Report No. 26169-RO Romania Poverty Assessment (In Two Volumes) Volume II: Background Papers September 30, 2003 Human Development Sector Unit Environmentally and Socially Sustainable Development Unit Europe and Central Asia Region Document of the World Bank FISCALYEAR January 1 - December 31 CURRENCYEQUIVALENTS CurrencyUnit =RomanianLEI (ROL) 1Lei =0.0003US$ US$1= 32,795 Lei WEIGHTSAND MEASURES Metric System ABBREVIATIONSAND ACRONYMS CASPIS Romania Anti-Povertyand SocialInclusion Commission CEM Country EconomicMemorandum ECA Europe and Central Asia ECSHD Europe and Central Asia Human DevelopmentDepartment EU European Union ER Employmentand Relocation FRDS GDP Gross Domestic Product IBRD International Bank for Reconstruction and Development IMF International Monetary Fund LFS Labor Force Survey MDG Millennium Development Goals MIG Minimum Income Guarantee MOLSS Ministry of Labor and Social Solidarity NAE National Agency for Employment OECD Organization for Economic Cooperation and Development SOE State Owned Enterprises TR Trainingandre-Training Vice President: ShigeoKatsu CountryDirector: h a n d Seth Acting SectorDirector: Maureen McLaughlin SectorManager: Arup Banerji Task Team Leaders: Cem Mete and Nicholas Burnett Acknowledgements This report was prepared by Cem Mete (ECSHD) and Nicholas Bumett (consultant), task team leaders. Contributors include Emil Tesliuc (HDNSP) and Lucian Pop (consultant) for the poverty profile; Roberta Gatti (DECRG) on growth and poverty; Cem Mete on labor, Emil Tesliuc, Richard Florescu (ECSHD) and Lucian Pop on social protection; and Maria Amelina (ECSSD), Stephen Knack (DECRG) and Dan Chiribuca (DECRG) on the poor in inter-household and community networks. Their full background reports are contained in Volume 11. Useful material was also provided by Rosalinda Quintanilla (ECSPE) on the macroeconomic background and by Henry Gordon (ECSSD) on agriculture and rural poverty. Shirley Liu, Denis Nikitin, Lu Wang and Min Ouyang helped to organize and analyze the data sets. A h a Barsan and Diana Marginean contributed to the design and implementation o f the household survey. Jennifer Manghinang and Anna Goodmanproduced the manuscript. Peer Reviewers for the report were Robert Chase, Philip Keefer, Peter Lanjouw and Dena Ringold. Helpful written comments were also received from Asad Alam, Ronald Hood and Kari Nyman. Randa el-Rashidi and the quality team o f the Social ProtectionAnchor helped inorganizing a Quality Enhancement Review, where the team benefited from the suggestions o f Gordon Betchennan, Margaret Grosh, Stefan0 Scarpetta and Quentin Wodon at an early stage o f the project. The team i s grateful to Ziad Alahdad, Mukesh Chawla, Annette Dixon, John Innes, Maureen Lewis, Alexandre Marc, Edmundo Murmgarra, Mamta Murthi, Reema Nayar, Catalin Pauna, Mansoora Rashid, Andrew Vorkink and M y l a Taylor Williams for their support and advise at various stages o f the project. Finally, in addition to those who are listed until this point, at the concept review stage the team received useful comments from Daniela Gressani, Dominic Haazen, Kathy Lindert, Pierella Paci, Silviu Radulescu, Ana Maria Sandi and Radwan Shaban. Invaluable assistance was provided in Romania by the government's Anti-Poverty and Social Inclusion Commission and by the National Institute o f Statistics. The report also benefited from discussions with the staff o f the Ministry o f Labor and Social Solidarity, the National Pension Authority, and National Employment Agency. The final views expressed by this report are those o f the World Bank team. Volume 2: Table of Contents Page Poverty inRomania: Profile and Trends during the 1995-2002 ................................ 1 Poverty and growthinRomania: 1995-2002 .............................................................. 55 Labor Force Participation. Unemployment and the Poor ........................................... 83 Protecting the Poor andVulnerable .......................................................................... 118 Mapped in or Mapped out? TheRomanianPoor inInter-Household and Community Networks ...................168 Poverty in Romania: Profile and Trends during the 1995-2002l Emil Tesliuc. Lucian Pop and Filofteia Panduru 1.Welfare and poverty measurement ...................................................................................................... 1 A.Welfare and poverty concepts ........................................................................................................ 1 B.Data sources.................................................................................................................................... 2 3 D.Estimation ofthe Poverty Lines ..................................................................................................... C. Constructiono f the welfare aggregate ............................................................................................ 5 7 F. Survey Design, Estimation and Inference....................................................................................... E.Poverty Indicators and their Estimation.......................................................................................... 8 2.Who 8 A Consumptionpoverty ...................................................... . Are Romania's Poor?.................................................................................................................. ........................................................ 8 B. Individual dimensions ofdeprivation .............................. .......................................... 10 C.Community poverty ...................................................................................................................... 11 D.Perceptions ofpoverty.................................................................................................................. . 12 13 F.Changes inInequality.................................................................................................................... E Intemational comparisons ............................................................................................................. 14 3. PovertyDynamics: 1995-2002 .......................................................................................................... 16 A.Poverty, growth and inequality..................................................................................................... 16 B.Assessing the Robustness of ObservedChanges inPoverty over Time....................................... 16 C.Assessing Changes inPoverty over Time without Poverty Lines ................................................ 17 D. PovertyDynamics andVulnerability............................................................................................ 18 4. Revenues and consumption................................................................................................................ 21 5. A Profile ofthe Poor in2002 ............................................................................................................. 24 A.Groups at HighPoverty Risk........................................................................................................ 24 B.A ConditionalProfile ofPoverty .................................................................................................. 32 36 Annex 1. Adult equivalence scale.......................................................................................................... 6. Conclusions and Policy Recommendations ....................................................................................... 40 Statistical Appendix ............................................................................................................................... 42 1 W e thank Nicholas Bumett. Peter Lanjouw. Cem Mete. Roberta Gatti and Richard Florescu for their comments and suggestions. 1. Welfare andpoverty measurement This paper focuses mainly on monetary dimensions o f well-being, or consumption, to informthe policy dialogue in Romania on the extent and dynamics of poverty, as well as the main factors associatedwith poverty. The results are based on two comparable, nationally representative surveys: (i) RomanianHouseholdBudgetSurvey(ABF,uponitsRomanianacronym) fortheperiod2001- the 2002; and (ii) the Integrated Household Survey (AIG) for the period 1995-2000. Other dimensions o f deprivations have beencaptured for 2002 using the Living Conditions Survey (ACOVI). While it concentrates on monetary poverty, the paper captures other dimensions o f material poverty such as caloric intake; crowding and living conditions, ownership o f essential durables. Inaddition, it assesses poverty defined as capability to function in society, or adequate level o f investments inhealth or education, or the lack o f unemployment. Finally, the report reviews community poverty and perceptions o f poverty. This section presents the mainelements o fthe methodology usedto estimate poverty inRomania. This i s not the first attempt to produce and analyze poverty counts in Romania (World Bank, 1997; Wagner et all. 1998; Molnar 1999; Chirca and Tesliuc, 1999; OECD 2000; Tesliuc, Pop and Tesliuc, 2001). It differs, however, from earlier analysis in using a new methodology to estimate poverty, paying particular attention to the comparability o f the welfare indicator across time, space and households o f different sizes and compositions. Households are ranked from the poorest to the richest based on their real consumption per adult equivalent, and indicator that: (a) includes only commodities not affected by seasonality; (b) uses an empirical equivalence scale that takes into account economies of scale and the relative cost of children over adults; and (c) deflated current expenditures with a robust price index that detects the variation inthe living standards across time and areas o f residence. A. Welfare and poverty concepts Throughout most of the paper, poverty i s measuredbased on per adult equivalent consumption. Poverty analysis requires a welfare measure, observable and measured in a consistent way across households, space and time. Two typical monetary measures o f welfare are income and consumption: The paper uses consumption to compare welfare across households, space and time because consumption reflects better than income a household's actual standard o f living and its ability to meet basic needs. Unlike income, consumption reflects the ability o f a household to borrow or mobilize other resources intime of economic stress. Inmany instances, consumption will capture better than income questions o f access to and availability o f goods and services. The choice o f a welfare indicator based on household consumption was also dictated by country- specific considerations. In Romania, consumption data i s collected more reliably than income data. The latter tends to suffer from incomplete measurement (informal income, from farming or small business, are poorly captured inthe surveys), underreporting, and seasonality. Although income as basic welfare indicator for EUpoverty monitoring and Romania i s on its road for EU accession, the paper does not report income-based poverty measures. Annual household income i s poorly measured inthe ABF, due to (i) a short recall period (one month) and a large share o f seasonal income sources in total income; and (ii) the poor recording o f informal sources o f income. Some regular (non-seasonal) sources o f income, such as wages and transfers, are accurately measured by the survey. Unfortunately, more volatile sources, such as farm or some types o f self-employment income cannot be determined at the household level, given the one-month recall-period. At one extreme, similarly endowed