ ~ ~ TWComd *OD ~ ~ ~ ~ ~ ~ ~ ~ O ,~~ I' I -H o 20 40 io so .0o mum. 7. OhWbudun of CmnmAopdu md lWyM of 7Tinudum Demo 100; ii - TOW C4-Anasm'd- 40 * 2O U a 10 so so8 leo cmundg*Sw uawe of pegumladn Sowum bmu'ia HS5 18 4.8 Private remittances are negligible among poor households in both rural and urban Romania (see Table 19). Private remittances are contributing significandy to household income for the richest income decile in urban Romania. 4.9 In Table 18, the decile-wise shares of unemployment allowance are contrasted with the contribution of each decile to total unemployment in the country. Alongside, expenditure shares of each decile are also given. The poorest four deciles are represented disproportionately among the unemployed, particularly in mral areas. These four deciles are receiving a substantial share of the unemployment allowance as well. In urban areas, the shares in unemployment allowance of the poorest two deciles are much lower than their shares in national unemployment. The poorest deciles are benefiting from unemployment allowances, more so in rural than in urban areas. 4.10 Six conclusions emerge from the above analysis: (a) While the expenditure distribution is more unequal (Gini ratio is higher) in urban than in rural areas, the distribution of cash transfers as a whole is more equitable in rural than in urban areas. (b) The distnrbution of in-kind transfers is regressive in both rural and urban areas. (c) The households headed by relatively more educated persons have a higher probability of receiving transfers, while households headed by women have a lower probability of receiving transfers than those headed by men. While the marginal effect of education is twice as high in rural than in urban areas, the margial effect of female-headness is slightly higher in urban than in rural areas. Child allowances are poorly targeted. A high proportion of eligible poor households are not receiving child allowances. (d) Moreover, the coverage of children's allowance is overly generous: many nonpoor households receive it (see section V below). Among the nonpoor households receiving child allowances, about 12 percent and 6 percent of households in the middle and high income deciles respectively, are ineligible-a clear reflection of the inefficiency in implementation. (e) Households living in the North-East region-the poorest in Romania-have a higher probability of receiving transfers than in other regions, though the margial effect is more substantial in urban North-East than in rural North-East. (f) One surprsing finding is with respect to social assistance pensions (including special aid for the handicapped). Its distnbutional impact appears to be very limited in both mral and urban areas. 19 Table 18: Shares of Expenditure, Nunber Une ployed and Unemployment Allowance Total Rural Urban Decile Expenditure Total Unemployment Expenditure Total Unemployment Expenditure Total Unemployment Unemployment Allowance Unemployment Allowance Unemployment Allowance 1 2.8 11.0 7.6 4.7 16.7 12.5 1.5 6.1 2.2 2 4.6 15.8 13.1 7.7 20.0 18.7 2.4 12.1 6.7 3 5.8 12.5 12.4 8.5 12.6 17.3 3.8 12.5 12.1 4 6.8 13.6 15.9 8.2 12.6 10.3 5.9 14.5 20.8 5 7.9 8.2 8.2 8.7 7.0 10.2 7.3 9.3 9.9 6 9.2 11.9 14.7 9.7 8.8 7.9 8.9 14.5 20.7 7 10.6 9.1 8.0 9.9 7.0 5.0 11.0 10.9 7.2 8 12.3 5.8 7.1 9.9 4.7 5.7 14.1 6.9 6.4 9 15.1 7.6 7.9 13.1 7.0 7.9 16.5 8.1 9.3 10 24.8 4.5 5.1 19.7 3.7 4.1 28.4 5.2 4.9 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 N 2650 463 2650 1261 215 1261 1389 248 1389 Source: Romania IHS, 1994. 20 Table 19: Distribution of Households by Remittances Received for Absent Members Ruri Urban Remitances as a Remittances as a share of total share of total Households receiving household Households receiving household Decile remittances (%) expenditures remitinces (%) expenditures 1 7.6 2.1 8.2 3.4 2 4.6 1.1 4.7 2.4 3 2.9 0.7 8.4 2.4 4 6.1 1.6 7.8 2.4 5 5.4 1.5 8.1 1.5 6 4.7 0.6 10.0 2.6 7 2.7 1.5 5.5 1.1 8 7.8 2.2 6.8 1.8 9 5.4 1.6 7.8 2.8 10 1.8 0.4 11.9 6.1 All 4.9 1.3 8.3 2.8 N 1261 1261 1389 1389 Source: Romana IHS, 1994. V. Transfers and Poverty 5.1 Whether considered in relation to gross income, net income or expendit , the prevailing transfers do contribute to a reduction in headcount ratio of poverty-the impact is more subsntial in urban areas (8 percentage points) than in rural areas (3 percenage points)-see Table 20. In terms of poverty gap ratio also, the impact of transfers is clearly more pronounced in urban than in nrual areas. The direction of results is similar whether one considers food poverty ratio, and poverty ratio. This means that overall, transfers have been more effective in urban than in rural areas. 