farm households will appear rich if surveyed after the harvest, and poor if surveyed before. Moreover, there is little information inthe survey to estimate a net income indicator for informal activities, such as self-employed in non-agricultural and agricultural activities. Without such information, one cannot compute a robust annual household income indicator. The income measure that can be derived from the household survey would tend to overestimate poverty for households deriving their livelihoods from informal (such as farming, self-employment) or seasonal (such as farming, tourism, or construction) activities. B. Data sources The information on consumption poverty i s derived fiom two comparable, nationally representative surveys: (i) RomanianHouseholdBudgetSurvey(ABF,uponitsRomanianacronym)fortheperiod2001-2002; the and(ii) IntegratedHouseholdSurvey (AIG) for the period 1995-2000. the The ABF (AIG)is a multi-purpose nationally representative survey adrmnisteredby the RomanianNational Institute for Statistics (INS) in cooperation with the Ministry o f Labor and Social Solidanty, and designed with the technical assistanceof the World Bank. The survey was first admmistered inApril 1994, and continued since. The survey aims for an annual sample o f 36,000 households, in fact 12 repeated cross- sections o f 3,000 households interviewed for one month during the yea?. Each month, responses are gatheredfrom 2,600-2,800 households out o f 3,000 selectedhouseholds. These households provide detailed informationregardingdemographics, assets, labor market activities, income, purchases and consumption for that month only. The information is collected using a household questionnaire (admmistered inthree visits by trained interviewers), complemented by a diary. The diary is usedto help the household keep track of cashflows: incomes, expenditures, and savings. The strength o f the ABF (AIG) is inmeasuring monthly current consumption, i.e. household purchases o f food, non-food and services, as well as consumption o f food out o f own production. The food consumption module collects information about the consumption o f 104 (83) commodities, usinga balance approach. The household reports the initial stock o f that commodity, the inflows and outflow during the month, and the final stock. The inflows are split between (the value and quantity of) goods bought (Bo),quantities produced on-farm, derived fiom processing, received ingift or inexchange withother commodities. Outflows consist o f (the value and quantity of) goods sold, processed, given as gifts, used as farm inputs, exchanged and goods consumed (Cs) by the members o f the household. One can determine consumption out o f own production as Sc =max (0, Cs -Bo), and the amount purchased as food for the household members as Pr = Cs - Sc. The non-food module collects information on the purchases of 121 (113) items, mostly as total monthly outlays. Similarly, the services module records information on the value o f the monthly purchases o f 89 (56) items. The survey does not record the self-consumption o f non-food or services. The survey collects information on few durable goods ownedbythe household. Although AI3F and AIG use different questionnaires - with the ABF questionnaire collecting more detailed consumption information than the previous AIG - the two surveys are perfectly comparable. The primary information for the two questionnaires i s collected using the same household diary, which captures the consumption o f food, non-food and services at a greater level o f detail. High level o f literacy and compliance by the respondents makes the diaries an important survey instrument. Moreover, the two surveys use the same team o f interviewers, and are based on the same sampling design. 2 Initially, the survey was designed as a rotating panel, with 50% of the households interviewed in a given year to be interviewed again the next year. This feature was not respected during implementation. Some households have been interviewed for 2, maximum 3 consecutive years, especially during 1995-1998. N o panel element was maintained since 2001, with the implementation of the ABF. Box 1:Are the ABF and AIG Surveys Comparable? The new household budget survey (ABF) that "replaced" the former AIG in 2001 collects consumption information at a greater level of detail, similar with the structure used in the European Union. A more detailedlist of consumptionitems can trigger higher recall from respondents, and result ina higher reported consumption.Ifthis is the case, thenthe level ofpoverty estimatedin2001 will be artificially low, the result of a change in survey instruments,not inthe economic fundamentals.This was not the case in Romania, as primary survey data continued to be collected using the same, very detailed household diary. The main function of the two questionnaireswas to aggregate the information collected in the diaries into a smaller number of items. Fortunately,the statementthat the two surveys report the same level of householdconsumptiondespite their different level of detail can also be demonstrated, not only be claimed. The reason is the implementation of a true random experiment in 2000, during the period of pre-testingthe ABF. The introduction of the ABF survey in 2001 was preceded by a 6 moths testing period, where the two surveys (ABF and AIG) were implemented in parallel, on two randomly chosen samples. We did not found significant statistical differences between the mean household consumptionof the parallel surveys, at national level or at more disaggregated level (by area of residence). Moreover, the distribution of householdconsumptioninthe new ABF was similar with the AIG (Figure l), for total consumption and by component (for food consumption, non-foodconsumptionand services). Figure 1. ABF and AIG Surveys Report the Similar Household Consumption Distribution of Total HouseholdConsumption Distribution of HouseholdConsumption Under 2 Alternative Questionnaire Designs Under 2 Alternative Questionnaire Designs Food Non-Food Services 0 1 HOUSehOldConsumption.3Million ROLS 2 4 5 C. Construction of the welfare aggregate The consumption aggregate used in this report to rank households includes the consumption of food (including consumption out o f own production); non-food; services; and selected durables3. Robust welfare comparisons require the estimation of a comprehensive consumption aggregate. However, despite our attempts to compile a comprehensive welfare indicator, the information included in the RIHS (AIG/ ABF) does not allow a precise enough estimation of a number o f important components of householdconsumption, such as housing, some durables, or fkee-of-charge in-kind services.Constrainedby data availability, our consumption aggregate ignores housing or consumption out o f household production of non- food or services. Other components of householdconsumption, such as publicly provided education or health services, or other in-kindpublic services, are ignorednot only because(some of them) are poorly capturedin the householdsurvey, but also because the difficulty in finding suitable prices. The same problem applies in a large measure to durables. As a result, only consumption of food, purchases of nonfood and services, and the use value for a small number of durables can be estimated with the available data. Constrained by data quality and 3 To ensure comparability across time during a high-inflation period, each component o f current consumption i s inflated with the corresponding CPI (for food, non-food and services) in December 2002 constant prices. During 1995 to 2002, food rural prices were systematically lower than urban one, indicating higher purchasing power in rural areas for the same nominal expenditures. To arrive at comparable consumption figures across regions and areas o f residence, location-specific consumption needs to be adjusted by area- specific deflators. Ignoring the problem and using the same price index across space would overstate poverty in rural areas (were prices are lower) and understate it in urban areas (were prices are higher). To account for differences in the cost of living across areas of residence,rural consumption is further inflated in urban prices, using a Laspeyres rural-urban price index constructed from the unit-value information collected inthe survey (Table Al, Statistical Annex). The cost o f living inrural area was between 2 to 7% cheaper inthis period. Seasonality in consumptionwas smoothedusing a seasonality index (Table A2, Statistical Annex). Household monthly consumption has a strong seasonal pattem, which raises the issue o f comparability between the monthly waves o f the survey. Given seasonality, the hypothesis that the randomly chosen households are coming from the same distribution o f consumption does not hold. In any year, December food consumption peaks about 25% above the national average, as households are hitby the Christmas spending spree. Thus, the consumption o f the December sample i s not similar with the other waves o f the survey. If we do not correct for seasonality, we found fewer poor during this month, despite the fact that all other structural parameters o f the population are similar with the other months. If the distribution of income or consumption is affected by seasonality, then poverty would be overestimated. To correct for consumption seasonality, indices4have been constructed for each month, quintile and area o fresidence. Differences in needs across households o f different size and demographic compositions were made comparable using adult equivalent and per capita measures. The results reported inthe paper are based on per adult equivalent consumption. The use o f equivalence scales i s common in the poverty literature and the practice o f the social protection policy' inRomania. To investigate how sensitive are the conclusions to the choice of equivalence scale, poverty statistics based on per capita consumption were also computed. Using per capita expenditure, one would found higher poverty in Romania, especially for larger households. Other findings remain, however, qualitatively similar. Equivalence scales acknowledge that the needs o f children may be different than those o f adults, and that larger households may need less than their individual members living separately to enjoy the same level o f welfare. The later i s attributable to economies o f scale in consumption, associated to the consumption o f non-rival (public) goods. The adult equivalence scale used in this paper is based on two parameters: the cost o f children relative to adults and the economies o f scale enjoyed at household level. The estimations made on the basis of FUHS 1999 and 2000 indicate that a child cost 50% o f an availability, we termed this incomplete indicator o f welfare as total household consumption. It is, however, only a partial welfare measure. The ratio of our mean household consumptionper capita to the corresponding System o f National Accounts estimate i s only 42 percent in 2001 In2001, the final consumption o f the household sector was estimated, in the System o f National Accounts, to 921,805.1 billion ROL. The annualized value o f household c onsumption e stimated using the ABF (HBS) c onsumption v ariable, adjusted withthe appropriate expansion factors, was estimated at 387,002.04 billion ROL, or 42 percent.. 4 The seasonality index for a given month, quintile and area i s the ratio o f the predicted value of consumption for that month, quintile and area to the national average. > The MinimumIncome Guarantee program uses a set o f thresholds that increase at a slower rate than the household size, acting as an economies-of-scale parameter. 4 adult, and an economies o f scale parameter 0 = 0.90 (see Appendix 1). The formula used to determinethe number of adult equivalentsis thus: AE = (A+0.5C)A0.9 These results are comparable with other recent studies on equivalence scale inRomania (Betti, Molnar and Panduru, 2002)6. The use o f this scale i s an improvement over other scales used in poverty analysis on Romania, most notably the scale devised based solely on the nutritional requirements o f individuals, by age and gender. The scale used in this report takes into account other dimensions o f welfare besides food consumption and nutritional requirements (i.e. economies o f scale inconsumption o f durables, utilities or housing). In contrast, the nutritional scale accounted only for difference in "needs" in food consumption. Among other things, the nutritional scale implicitly assumed that females and, respectively, elderly would be better off than adult men even if they would consume 15% or 40% less than an adult men. Such judgments can hardly pass as valid in Romania, and elsewhere. Due to the implicit assumptions embodied in the construction o f the nutritionist scale, earlier studies found no differences in poverty rate by gender, or for the elderly. The new methodology eliminates the bias against women, elderly and rural inhabitants, associated with the former methodology used in Romania until 2001. For each year, households are ranked by the level o f real household consumption per adult equivalent. This variable is usedindistributional analysis throughout the report. D. Estimationof the PovertyLines As most poverty assessments, we use the cost of basic needs method to determine two poverty lines for December 2 002. The two 1ines s eparate the (total) poor and extreme poor from the rest o f t he population. Each poverty line include a food component (common to both lines), plus an allowance for essential non-foods and services (different for each line). The food component o f the poverty line i s determined as the cost o f a food basked preferred by the individuals from the second and third quintile, priced at the unit values faced by this group, with quantities scaled up proportionally to give a caloric intake o f 2550 calories per adult per day' (Table A3). The 2550 calories/adult/day conforms to the recommendations o f the FAO/OMS/UN Expert Group onNutrition( 1989,a s well with the nationalrequirements (Ministryof Health, 2 002). For instance, the average consumption vector consumed in 2002 by the individuals from the second and third decile corresponds to a caloric intake o f 2610 caloriedper aduldday, or 4% higher than the requirement. The vector o f quantities consumed was thus multiplied by 0.962, and then priced at the unitvalues specific to this group. The resultingfood component ofthe poverty line, expressed inDec- 02 urban prices, was 872,005 ROL. The extreme poverty line is determined by summing up the food component of the poverty line with the amount o f non-food and services typically consumed by those whose total consumption equals the 6 The scale proposed by Betti, Molnar and Panduru attributes the weight o f one for the first adult aged 19-64, 0.8 for any subsequent adult or children aged 15-18, 0.6 for subsequent adults aged 65 or over, 0.5 for children aged 6-14 and 0.3 for children aged 0-5. 7 For per capita consumption, the food component o f the poverty line was estimated such as it would provide a recommendedcaloric intake o f 2400 caloriesiper capitaiday. The value o f the per capita poverty line i s 820,472 ROLiper capita/month inDecember 2002 prices. Table A 4 inthe annex presents the main elements used inthe computation ofthis line. 5 food requirement. Ifhouseholds that can cover only their food requirements gave up on food for other consumption items, these items should be necessities. Thus, the extreme poverty line i s the sum o f food and other non-food necessities. Individuals will be classified as extreme poor if their consumption per adult equivalent will be lower than the extreme poverty line. The extreme poverty line, expressed inDec-02 prices, i s 1,060,658 ROL8. The total poverty line is determined by adding to the food component the amount spent on non-food and services by those households whose food consumption equals the food component o f the poverty line. Inthis variant, the definition o f non-food necessities is broader. Individuals will be classified as (total) poor iftheir consumption per adult equivalent will be lower than the total poverty line. The total poverty line, expressed inDec-02 prices, is 1,535,570 ROL9. The poverty lines are obtained by adding together the food and non-food (and services) components. The absolute poverty line determined this way has atangible meaning: itrepresentthe cost o f satisfylng the basic needs, the cost ofthe bundle o f food products consumedby the relatively poor that provides arecommended caloric intake, plusthe cost o fother essentialnon-foods. Figure 2 Extreme andTotal Poverty Lines Individuals, ordered by Real Consumption We illustrate graphically the estimation o f the extreme and total poverty lines inFigure 2. Individuals are ordered fi-om lower to higher consumption level (per capita or per adult equivalent) on the x axis. We plot two distributions: total consumption (the 45 degree line) and the food consumption (the curve). Next, we draw the horizontal line P equal to the food poverty line. At the intersection of this line and total consumption we found the household whose total consumption is just enough to buy the bundle o f food having the recommended energy intake. On the red vertical line corresponding to t h s household, we read how much it spends on non-food and services, as the vertical distance between the total and food , 8 The extreme poverty line for per capita consumption Was estimated at 1,009:690 ROL/per capitalmonth. 9 The total poverty line for per capita consumptionwas estimated at 1,448,175 ROL/per capitalmonth. 6 consumption curves. To find the extreme poverty line, we add a segment equal to this distanceon the top o f the red line. The heightofthis line, at Pm, gives the extreme poverty line. Findingthe total poverty line is similarly easy. First, we findthe householdwhose food consumption is equal to the absolutepoverty line at the intersection between the horizontal line P and the food consumption curve. We draw a vertical line passing through this point, until it reaches the total consumption curve. The height o f this segment is the total poverty line (Pt). Box 1. Three Indices usedto Measure Poverty The poverty headcount is the share of the population which is poor, Le. the proportion of the population for whom consumption or income y is less than the poverty line z. Suppose we have a population of size n in which q people are poor. Then the headcountindex is definedas H = q/n. The poverty gap, which is often considered as representing the depth of poverty, i s the mean distance separating the population from the poverty line, with the non-poor being given a distanceo fzero. The poverty gap is a measureof the poverty deficit ofthe entire population, where the notion of poverty deficit captures the resourcesthat would be neededto lift all the poor out ofpoverty through perfectly targetedcashtransfers. It is definedas follows: where yi is the income of individual i, and the sum is taken only on those individuals who are poor. The poverty gap canbe written as being equalto the product of the income gap ratio and the headcount index of poverty, where the poverty gap ratio is itself defined as PG=I*H. It must be emphasized that the income gap ratio I in itself is not a good measure of poverty. Assume that some households or individuals who are poor but close to the poverty line are improvingtheir standards of living over time, and thereby become non-poor. The Income gap ratio will increase becausethe mean distance separatingthe poor from the poverty line will increase(this happensbecause some o fthose who were less poor have emerged kom poverty so that those still in poverty are on average further away from the poverty line), suggesting a deterioration in welfare, while nobody is worst off and some people are actually better off. Although the income gap ratio Iwill increase, the poverty gap itself PG will decrease, because the headcount index of poverty will decrease, suggesting an improvementtowards poverty reduction. The problem with the income gap ratio is that it is defined only on the population that is poor, while the poverty gap is defined over the population as awhole. The squaredpoverty gap is often described as a measure of the severity of poverty. While the poverty gap takes into account the distance separatingthe poor from the poverty line, the squared poverty gap takes the square of that distance into account. When using the squared poverty gap, the poverty gap is weighted by itself, so as to give more weight to the very poor. Said differently, the squaredpoverty gap takes into account the inequality amongthe poor. It is obtainedas follows: E. PovertyIndicatorsandtheir Estimation The results reported in the paper use the Foster-Greer-Thorbecke class o f poverty measures: the headcount ratio, poverty gap and squared poverty gap measures (Foster, Greer, Thorbecke, 1984). All estimations, from simple comparison o f means to regressions, incorporate the survey design in computing the standard errors. Tabulations use expansion factors that correct for non-response and recalibrate the marginal distribution o f the sample toward the population distribution by age group, sex, judets, and area o f residence. 7 F. SurveyDesign,Estimation andInference Throughout the paper we use the survey design parameters to estimate standard errors and confidence intervals with usual confidence (5%). The three key parameters are stratification, clustering and expansion factors. The ABF and the AIG has no stratification, and i s based on 501 clusters (primary sampling units). To expand sample statistics to population, we use a set of post-stratification weights estimated by National Statistical Institute. The simple fact that poverty estimates are going up or down may be insufficient to draw robust conclusions on the dynamics o f poverty. One should check if the differences in poverty over time pass the conventional levels of statistical significance, and are not simply due to sampling error. 2. Who Are Romania's Poor? This section focuses on the magnitude of the poverty in Romania. There is not a unique or simple answer to this question. The answer depend on the dimension o f well-being being considered (consumption, education, health, employment), or on the level o f poverty line being considered within the same dimension. This section surveys the extent of deprivation ina number of such dimensions, to provide an as-complete-as-possible picture o f poverty. An emphasis i s placed on consumption- poverty, given its high correlation with other forms o f deprivation. The poor described inthis paper cover only one part - the largest part - o f Romania's population: all people living in dwellings, families or single-persons living alone or forming extended households. What i s it not covered are the institutionalized population (military conscripts, institutionalized children and elderly) and the homeless (including street children). The s ize o f these groups is not large, but should concern the public action for poverty reduction. The homeless" tend to concentrate extreme pockets of poverty, groups for which poverty tends to get inherited from one generation to the next. The institutionalized children and elderly" are two extremely vulnerable groups, whose well- being i s entirely depended on the availability and the quality o f care provided by the public institutions. A. Consumption poverty H o w many Romanians were consumption-poor in 2002? H o w many persons consume less than the social minimumrepresented by the poverty line? Table 1presents a snapshot of poverty in2002: 0 About 6.5 million Romanians or 28.9% o f the population consumed less than the poverty line o f 1.5 million ROL in Dec-2002 prices. On average, the consumption o f this group was 26.3% lower than the poverty line, signaling a concentration o f the poorjust below the poverty line. 0 From this group, about 2.5 million people or 10.9% o f the population were living in extreme poverty. Their consumption was 1ess than 1 million ROL in Dec-2002 prices, or the extreme poverty line. Extreme poverty is shallow: the average consumption o f the group falls short o f the extreme poverty line by only 21.9%. lo There are no national estimates regarding the number o f homeless. The number o f street children is estimatedat 1500(of which 400 livinginBucharest) by the National Agency for Children Rights Protection. " The number of children living inpublic andprivate placementcenters was over 43 thousands at the end o f 2002, as reported by the National Agency for Children Rights Protection. The reported number o f elderly living ininstitutions at the endof 2000 was about 2200. 