5.2 The impact of transfers on poverty and poverty gap ratio is very pronounced in relation to expenditures (Table 20) than incomes, in both rural and urban areas. The effect of transfers on the reduction in the ineqality of expenditures is more significant than in the reduction of inequality of incomes. 21 Table 20: Effect of Transfers on Gini and Poverty Ratios Poor Ultrapoor Without Transfers With Transfers Without Transfers With Transfers ToMl Total Total Total Net Gross Expend- Net Gross Expend- Net Gross Expend- Net Gross Expend- Income Income iture Income Incorme iture Income Income iture Income Income iture (excluding transfers) (excluding tAnsfers) Rur Gini 0.49 0.47 0.37 0.47 0.45 0.34 0.49 0.47 0.38 0.47 0.45 0.34 HC Ratio 22.6 19.5 34.0 18.3 16.4 26.6 17.3 15.6 24.2 13.7 12.6 16.5 PG 1 10.5 9.5 12.8 7.8 7.2 8.2 8.5 7.7 9.2 6.0 5.5 5.3 PG 2 7.1 6.5 7.7 4.9 4.5 4.0 6.0 5.5 6.1 3.9 3.6 2.7 Urban Gini 0.39 0.38 0.32 0.35 0.35 0.30 0.39 0.38 0.32 0.35 0.34 0.30 HC Raio 28.0 20.5 20.4 20.3 14.7 13.9 18.9 13.6 12.8 13.3 9.95 7.2 PG 1 11.9 8.7 6.2 7.2 5.2 3.4 8.8 6.5 3.5 4.5 3.2 1.4 PG02 9.4 7.1 2.8 3.7 2.7 1.2 8.7 6.7 1.5 2.3 1.7 0.4 Source: Romania IHS, 1994. 5.3 How does the existing child allowance transfer compare with a uniformly distributed lumpsum tansfer to every household? The results of this simlation are compared with the actual (current) allocation of child allowance transfer in Table 21. Actual (current) has a greater impact on the depth and severity of poverty than a uniform wansfer, confirming the favorable impact of child allowance (as they are currn y operating) on aggregate poverty. Table 21: Simulations - Based on Total Consumption Redistribute children's allowances equally to all households (lumpsum transfer to every household) Poor Ultrapoor Rural At Present Simulation At Present Simulation Gini 0.34 0.35 0.34 0.35 HC Ratio 26.6 27.0 16.5 17.9 PG 1 8.2 8.6 5.3 5.7 PG 2 4.0 4.3 2.7 2.9 Urban Gini 0.30 0.30 0.30 0.30 HC Ratio 13.9 15.1 7.2 7.6 PG 1 3.4 3.9 1.4 1.8 PG 2 1.2 1.5 0.4 0.6 Source: Romania IHS, 1994. 22 5.4 How much scope exists for further targeting of child allowances? Table 22 presents two simulations where children's allowance is denied to the top 20 percent of households, and top 40 percent of the households, and is redistributed to the bottom 20 percent in accordance with number of children. The simulation assumes that the resources saved are redistributed in such a way the poorest households receive it-whether located in rual or urban areas. Since the objective is to analyze welfare gains, the impact on various poverty measures is estimated with reference to total expenditures. Simulations are done both for poverty and food poverty. Table 22: Simulations - Based on Total Consumption 1. Redistribute top 20% children's allowance to bottom 20% based on number of children Poor Ultrapoor At Present Sxnulation At Present Simulation Rural Gini 0.34 0.33 0.34 0.33 HC Ratio 26.6 24.5 16.5 14.6 PG 1 8.2 7.3 5.3 4.6 PG 2 4.0 3.4 2.7 2.2 Urban Gini 0.30 0.29 0.30 0.29 HC Ratio 13.9 13.9 7.2 6.6 PG 1 3.4 3.0 1.4 1.0 PG 2 1.2 0.9 0.4 0.3 2. Redistribute children's allowance from top 40% to bottom 20% based on number of children Poor Ultrapoor At Present Simulation At Present Simulation Rural Gini 0.34 0.33 0.34 0.33 HC Ratio 26.6 21.6 16.5 13.4 PG 1 8.2 6.2 5.3 3.9 PG 2 4.0 2.9 2.7 1.7 Urban Gini 0.30 0.29 0.30 0.29 HC Ratio 13.9 13.4 7.2 4.8 PG 1 3.4 2.4 1.4 0.7 PG 2 1.2 0.7 0.4 0.2 Source: Romania IHS, 1994. 5.5 The gains are substantial when the top 40 percent of the households are denied access to child allowance, particularly m rural areas. The headcount poverty rato falls by 5 percentage 23 points, and the poverty gap ratio falls by 2 percentage points. The gains to the poor in urban areas are mininal: the redistribution is essentially occurring from the urban nonpoor to the rural poor. The direction of results are similar whether considered for poverty or for food poverty. 5.6 Child allowances are currently reaching households way above the poverty line. In the sample, 81.8 percent of child allowances (as well as other transfers) are being received by noupoor households (see Table 23). The above simulations suggest that scope exists for redistributive gains if child allowances are provided only households below a certain income threshold. However, the threshold needs to be carefully set: any level equal or close to the poverty line should be avoided, for two reasons. First, many households are above the poverty line currently because they are receiving child allowances. Second, households-particularly in economies in transition-constntly move up anxd down the poverty line; so it would be both impractical and inappropriate to use a level around the poverty line as the cut-off point for targeting child allowances. 