8 Finally, 1.2 million Romanians or 5.6% o f the population do not have enough to cover their basic food needs. The minimal food needs were estimated at 875 thousand ROL inDec-2002 prices, and are e quivalent w ith the c ost of a n non diversified food basked c onsumed by the people in the second and third consumption decile that provides 2,550 calories/adult/day. On average, the consumption o f the food-poor is one fifthbelow the food poverty line. Table 1.ConsumptionPoverty in2002 Poverty Line Food Extreme Total Number ofPoorPersons 1,250,410 2,443,668 6,470,551 5.6% 10.9% 28.9% Poverty Gap - Average ConsumptionDeficit PovertyHeadcount- % Poor inPopulation 0.011 0.024 0.076 Average ConsumptionDeficit amongthe Poor 0.204 0.219 0.263 Poverty Severity 0.004 0.008 0.030 Source: WorldBank estimations based on ABF 2002 Inthe remaining sections, consumption poverty i s contrastedagainst anumbero fnarrower dimensions o f w ell-being, such as nutrition, housing amenities and ownership o f e ssential durables, e ducation, health, employment, community poverty, or subjective perceptions o f well-being. As expected, there i s a high degree o f correlation (overlap in identifying the poor) in various dimensions o f well-being, especially for those dimensions related to material well-being that are included in (per adult- equivalent) consumption. Consumption poverty i s highly correlated with other measures o f material well-being, such as adequate nutritional intake, availability o f essential durables, or adequate housing amenities" (Table 2): 0 Nutrition-poverty. Against 29% o f the people who are consumption-poor, only 20% suffer from inadequate nutritional intake. Among the consumption-poor, insufficient nutritional intake touches 35% o f the people. This should be a high-risk group, as the nutritional deficit is likely caused by inadequate means, and not a deliberate choice. In contrast, 14% of the non-poor consume less than the recommended caloric intake, but have sufficient means to cover that deficit. Nutrition-poverty i s more prevalent among the urbanpoor (50%) compared to ruralpoor (28%). 0 Lack-of-essential-durables. Nationwide, 14% o f people do not have a gas stove, and 22% a refrigerator. As expected, endowment with e ssential durables is better among non poor and in urban areas. 0 Housing-poverty. Housing poverty has different features depending o f area o f residence. At national level, 30% o f the consumption-poor live in crowded households, half o f them do not have a flush toiled, and more than 70% have no bathroom. Lack o f access to basic utilities such as bathroom, flush toilet and hot water characterizes almost all rural inhabitants, while inurban areas i s found mainly among poor households. A relatively large percent o f urban poor find themselves l2 As with consumption, these indicators are estimatedat householdlevel, implicitly assumingfair sharing across householdmembers. Consumption o f food and durables in included in per-adult-equivalent consumption. As such, it quantifies the ability to purchase sufficient food for adequate nutrition or "rent" essential durables. Consumption ranks households as poor form a welfarist perspective: if a householdhas sufficient money to buy food and does not, it is not poor. However, individual preferences are heterogeneous, and some individuals consume less than the average social norms insome dimensions inorder to consume more in others. Nutritionally-poor are ranked using a normhtive perspective: if a household does not consume enough calories (given the gender-age compositionof the householdsand nutritional norms) it is poor. The same appliesfor the durables-poor. 9 inthe impossibility to cope with the public utilities costs, thus accumulating debts (over 50%) or being disconnected from the heating a n d or hot water district system (8%). Table 2. DeprivationinSpecificDimensions of MaterialConsumption National Urban Rural lonPoor Poor Total ;on Poor Poor Total JonPoor Poor Total overty status TotalPoverty 11 29 100 82 18 100 58 42 100 yPopulationinthe group 100 100 100 100 100 100 100 100 100 iwhidi: `utrition consume less than aminimum caloric intake 14 35 20 21 50 26 4 28 14 ackingconsumer assets no gas stove 1 33 14 3 10 4 15 44 27 no refrigerator 13 42 22 9 21 11 20 53 34 no washingmachine 34 14 45 22 54 28 54 84 61 lousingpoverty 3 or morepersonsper room 1 28 13 1 30 11 8 28 16 less than 5 sq meterslperson 3 16 7 2 14 4 4 18 10 no bathroom 35 13 46 9 33 13 78 93 84 no flush toilet 26 58 36 5 20 8 62 17 68 no hot water system 3 1 1 5 48 11 36 15 80 95 86 disconnectedfrom the districthot water system 2 3 3 4 8 5 0 0 0 disconnectedfrom the districtheatingsystem 3 3 3 4 8 5 0 0 0 couldnotpay intime electricitybill 12 23 15 11 26 14 14 21 17 couldnot pay intime public utlities 21 29 28 32 52 36 18 18 18 wrce: WorldBank estimations based on A. 72002 Consumption-poverty i s highly correlated with other dimensions o f material well-being, such as access to and use o f education and health services, where effective use o f publicly provided services is conditional on the investment o f the households in complementary inputs such as teaching materials or medicines (World Bank, 2002). B. Individualdimensions of deprivation All poverty measures used so far are household-level attributes that rest on the assumption that consumption i s fairly distnbuted within the household. This assumption is used because the measurement o f individual well-being in a given money-metric, such as income or consumption, i s difficult. However, education, health and employment are individual-level attributes where failure to function normally in a society can be easily measured against a given benchmark. Somebody may be considered education-poor if does not achieved a minimal stock o f education by certain age, while other may be considered health-poor if chronic illness or disability impinge upon his ability to work and function normally within a society. This type of poverty i s close to what Amartya Sen (Sen, 2000) refers at as functionings or "capabilities". Table 3 presents a number o f definitions o f education-, health- and employment-poverty, and e stimates the incidence of the phenomenon by consumption- poverty, nationally and separate for rural and urban areas. 10 Table 3. Lack of Capabilities: Education, Health and Employment Kationai Urban Rural Jon Poor Poor Total Jon Poor Poor Total IonPoor Poor Total Poverty status Total Poverty 71 29 100 82 18 100 58 42 100 Populationinthe group 100 100 100 100 100 100 100 100 100 Of which: Education Poverty age group 8-14: children not attendingschool 1.8 6.4 3.3 1.1 5.4 2.0 2.9 6.9 4.8 age group 15-48: not in school anddid not attendigraduate middle school 1.3 10.5 4.0 0.8 6.1 1.8 2.4 13.3 7.3 age group 15-24:not in school anddid not attend vocational or high-school 9.8 39.4 20.7 4.8 19.6 8.5 21.1 51.9 37.8 Health Poverty having a chronic illness that precluded usual activities for more than 2 weeks inthe last month, or disabled 7.0 5.8 6.6 1.4 6.6 7.3 6.3 5.4 5.9 Employment poverty age group 16-59:not working but looking for ajob inthe last week andready to work ifhave job opportunity 5.6 11.8 7.3 6.9 21.2 9.4 3.0 6.0 4.3 Source: WorldBank estimations based on ABF 2002 Consumption-poverty correlates strongly with the education-poverty or, in urban areas, with employment-poverty, and is less sensitive to health-poverty: Education-poverty was defined as (i) drop-out for children up to 14 years old; (ii) school drop-out or non-enrolment in middle-school for people 15-48 years old; or more restrictively as (iii) non- enrolment invocational or high-school for youth aged 15-24. By the first two definitions, there are relatively few education-poor individuals inRomania, estimated to 3%-4% nationwide. Using the more restrictive definition, one in five Romanians aged 15-24 i s education-poor. Irrespective o f the poverty measure and the reference group considered, education-poverty affects disproportionally more the poor, the youth and the rural areas. More than 50% o f rural poor aged 15-24 are not in school and did not attend vocational- or high-school. It is important to note that the relative risk o f education-poverty i s much higher for the poor inurban areas. Controlling for the area o fresidence, the risko f education-poverty is 4-8 times higher inurban areas, compared to 2-4 times higher inrural areas. Health-poverty was defined as suffering from chronic medical conditions or disability. Against this benchmark, about 6.6% o f the population is found health-poor, with small differences across areas o f residence or consumption-poverty status. This finding may indicate that health-poverty i s triggered by other, more powerful, factors such as age, gender, lifestyle and exposure to health hazards. As noted in other analyses (World Bank, 2002), the health data in the Romanian HBS may suffer from miss-reporting, Employment-poverty was defined for the working-age cohort as being unemployed. About 7% o f individuals o f working age suffer from employment-poverty, more inurban than inrural area. The highest risko f unemployment i s found for the poor urban inhabitants (21.2%). C. Communitypoverty While consumption-poverty captures deprivation inthe appropriation o f private goods - such as food, non-food, services, durables or housing amenities --,community-poverty measures the availability o f publicly provided services within a ~ommunity'~.Among the dimensions investigated are (for the rural area only) the availability o f post-office, health center, pharmacy, cultural hall or playground. Other l3 W e exploit the information collected in the Living Conditions' Survey (ACOVI 2002) to assess community-poverty, defined a$ availability o f a public service within the community. For the purposes of the report, a public service i s not available if at least 80% o f the households ina primary sampling unit reported so. 11 dimensions investigated are the perceived security o f the neighborhood and the quality o f transport infrastructure. Table 4. CommunityPoverty National Urban Rural Con Poor Poor Total i o n Poor Poor Total Ion Poor Poor Total Poverty status Total Poverty 71 29 100 82 18 100 58 42 100 Populationin the group 100 100 100 100 100 100 100 100 100 OJ-wkicli : Community lack ofpost office in the locality 6 12 7 15 18 16 lack ofhealth service in the locality 5 11 7 14 17 15 lack ofpharmacy inthe locality 14 28 18 0 1 0 38 41 40 lack o f culturallentertainmentcentre inthe locality 10 19 13 2 2 2 24 27 25 lack ofplayground for children inthe locality 36 63 44 7 12 8 85 89 87 living inan highly insecure area (affectedby violence, burglaries, etc.) 2 2 2 3 4 3 1 1 1 I living inan area wth a highly deteriorated roads infrastructure 17 ._ 17 14 i n 7 9 19 21 20 Among all dimensions o f well-being analyzed so far, community-poverty has the lowest correlation with consumption-p~verty'~(Table 4). However, it is important to note that: The incidence o f community-poverty is, for most dimensions, lower than that o f consumption- poverty. Inother words, people are more concerned with their private consumption, than with the availability o f publicly provided services. However, the measure o f community-poverty used here does not take into account potential differences in the quality o f the services among poor and not- poor neighborhoods, or potential discrimination o f the consumption-poor inthe access to such services within a certain neighborhood. There i s a strong difference in the type o f community-poverty by areas o f residence. While the rural areas are affected by lack o f access to services and infrastructure, urban areas are affected mostly by insecurity. D. Perceptions ofpoverty While objective measures o f consumption-poverty place 29% o f the Romanians as poor, with only 11% in extreme poverty, most households perceive their economic situation inharsher tones. About half o f the people declare that they cannot buy enough food, while two thirds consider than their income do not match their current expenditures. These percentages are higher for the poor than for the non poor, and do not vary substantially inrural versus urban areas. The dissatisfaction with the current level o f consumption i s substantial even for the non-poor: 42% o f them consider their income insufficient to cover their food needs, and 59% assess than current incomes are not enough to cover current expenditures. l4 It is important to note that a weak positive correlation between consumption and community poverty exist in Romania. This is not necessarily the c ase. Recent Poverty Assessments (Guatemala o r Benin) found differences between households' perception o f consumption- and community-poverty, with reductions in community-poverty and increases inconsumption-poverty. 12 Table 5. Perceptions of Poverty Fiational Urban Rural NonPoor Poor Total NonPoor Poor Total NonPoor Poor Total Poverty status Total Poverty 71 29 100 82 18 100 58 42 100 04ii.iiiCll Nutrition couldnot buy enoughfood 42 69 50 41 68 45 44 70 55 Incomescannotcover current expenditures 59 86 67 59 88 64 59 84 70 Inpart, the gap between the objective and subjective consumption-poverty may be due to the use of different benchmarks in assessing current economic status. The objective consumption-poverty identify the share o f individuals who, given current prices and incomes, cannot cover minimal food and non-food needs. Incontrast, households asked to report their economic status may compare their current economic potential with some old - possibly idealized - benchmark, such as their purchasing power before 1997.For most households, only in2002 real consumption exceeded the 1997 level. E.International comparisons Inthis subsection, we present comparative estimates of the incidence ofpoverty inthe Europe and Central Asia region, against two international poverty lines (of $2 and respectively $4per capita per day). The most recent estimates o fpoverty, which covers a large number of countries inthe region for the same period are from 1996-99. The sharp increase in poverty over the last decade, although unprecedented outside the transition countnes, is a common feature o f this group that is mainly due to the fall in output and sharp increases ininequality. Froma regional perspective (ECA), Romania has moderate levels o fpoverty, lower thaninthe Commonwealth ofIndependent States (CIS) and SoutheastemEurope (SEE), buthigher thanin Central Europeancountries Figure 4.2)" l5 Here poverty is defined as the number o f people who are living on less than $2 or $4 a day, respectively, at the 1996 purchasing power parity o f the dollar. The poverty headcount i s computed for 1998 for Romania and for 1996-98 for the other countries in the region. Depending on the available data, the welfare indicator used to measure poverty was either income or consumption, and the coverage of the indicator, or its ratio to the National Accounts estimate varied between 60 percent and 120percent. 13 Figure 1. RegionalPerspectiveson Poverty shave ofpeople living on less than $2 and$4per d q in Europeand CentralAsia Region, 1996-98 Source: Making transition workfor everyone, WorldBank, 2000 F. Changes inInequality Compared with other countries in Central Table 6. Consumptionbased Gini coefficient in selected and Eastern Europe, Romania has low countries levels o f inequality (Table 6). In 2002, the ConsumptionbasedGini (per adult equivalent) consumption Gini coefficient index was 0.29. Bulgaria 0.27 Hungary 0.28 This outcome is the result three factors. Slovenia 0.28 First, the largest share in primary incomes Romania 0.29 comes from wages, income source that is Croatia 0.32 equally distributed in Romania compared Armenia 0.32 to neighboring countries. Second, self- Tajikistan 0.32 employment earnings - a big contributor to Latvia 0.34 inequality elsewhere - are less unevenly Poland 0.34 distributed in Romania, as they are Macedonia, FYR 0.34 dominated by farming and low- Georgia 0.37 productivity non-agricultural sectors such Moldova 0.41 as trade and services, and constitute a Kyrgyz Republic 0.42 buffer sector for those unabsorbed by the RussianFederation 0.47 formal sector rather than a reward for Source: entrepreneurship. Lastly, the extent o f Makingtransitionwork for everyone, World Bank2000, Table4.2 For Romania authors own calculations from HBS2002 - redistribution undertaken by the Romanian state i s substantial. The Government redistributes about 10% of GDP in social protection benefit, a system o f progressive transfers that reduce the inequality in secondary (post-transfer and net o f taxes) incomes compared to primary (gross) incomes. 14 Duringthe last eight years, successive government placed a strong emphasis on maintaining low level o f inequality. Low inequality was maintained during the period o f GDP decline, from 1996 to 2000, during which (per adult equivalent) consumption inequality went down from a Gini index of 0.31 in 1996 to 0.28 in2000. Figure 2. Despite 14 Years of Difficult Transition, Romania Succeededto Secure Low Inequality ConsumptionInequalityin Romania Gini Indexof Per Adult EquivalentConsumption 0.32 0'2 0.26 1995 1996 1997 1998year1999 2000 2001 2002 Source: Romania IHS 1995.2000. HBS 2001-2002 The resumption of economic growth since 2000 was accompanied a modest increase ininequality. The Gini index rose from 0.28 in 2000 to 0.29 in 2002. There is a growing consensus in the economic literature that low levels o f inequality is a precondition for sustainable growth, a favorable factor that the Romanian Government should bank on. 15 3. Poverty Dynamics: 1995-2002 Poverty dynamics i s the result o f two proximate causes: the GDP growth and the changes in the level o f inequality. Given the modest changes in inequality, changes in poverty responded mainly to changes inGDP. A. Poverty,growthand inequality The evolution o f poverty in Romania during 1995-2002 confirms the robust relationship between poverty and growth, the two indicators going in different directions (Table 6 and Figure 3). For any poverty line - food, extreme or total -the poverty numbers: declined in 1996, on the background o f a growing economy; registered a sharp increase in 1997, and modest increases thereafter up to 2000, during a four-year recession period; and began to fall again in 2001-02, with the retumo f the economic growth. The changes inpoverty were more pronounced for total poverty, compared to extreme or food poverty. Despite the recession that engulfed the country after 1996, extreme poverty changed very little. Inpart, this is due to (i) active social protection policy that aimed to mitigate the social costs associated an with the attempts to stabilize the macro economy and induce structural adjustment; and to (ii) the success o f the reforms initiated in 1997, that reduced the relative price o f food during the last years o f the `90s. The resumption o f growth in 2001 succeeded to make the first significant dent in total poverty since 1997. This re-emphasizes that growth is essential for sustained poverty reduction, and structural adjustment policies accompaniedby effective social safety nets may shield the extreme poor from deeper destitution while planting the seeds for a sustainable economic recovery. Table 7. Evolutionof ConsumptionPovertyandInequalityinRomania,1995-2002 Year Total I Extreme I Food /Inequality Headcount Gap Severity1 Headcount Gap Severity1 Headcount Gap Severity[ Gin 1996 0.2007 0.0477 0.0169 0.0627 0,0120 0.0037 0.0283 0.0050 0.0015 0.3081 1997 0.3026 0 0793 0.0307 0.1123 0.0246 0.0083 0.0576 0.0116 0.0038 0.2961 1998 0.3080 0.0797 0.0306 0.1132 0.0241 0.0082 0.0559 0.0113 0.0038 0.2931 1999 0.3320 0,0880 0.0346 0 1250 0.0283 0.0098 0.0674 0.0137 0 0045 0.2857 2000 0.3586 0.0957 0.0373 0.1379 0.0302 0,0101 0.0730 0.0142 0.0045 0.2803 2001 0.3057 0.0792 0.0306 0.1136 0.0247 0 0082 0.0581 0.0115 0.0036 0.2836 2002 0.2890 0.0759 0.0295 0.1091 0.0239 0.0061 0.0558 0.0114 0.0037 0.2876 B. Assessingthe Robustnessof ObservedChangesin Poverty over Time Are the changes in poverty numbers statistically significant? The poverty numbers presented in Table 7 are estimated based on a sample o f the population. Given the stochastic nature o f our estimation, it i s legitimate to ask if the observed changes in poverty levels are sufficiently large to represent true changes inpoverty, or may be simply due to sampling error. InRomania, such question i s more relevant than elsewhere, given the over-clustered design o f the HBS which tend to reduce the precision o f the survey in detecting changes in poverty. Figure 3 depicts the dynamics o f total and extreme poverty headcount with their corresponding 95% confidence intervals. It becomes apparent that insome years small changes inpoverty are hard to detect with sufficient precision. 16 From 1995 to 2002, only in three years were changes in poverty sufficiently large to be confident in their trend: 1996 (poverty fell), 1997 (poverty rose) and 2001 (poverty fell). One can identify significant changes in the level o f poverty in 1996 compared to 1995; or 1997 compared to 1996. After 1997, the changes in the poverty headcount are not significantly different from one year to the other, although total poverty is significantly higher in 2000 compared to 1997/98. The reduction in poverty headcount in 2001 compared to 2000 i s statistically significant, while the one registered in 2002/01 does not pass the usual tests o f statistical significance. Figure3. Significant ReductioninPoverty Were Achieved since 2001 Poverty Dynamics in Romania, 1995-2002 Poverty Headcount and 95% CI I I I , I , I 1995 1996 1997 1998 1999 2000 2001 2002 Max/Min -Total Poverty Headcount Max/Min - - ~ _Extreme Poverty Headcoun - Source: Romania IHS 1995-2000. HBS 2001-2002 Note: based on per adult equivalent consumption C. AssessingChangesinPovertyover TimewithoutPovertyLines In setting the total and extreme poverty line there is a certain subjective element, the most important being the choice o f the caloric requirement that would "anchor" the food component o f the line. This raises the question: How robust are the poverty rankings across time to the choice o f the poverty line? One way to make robust comparisons without choosing a particular poverty line i s by testing for stochastic dominance, i.e. by comparing poverty incidence curves (or cumulative density curves) that summarize the levels and the distribution o f income. A poverty incidence curve has the cumulative share of population on the vertical axis, and the real consumption per adult equivalent (or the welfare ratio or other welfare proxy) on the horizontal axis. The points along the horizontal axis can be considered as the complete set o fpoverty lines. The proportion o f the poor can be found by reading o f f the proportion o fthe population on the vertical axis whose welfare ratio is less than a given level. If, at any poverty line, one poverty incidence curve [l]is above the other [2], the amount o f poverty in population [13 will be always greater than in population [2] for any given poverty line. Iftwo poverty dominance curves intersect, then the comparisons will give different results for different ranges o f poverty lines. However, this was not the case for Romania inthe period 1995-2002. 