5.7 Minmumn Income Guarantee. The Government is considering a proposal to provide social income support to families as follows: single person families 45,000 lei two-person families 81,000 lei three-person families 113,000 lei four-person families 142,000 five-person families 169,000 For every person beyond 5-persons: 25,000 lei per person. The social income support rate is defined as the differental between the current monthly net income of a family (net of child allowance) and the social income levels shown above. The above cut-offs rates are to be increased by 5,000 lei for each working member of the family who supplies evidence of wage-employment or self-employment. The proposal virually amounts to a Minimum Income Guarantee (MIG) for every household. 5.8 Using the IHS for 1994, we derived the cost of the above scheme under two scenarios (a) excluding child allowance in the calculation of exising monthly net mcome of a family, and (b) including child allowance m the monthly net income of a family. Assuming the survey is representative of the whole country, the cost of extending the scherme to the whole population is estimated (Table 24). As percent of GDP, the scheme costs 2.6 percent if child allowances are given on the top of a minimum income guarantee, and 2.3 percent if othermwse. The scheme would account for 6 percent of the budget expenditure, and about 55 percent of the budget deficit. Considering that child allowances have declined from 2.8 percent of GDP in 1991 to 0.8 percent of GDP in 1994, the proposed scheme basically restores the present position of public (transfer) support to the situation prevalent in 1991. 5.9 The above estimates are based on "perfect targeting", i.e., the Govermnent is able to correctly assess and monitor incomes of every member of every household. The fact that child allowances, based on the easily identifiable criterion of rmmber of children in a particular age group, are characterized by a high proportion of exclusion errors casts doubts on the prospects of MIG based on income-based means-testing. MIG is particularly difficult to implement in a period 24 of rapidly changing sources of income; the growing incomes in the informal sectors are difficult to assess and monitor. To avoid exclusion errors, it may be preferable to reinforce income-based meanstests by indicator targeting such as the number of children, age and sex of the head of the household, etc. 5.10 Moreover, there appears to be no valid argument for retaining child allowances when every household is guaranteed a minimum income. It may be discontinued altogether, or restricted to very poor families. If the latter option is chosen, the families accessing child allowances-the ultrapoor in the country-can then become an already "screened" vulnerable group to which assistance in other sectors (e.g., health, education) can be targeted. 5.11 Finally, the implications of MIG for work incentives need to be fully explored. If social assistance fully covers the gap between the actual income and the poverty line, most households at margin may prefer not to work. This is a classic problem in all such income compensatory welIfre schemes; the scheme as is being implemented in Romania does not appear to have adequate built-in safeguards to avoid adverse incentives for work (other than monitoring incomes every three months, which is extremely expensive in terms of administrative cost). Table 23: Shares of Transfers: Poor and Nonpoor Percent Distribution of: Households Children's Social Unemployment Total In-kind No. of Allowances Assistane Allowance Cash Transfer Children Transfer Top 20% 15.9 38.8 13.1 17.7 47.5 12.8 Top 40% 35.5 52.2 28.1 36.1 73.4 30.6 Top 60% 56.7 66.7 51.0 57.0 88.3 49.3 Nonpoor 81.3 83.7 84.5 83.2 95.9 75.8 Total 11.22 3.76 1.00 2.65 2.72 Transfer amount received (in million lei) (52.7) (17.7) (4.7) (12.4) (12.8) Note: The number in parenthesis is the peramage to total social assis. Source: Romania IHS, 1994. 25 Table 24: Cost of the Proposed Minimum Income Guarantee Scheme Household Size Number of Households Eligible .