17 Figure 4 Comparing Poverty without Poverty Lines: Poverty Incidence Curves Poverty Incidence Curves Romania, 1995-2002 7 EC9 3 -P 8 I I $Y a ac\! 0 0 0 1000 2000 3000 4000 5000 Consumption per Adult Equivalent,Thousands ROL Dec 2002 1995 -1996 1997 1998 -- ~ 1999 -2000 2001 2002 ~ The two vertical lines correspond to the extreme and total poverty lines Read on y axis the corresponding poverty headcount for any poverty line on x axis Figure4 compares poverty levels during1995-2002 without using aparticular poverty line. The graph ranks eight poverty incidence curves for 1995-2002 for all households consuming up to 5 million ROLs. This additional test confirms the earlier findings. At any poverty line, poverty is lower in 1996, then in 1995, 2002, and then 2001-1997-1998, 1999 and 2000. Thus, we can conclude that poverty, measured against any poverty line, fell in 1996, to increase in 1997 at higher levels than 1995. Poverty changedimperceptibly in 1998,to rise again in 1999 and, to a smaller extent, in 2000. A large reduction in poverty was achieved in 2001, followed by a smaller reduction in 2002. Figure 4 illustrates the magnitude of the changes in poverty measured against the total and extreme poverty lines used in the study. These changes were more pronounced for total poverty compared to extreme poverty. D. Poverty Dynamics and Vulnerability For the formulation of an effective poverty-reduction strategy, understanding if poverty i s transient or chronic in nature i s of outmost importance. Transient poverty signals the inability of households to smooth consumption across time, while chronic poverty signals that poverty i s due to low asset endowments and low returns on those assets. However, without panel data it is not possible to assess if observed poverty i s chronic (the same households are poor in all periods) or transient (some householdsare poor in some periods, but not in others). As the ABF does not observe the same householdsacross time, an analysis of poverty dynamics after 2000 was not possible. Instead, this section draws on an earlier analysis (Tesliuc, Pop, Tesliuc, 2001), which investigates the chronic and transient character of poverty during 1995-1997, based on a three- period panel of approx. 3,000 households from the AIG. The analysis presented in this section replicate the earlier analysis using the new welfare indicator -- per adult equivalent consumption -- and the poverty lines reported inthis paper. 18 Conceptually, there are two situations that may trigger poverty: an income shock o f sufficient strength to push a household into poverty, or low endowment with assets insufficient to generate enough income to escape poverty. Income shocks may impoverish households temporarily. In the poverty literature, such households are called the transient poor. They would escape poverty even without outside help, after a period that is proportionate with the fall in income caused by the income shock and the return o f the assets (including labor) they own. Inthis category one would include the unemployedwho, inperiod o f economic recession, loose their jobs. When economy recovers and employment increase, such individuals would re- enter the labor force and may escape poverty. Other households would no be able to escape poverty even when economy recovers, becausethe assets they own do not generate sufficient income to lift them over the poverty threshold. The meager volume o f assets such households' own have similar returns under boom period as under recession. Such households are called permanent poor. Typically, in this category are includedthe disabled, or poor elderly unable to work. The distinction among transient and permanent poor is not straightforward. Some authors (World Bank, 1997)consider as permanent poor those who, for some period o ftime, do not escape from the poverty pool. Other authors (Sen, 2000) consider as permanent poor individuals without the capacity to adjust and exit from the poverty pool, irrespective of the fact that such assumed capacity was exercised or not. The first classification is based on the dynamics o f poverty, and is relatively simple to measure. The second classification is conceptually better, butrequires a number of subjective assumptionsto allow quantification. We estimated both o f them, as they look at poverty from slightly different angles, bringing thus additional insights on the poverty processes. Poverty Dynamics During Transition. To investigate the dynamics o f the poverty in recent years, we used a subsample of households observed continuously in the M S from 1995 to 1997. Out o f roughly 32,000 households surveyed each year, we found - thanks to the rotating panel feature o f the data - about 3,000 households surveyed eachyear. We usedthe panel to quantifythe entry into and exit from poverty pool, and to test ifpast poverty is associatedwith current poverty. From all the individuals inthe panel, less thantwo thirds (56.1 percent) were not poor during the period (Table 8). The rest o fthe householdswere poor in at least one year. Table 8 Entry Into And Exit FromPoverty Pool, 1995-97 Poor in..? IndividualsBelongingto an HouseholdHeadedby ., in 1997: . 1995 1996 1997 Employee Self- Farmer Unemployed Pensioner Total employed NotPoorBetween1995-97 No No No 71.3 31.8 27.2 30.9 52.1 56.1 PermanentPoor Yes Yes Yes 3.3 22.7 28.4 13.6 10.1 9.6 TransientPoor No No Yes 8.2 9.1 11.6 23.6 10.7 10.2 Yes No No 6.8 7.6 10.4 5.5 10.1 8.5 Yes No Yes 4.3 15.2 10.0 10.0 7.8 6.9 AtypiealPoor No Yes No 1.6 1.5 3.6 2.9 2.3 No Yes Yes 2.1 9.1 4.8 12.7 2.6 3.2 Yes Yes No 2.3 3.0 4.0 3.6 3.6 3.1 100.0 100.0 100.0 100.0 100.0 100.0 Source. Own estimations based on AIG 1995-97 19 The households that were poor in at least one year were divided inthree groups. Firstare those who were poor throughout the period, fitting our first definition o f permanent poor. Second, are those who exited from poverty in good years (1996), but entered in bad years (say, 1997).This group is close to our definition o f transient poverty. The remaining group contains the exemptions, households that either fell into poverty when the economy went well, or exited from poverty inperiods of recession. We have calledthem atypical poorI6.InTable 9 we presentedthe structure o f the sample o f the "poor at least inone year", for the whole sample and diss-aggregatedbythe occupation o fthe householdhead in 1997. Judged by the dynamics o f at risk o f poverty, most o f it - 58.4 percent - is transient poverty. Surprisingly, permanent poverty is only 21.9 percent, very close to the atypical poor figure (21.4 percent). This aggregate dynamics may hide significant behavioral differences among various types o f households, grouped by the occupation o f the household head. Extremely low levels o fpermanent poverty are noted for the employee - headedhouseholds. Such households seem to be able to restoretheir consumption above the poverty level in one or two years after the income shock. In contrast, a large proportion o f the self-employed or farmer- headed households "at least once poor" bear their poverty stigma year after year. This exercise is o f immediate interest for social policy, because we identified the groups that were able to cope with the hardship of transition, and contrast these with he ones that fare worse intime. As expected, we found that poor households headed by employees or pensioners are more able to exit from poverty than the others, notably the unemployed-, farmer- or self-employed-headed households. Social policy-makers should pay more attentionto the latter category. As a feature o f the transition process, one would notice the relative stability of the proportion of "atypical poor" for all types o f households. Inour opinion, this is indicative of the transformations that occur inthe real economy -that affects primarymarket incomes such as wages and entrepreneurial income-, but also in the cashbenefit system. Table9 The Structureofthe Sampleof "Poor inat LeastOneYear", byDynamics,1995-97 IndividualsBelongingto anHouseholdHeadedby ... in 1997: Employee Self-employed Farmer Unemployed Pensioner Total PermanentPoor 11.5 33.3 39.0 19.7 21.1 21.9 TransientPoor 67.3 46.7 44.0 56.6 59.8 58.4 AtypicalPoor 21.2 20.0 17.0 23.7 19.1 19.7 Total 100.0 100.0 100.0 100.0 100.0 100.0 Source.World Bank Estimations, AIG 1995-1997 Various hypothesesmay be formulated to explain the high tumover into and from poverty pool. First, the shallow poverty in Romania may be responsible for this result. About 8 percent o f Romanians were 5 percent above or below the poverty line in 1997.Relatively small income shocks may change their welfare position a little, butthis little is sufficient to change their poverty stams. Second, the 1995-97 sub-period was the most dynamic period fi-om 1995 to 2002, with one period of growth (199516)andthe largest economic decline (1996/7). As mentioned already, the enterprise reforms impacted on the wages and the welfare o f some employee-headed households, as well as on those affected by the subsequent redundancies. The containment o f the aggregate demand reduced the profitability of many businesses, hurting for instance the self-employed. Changes in the parameters o f social programs, such as lack o f timely readjustment o f pensions, child allowances or social aid, produced another type of shocks. Although such hypothesis may be plausible, we caution the reader that they were not demonstrated, and their validation or rejection requires further research. l6Itworth to note that inthe case o funemployedthere is not an "atypical" situation, since most o fthem appear to becameunemployedin 1996/ 1997.This explains the large percentage of "atypical" poor insidethis category. 20 Table 9 HouseholdCapabilities to EscapePoverty: Transient versus Permanent Poor PovertyHeadcount(YO) Poverty Structure(%) Yew 1995 1996 1997 1998 1995 1996 1997 1998 TransientPoverty 71.4 72.0 75.5 76.0 PermanentPoverty 28.6 28.0 24.5 24.0 -head is over 60 years old 15.6 10.6 17.7 19.7 14.4 12.7 13.6 14.4 -- havingfour heador membercannot work 42.0 42.4 54.2 56.5 2.2 3.1 2.3 1.9 childrenormore 71.0 66.7 79.7 83.6 12.0 12.3 8.6 7.7 The Poor Rat Needs Our Help. The second classification o f poor into transient or permanent uses as criterion for determining the permanent poor the lack o f means to overcome poverty. Here, we classify as permanent poor the disabled, some o f the elderly and families with a large number o f children (four or above). According to this criterion, about 28 percent o f the poor were in permanent poverty in 1995 and 1996 (Table 9). As expected, in 1997 and 1998, years o f economic decline, this proportion dropped to 24 percent as an effect o f the large entry o f "transient poor". This defmition gives us higher estimates than the "dynamic" interpretation, but ina relatively smaller range.Inthe end, it carries the same message, that much o f the poverty inRomania is transient in nature. The positive messagethis analysis carries is that much of poverty faced by Romanians today is transient in nature. Transient poverty would shrink when growth resumes. This speaks about the prime role that macrostabilization and restoration o f sustainable growth should play ina poverty alleviation strategy inRomania. For the remaining poor, social programs that shield them from poverty should be implemented. 4. Revenues and consumption This section looks at the sources o f income and the destination o f consumption o f the Romanian households, in 2002 (ranked into quintiles according to their per adult-equivalent consumption, by area o f residence or poverty status) and in dynamics, from 1995-2002. The most important difference in the income structure o f the rich and poor households (classified by quintile; or by poverty status) was due to the size and share of the wage income in total household income. A better understanding o f the labor market is crucial for the formulation o f an effective poverty reduction strategy. 0 For the richest 20%, wages accounted 60% o f their income, compared to only 20% for the poorest 20%. 0 The poorest tend to depend more on subsistence agriculture, self-employment and other social protection programs than pensions. 0 The largest share o f pension income is captured by the people inthe third and fourth quintile. 0 By area of residence, rural residents earn on average substantially less wage income, and more agricultural income (especially from subsistence agriculture). They also depend more on other social protection programs than pensions. 21 Figure 5. IncomeStructurein 2002 Income Structure by Quintiles in 2002 Income structure in 2002 100% 90% Other inmme ~~~ ~ ~ SO% Otherinmme 70% mOthwsoaa1 pmteulon DGther~~aal pmleRiOn 60% ~SOcialinsurancepension DSOClal lns"ranc pension 50% 0Self-employment 0Self-employment 40% GSale of agn-focdr&lnertock 0Sale of agn-hDdS8l~ves10Ck 30% IIFocd SelFmnsUmption 0FoodSelf-mnsumption 20% ~ G m s ssalaw O G m ~ i i a l a ~ y '0% 0% national urban rural mnpoor poor extreme Q i 02 Q3 Q4 C5 Total poor Source: World Bank estimations based on AIG 1995-2000 and ABF 2001-2002 The average structure of the household income was relatively stable during 1995 to 2002. During the recession period (1997-2002), the share o f subsistence agriculture increased at the expense of wages. Figure 6. ChangesinIncome Structure, 1995-2002 income Structure Dynamics 1004 90% SO% 1 70% mother income 60% UOther social protection ESocial insurancepension 50% 0Self-employment 40% OSale of agnfoods&livestock 30% Food Self-consumption 20% 0Grosssalary 10% OX 1995 1996 1997 1998 1999 2000 2001 2002 I Source: WorldBank estimations based on AIG 1995-2000 and ABF 2001-2002 On average, the poor spend a larger fraction o f their income on food, consistent with the empirical regularity knows as the Engel's Law (Figure 7). The share o f both taxes and savings tend to increase with income. The fact that richer households bear a larger share o f the tax burden signals the progressive structure o f the tax system in Romania, built upon the concept of vertical equity. Together with the system o f public transfers, the tax system contributes to the reduction in consumption inequality. By area o f residence, rural households spend less o f services, bear a lower tax burden and have a higher saving rate. 22 Figure7. The StructureHouseholdConsumptionin2002 Expenditure structure by Quintiles in 2002 Expenditure Structure in 2002 lWld ma B4 Otherexpenditures 80% 70% 70% aSavings 60% SO% OTaxes and mmnbutions 50% EU% OPayforsewices 40% 40% 0Purchase of "on-bod 0Purchase ofnon-food 30% 30% OPurchase of food Purchase of food 20% 20% 10% 10% 0% 0% Qi Q.2 Q3 Q4 Q5 Total Source: WorldBank estimations based on AIG 1995-2000 and ABF 2001-2002 The average structure of the household consumption recorded a number of changes during 1995 to 2002: The share of food in the household budget continued to increase during 1995-2000, to moderately decrease thereafter. By source, purchases of food items registered higher changes compared to the share o f subsistence agriculture. The household tax burden (direct taxes on wages and social security contributions, agricultural tax, etc.) went down in 2000, an electoral year that coincided with the implementation of the global income tax. Figure8. Changesin Structure ofHouseholdConsumption,1995-2002 I Expenditure Structure Dynamics Other expenditures 70% 60% OTaxes and contributions 50% 40% Purchaseof non-food 30% Purchaseof food 20% Food Self-consumption 10% 0% 1995 1996 1997 1998 1999 2000 2001 2002 Source: WorldBank estimations based on AIG 1995-2000 and ABF 2001-20 23 5. A Profile of the Poor in2002 This section presents what individual, household or location characteristics are associated with poverty. Following the standard practice, we slice total and extreme poverty by characteristics such as: e household demographics: household size, the number o f children, age, gender and nationality; endowment with assets: education and occupation; and e location: area o f residence and region. The analysis reveals the groups at high riskof poverty groups where the incidence o fpoverty, or the - poverty headcount, i s above the national average - as well as the largest groups of poor. To simplify the presentation, we use simple graphics to illustrate the groups with highest risk o f poverty or largest number o f poor in 2002, and changes inthe risk o f being poor over the period 1995-2002 for a subset o f characteristics where significant changes were observed. The main text concentrates on total poverty; comparative information on extreme poverty i s annexed. As inanumber o f other studies that profiled poverty inRomania, we found that larger households face higher poverty r isks, and that education and e mployment in formal economy is a n effective shield against poverty. Unlike those studies, however, we detect higher poverty among elderly and female- headedhouseholds, and we signal the very high riskof poverty among the Roma minority. We analyze the strength o f the association between endowment with assets and poverty status, acknowledging the circularity o f the relationship between the two: having more assets now contribute to higher incomes and consumption tomorrow, while more income today generate more savings, investment and hence greater wealth tomorrow. By location, we found higher poverty in rural areas and in North-East, South-East and South, and acknowledge that our results are affected by consumption seasonality. The last part o fthis section uses a set o fpoverty regressions to disentangle the role that demographics, economic and location factors are playing the influencing household consumption. Given the large structural differences that exist between rural and urban areas, we estimate separate models by area o f residence and year. In each o f the models, expected consumption is the product o f a vector o f household resources (assets and characteristics) and their productivity. We conclude by pointing out the relative roles that changes inresources and changes inthe productivity o f these resources have had on the changes inhousehold consumption over the period 1995-2002. A. Groups at HighPovertyRisk Household Size and Poverty. Larger households have a higher poverty risk, even after accounting for economies o f scale inconsumption and the differential cost o f children compared to adults (Figure 9). This relationship is extremely robust to the choice of poverty line (holds for any poverty line) and year, Compared to earlier studies, we do not find a statistically significant difference in the risk o f being poor for families o f one, two or three persons. While larger families (five members and more) have a higher incidence o f total and extreme poverty in 2002 (49% total poverty headcount compared with 29% on average, respectively 23% extreme poverty headcount compared with 11% on average), they represent only 47% o f the total poor, respectively 57% o f the extreme poor. Thus, information on the household size alone i s a rather poor predictor o f poverty status, and its use for targeting transfers to the poor will entail large inclusion and exclusion errors. In fact, both pie graphs in Figure 9 show that poverty is a characteristic that permeates all types o f households. 24 Over 1995-2000, the relative risk o f poverty increased for families with four members, and remained largely unchanged for households with less than four members. Larger households, with five or more members, faced an almost constant higher risk o f poverty (71% higher than the average for total poverty, and 110% higher for extreme poverty) throughout the period. Figure 9. Larger householdsface higher risk of poverty Poverty Headcount by HouseholdSize Risk of Being Pwr (%) Shareamongthe Poor PopulationShares (Persons) NMe:Pave@ headmunlbaredon per adult equivalentmnsumpllon Source: 2002 HBS, Romania Number of Children and Poverty. About half of the Romanian households have children. Among families with children, 51% have one child, 35% have two children, and 14% three or more children. The risk of being poor increases with the number o f children, moderately up to two children but steeply thereafter. This difference is observed even after accounting for the lower cost o f children compared to adults. Although families with more than three children do not represent a large fraction o f the poor, they are a deep pocket o f poverty. Inparticular, two thirds o f the families with 3 or more children live inpoverty. Figure 10. Householdswith 3 or more Children face HighRiskof Poverty Poverty Headcount by Number of Children under 15 in the Household Risk of BeingPoor (%) Shareamongthe Poor PopulationShares (Persons) Note. PovertyheadmuntbaredOn peradult eWYBlenl mnsvmplion Source:2002 HBS. Romania Families with children pass a period during the live cycle when adults cannot devote their full labor endowment to lucrative activities, as they need to provide care for the children. Inthe last part o f the O OS, the opportunity cost o f raising children rose, as the cost o f childcare increased while the supply 25 o f such services fell. A larger number o f adults, especially mothers, are dropping out o f the labor force to nurse children. But the main response to the crisis was more family planning, and less children being born. Successive governments instituted policies providing incentives for families with children, driven by demographic and economic concerns. Working mothers benefit for up to two years matemity leave paid out o f the Social Insurance budget, if they choose to nurse their infants. Similar analyses determined the Romanian Government to institute a dual system o f family allowances, where a flat allowance per child i s given to all families and a supplementary allowance for the secondto fourth child. For families with one or two children, the risk o f poverty is further associated with mono-parenting. Families with one parent face higher poverty risk than families with two parents. Again, mono- parental families are one deep pocket o f poverty. They represent only 11% o fthe total number o f poor (or extreme poor), but face 30% to 50% higher risko fpoverty than the comparator households. Age and Poverty. Who ends up being poorer, the children, the elderly or the adults? By age, the highest risk of poverty is found among children, especially during the adolescent period (15-24 years old) (Figure 11). Indicator variables such as the age o f the household head, the size o f the household or the number o f children point, to a large extent, to the same group o f households. Figure 11.ChildrenFaceHigherRiskof Poveity Comparedto Adults or Elderly Poverty Headcount by Age of individuals Risk of Being Poor (%) Share among the Poor Population Shares Note:P w m hesdmrnl basedon per adult equivalentmnsumptim Source: 2002 HBS, Romania As compared to 1995,the relative risk o f poverty went down for the elderly, inpart due to the reforms o f the pension system (the recorrelationo f pensions implementedduring2000-02), while for children it increased throughout the period. 