____________ Households for Assistance Expenditures 1 2 3 4 5a 6b 1 415 77 77 1,252,247 1,252,247 2 674 103 102 2,507,150 2,432,315 3 541 127 122 5,070,989 4,702,546 4 536 158 139 7,972,271 6,631,860 5+ 435 202 187 18,392,498 15,881,857 Total 2601 667 627 35,195,155 30,900,825 Per capita cost of transfer per annum: 52,296 45,915 Total cost income (in billion lei): 1,202.8 1,056.0 As percent of GDP (1994): 2.5 2.2 As percent of budget expenditure (1994): 6.8 5.9 As percent of budget deficit (1994): 62.1 54.5 a Children's allowances are protected (excluded) while calculating current net income. b Children's allowances are not excluded from current net income. Source: Romania IHS, 1994. 5.12 Analysis also suggest that scope exists for targeting of social assistance pensions including special aid for the handicapped and aid from the social assistance fund. For example, simulations suggest that denying social assistance to the top 20 (and 40) percent of households and redistributing the saved resources to the bottom 20 percent would reduce headcount ratio and poverty gap ratio, though not to the same extent as child allowances (see Table 25). 26 Table 25: Simulations - Based on Consumption 1. Redistribute top 20% Social Assistance Pensions* to bottom 20% Poor Ultrapoor At Present Simulation At Present Simulation Rural Gini 0.34 0.34 0.34 0.34 HCRatio 26.6 24.8 16.5 15.5 PG 1 8.2 7.4 5.3 4.7 PG 2 4.0 3.6 2.7 2.4 Urban Gini 0.30 0.29 0.30 0.29 HC Ratio 13.9 13.9 7.2 6.5 PG 1 3.4 3.1 1.4 1.2 PG2 1.2 1.0 0.4 0.3 2. Redistribute Social Assistance Pensions* from top 40% to bottom 20 % Poor Ultrapoor At Present Simulation At Present Simulation Rural Gini 0.34 0.33 0.34 0.33 HC Ratio 26.6 24.3 16.5 14.6 PG 1 8.2 7.1 5.3 4.5 PG2 4.0 3.4 2.7 2.2 Urban Gini 0.30 0.29 0.30 0.29 HC Ratio 13.9 13.9 7.2 6.3 PG1 3.4 3.0 1.4 1.1 PGS2 1e1.2 1.0 0.4 0.3 Source: Romania TIHS. 1994. * These include social assistance pensions, special aid for the handicapped and aid from the social assistance fund. 5.13 The analysis also suggests that the case for targeting of in-kind transfers appears particularly strong. In fact, most in-kind transfers are currently accruing only to the nonpoor. VI. Condusions and Policy R mmadations 6.1 Though cash transfers are contibuting to a reduction in headcount poverty and poverty gap ratio, the transfers are currently mistargeted substantially to the nonpoor, while many deserving poor-such as households with the least education of household head and female-headed households-may not be receiving child allowances particularly in rural Romania. Simulations suggest that significant gains can be realized by better targeting. Nearly a half of the current spending on child allowances can be saved by targeting to households below, say, twice the level of poverty thrshold income. However, the threshold level for denying child allowance should not be 27 set too close to the poverty line, to avoid a worsening of headcount poverty and poverty gap ratio. Moverover, while targeting child allowances to the most needy, care needs to be exercised that any deserving household is not excluded. This appears to be the case at present. 6.2 Unemployment allowance is clearly better targeted and more equally distnbuted. 6.3 There is a strong case for excluding households in the top three income deciles from social assistance benefits, and in-kind transfers. 6.4 Rual-urban disparities in transfer receipts are cuntly substantial. Targeting and refocusing of child allowance would result not only in reducing the poverty gap overall, but also would contribute to a narrowing of the rural/urban gap in transfers, by redistributing incomes from urban nonpoor to the rural poor. 6.5 The proposed Minimum Income Guarantee (MIG) scheme, under conditions of perfect targeting (zero leakage), is estimated to cost the govemment 2.6 percent of GDP or 6 percent of current budget expendte. Given ineffectve targeting of child allowances at present, it appears difficult to implement MIG based on income-based meanstests. To avoid exclusion errors, it may be preferable to reinforce meanstests with indicator-based targeting. Moreover, the implicatons of MIG for (a) targeting efficiency, (b) insitudonal capacity and administatve cost per lei of income transferred, and (c) incentives for work and labor markets, need careful thought before it is implemented. 28 References Milanovic, Branko. 1994. "Cash Social Transfers, Direct Taxes, and Income Distribution in Late Socialism," Journal of Comparative Economics, 18:2, pp. 175-197. Rashid, Mansoura. 1994. "Household Welfare in a Transition Economy: Growth, Equity and Poverty in Romania, 1989-92." Mimeo, EC1HR. The World Bank. van de Welle, Dominique, M. Ravallion and M. Gautam. 1994. How Well does the Social Safety Net Work: The Incidence of Cash Benefits in Hungary, 1987-89. LSMS Working Paper 102. 29 I