26 Table 10. Evolution of the Riskof Poverty by Age, 1995-2002 Statistic: Poverty headcount Age 1995 1996 1997 1998 1999 2000 2001 2002 0-6 years old 0.30 0.24 0.35 0.35 0.39 0.42 0.38 0.35 7-14 0.27 0.21 0.32 0.34 0.37 0.41 0.34 0.34 15-24 0.32 0.27 0.38 0.38 0.41 0.45 0.39 0.37 25-34 0.20 0.16 0.25 0.26 0.29 0.31 0.27 0.25 35-44 0.20 0.16 0.25 0.26 0.29 0.33 0.26 0.25 45-54 0.21 0.18 0.27 0.27 0.29 0.31 0.26 0.24 55-64 0.23 0.17 0.27 0.27 0.27 0.30 0.24 0.23 65 and over 0.31 0.22 0.34 0.34 0.35 0.35 0.32 0.29 Source: WorldBank estimations based on AIG 1995-2000 andABF 2001-2002 Gender and Poverty. Most recent poverty studies produced inRanania found that female, and female headed households inparticular, face lower risk o f poverty than men. To a large extend, the findings were driven by the assumptions embodies inthe particular methodology usedto measurepoverty. The cited studies used an adult equivalence scheme centered on the nutritional requirements by gender and age, where females would need, in average, only 80% o f the consumption o f the males to achieve a similar level ofwelfare. This assumption is hardlytenable, at least becausenutritional requirement are only one part o f individual, or householdwelfare. Moreover, a Rothbarth model o f the cost o f children relative to adults run separately for male- and female-headed households gave no significant differences inthe "cost" o fa male or a female compared to that o f a child. At individual level, there are no differences in the incidence o f poverty by gender, throughout the period. Female-headed household, however, face higher risk o fpoverty compared to males, due to the higher share o f mono-parental households and old widows living on low survivorship pensions that are found in this category. Overall, the share o f female headed-household in total or extreme poverty i s 21%. The relative risk o fpoverty between female and male-headed households dropped continuously during 1995-2002, with the steepest reduction occurring in2002. Figure 12. Female -Headed Households Face Higher Riskof Poverty Poverty Headcount by Gender of Household Head Risk of Beina Poor I%) Share among the Poor Population Shares (Persons) Note PweW headmun: based on peradult equivalent consump:ion Source: 2002 HBS, Romania 27 Female headedhouseholds facenot only a higher poverty risk,but also a lower capacity o f escapingpoverty (Table 11). The analysis o f the panel data presentedin section 3.Drevealed that the gender dimension is an important one when speakmg about the vulnerability: the risk o f a female headedhousehold o fnot escaping poverty is 1.3 higher than the average. The most vulnerable categories o f female headedhouseholds are the rural and the elderly. One could expect a large overlap o f these categories given the ageing process o f Romanian rural population. In most cases the elderly female headed households are single member households which are not able to work the land and are left without any form o f support. Inparticular, they may be excluded fiom the targeted social assistanceprograms (such as the MIG) due to the fact that they own land. Table 11.Permanentpoverty risk (YO)by area andgender Male Female Total Age group of household head 15-39 6.3 6.9 6.4 40-48 9.8 8.8 9.7 49-58 9.8 10.9 10.0 59-67 8.6 9.9 8.9 67 and over 9.9 17.4 12.9 By area of residence 9.6 Urban 3.9 6.2 4.4 Rural 13.3 18.4 14.4 Nuti0nu1 8.8 12.4 9.6 Source: World Bank estimations based on apanel sub-sample fiom theAIG 1995-1997 Romas, a Deep Poverty Pocket. Beside the majority population, the other three main ethnic minorities that are living in Romania are Hungarians (5.9%), Roma (2.5%) and Germans (0.5%). There are no significant differences in the average level o f welfare, total or extreme poverty between ethnic Hungarians or German and the majority population. In contrast, there are wide disparities in welfare between Romas and the rest. These disparities increased in 1996 compared to 1995, and decreased thereafter. By 2002, Romas were 2.7 more likely to be found among the poor than the rest o f the population, and 5 times more for the extreme poor. Romas account for 7% o f the total poor, and 12.5% o f the extreme poor. In fact, three out o f five Romas live inextreme poverty, and only one out o f five is n o t poor. H owever, the poverty problem is not a n ethnic one inRomania. T he 1argest number o f total (87%) and extreme (82%) poor are ethnic Romanians. 28 Figure 13. Rromas are a Deep Pocket of Poverty Poverty Headcount by Nationalityof the HouseholdHead Risk of BeingPoor(%) Shareamongthe Poor Population Shares Pow* hesdoovnl Index Note:Pow* headmunlbaJedon per adultequivalentmnrumption Source:2002 HBS, Romania Education and Poverty. Education is one o f the best proxies for the household's human capital. The risk of poverty d-ops substantially with better education, for any poverty line. One deep poverty pocket is represented by household whose head has no school, for which the changes o f being in poverty are near 67%. This group represents 7% o f total poor, or 10% o f the extreme poor. The majority o f the poor, however, are to be found in households whose head finished at most middle or vocational school. The relative risk o fpoverty for household whose head has no education was falling after 1998. Households headedby persons with vocational training fare worse from a year to the next, as they are most hit by the process o f industrial restructuring taking place in Romania. Household headed by individuals with highschool or university education registered a constant risk dpoverty throughout the period. Figure 14. Farmers and the Unemployed Face Higher Riskof Poverty Poverty Headcount by Occupational Status of the HH Head Risk of Being Poor (%) Share among the Poor Population Shares (Persons) Not@.Poverty headmunt based on peradullqu~v&ntc"mpticm Source: 2002 HBS,Romania Occupation of the Household Head and Poverty. Another characteristics associated with the household's human capital i s the occupation o f the household head. Occupation also reveals the activity status of the main breadwinner in the household (pensioners or others mean inactive), and participation in the formal or informal economy (Figure 14). Throughout the period, households 29 whose breadwinner derives income from the formal economy, such as employers or employee, had the lowest incidence o f poverty. Pensioners, especially those eaming a social security pension due to their past employment in the formal economy, come third in terms o f low risk o f poverty. Households headed by the unemployed, or whose head works in the informal economy - as farmer or self- employed in non-agricultural activities - face the highest poverty risk, between 50 and 61%. Such high risk of poverty among those engaged in informa1 activities suggest that the self-employment status if,for many, not a matter o f choice inthe pursuit o f an entrepreneurial idea, but an employment buffer offering low returns. The largest groups o fpoor, however, are employee- andpensioner-headed households. Together, these two groups account for 76% o f the poor. Within the group o f employees, households with a larger number o f dependents, with only one wage-eamer, or with low wages (from farming, trade or some services) are facing higher poverty risks. Within the group o f pensioners, households subsisting on survivor pension inurban areas, or pensioners with little land inrural areas are more often among the poor. During 1995-2002, the relative risk of employee-headed households rose during 19961999, and begins to fell after 2000 with the resumption o f economic growth. For pensioners and the unemployed, the relative risk o f poverty was falling constantly after 1996, reflecting the concern o f successive governments to minimize the social costs cf the economic transformation. Throughout the period, poverty among farmer-headed households' i s the biggest. The same results are obtained by comparing the poverty rate among individuals, grouped according to their occupation (Figures 15 and 16). Figure 15. Riskof Povertyby the OccupationalStatus ofthe Individual,2002 Poverty Headcount by OccupationalStatus of individuals Riskof Beina Poor (%) Shareamongthe Poor PopulationShares 30 Figure 16. Evolution of the Riskof Poverty by Occupational Status, 1995-2002 0.55 0.45 0.35 0.25 0 15 v "I I - - - 1995 1996 IS97 1998 1999 2000 2001 2W2 employee - 4 non-agriculturalself-employed - -- I*W_n_ agriculturalself-employed unemployed 4-pensioner pupil,student Poverty and Location. Both the incidence and the number o f poor is higher in rural than in urban areas (Figure 17a). Despite a constant trend toward convergence, in2002 the risk of poverty was still more than double inrural than inurbanareas. Figure 17a Poverty is Mostly Rural Poverty Headcount by Area of Residence Risk of Beina Poor (Oh) Share among the Poor Population Shares Nob: Poverty hmdmunt basedon persduHequlvaient mnsumptlon Source: 2002 HBS, Romania Regional dimensions ofpoverty. There are substantial differences inthe incidence ofpoverty by region (Figure 17b). The higher risk of poverty is inNorth-East, 47% higher than the national average (77% higher rate of extreme poverty as well). This region also hosts the larger number of poor (25% o f total poor, and 30% of extreme poor). Bucharest, the capital city, enjoys the lowest risk of poverty, about one third o f the national average. After 1996, regional disparities in total and extreme poverty attenuated slightly. 31 Figure 17b. The Highest Riskof Poverty i s in NortkEast, the Lowest inBucharest City Poverty Headcount by Region Risk of Being Poor (Oh) Share among the Poor Population Shares Norlh-Earl Not- Pow* hesdmuntbasedon p r .dUlt84U,YalBnlmnJumpti~n Source: 2002 HBS, Romania B. A ConditionalProfile of Poverty The first part o fthis section analyzed the individual andhousehold characteristics associated with the risk of being poor. However, the resulting poverty profile, while informative, does not provide a deeper understanding o f the contribution o f various factors to household poverty. For instance, we have seen that poverty is higher in rural areas, among farming households and low educated households. However, we cannot tell what portion o f the observed rural poverty i s due to low educational achievements, and how much is due to an occupation with meager earnings. Next, we analyze the determinants of (the logarithm of) consumption per capita in a multivariate regression setting. The analysis is useful, first, to verify the relative role o fvarious household characteristics and endowments in determining poverty status, and second, to assess the potential impact that policy- inducing changes inthese factors are likely to have on poverty, holding all other factors constant. It is important to note the limitations o fthis type of analysis at the outset. First,the analysis does not capture the dynamic impact o f certain causes o f poverty over time. For example, the impact cf changes ineconomic growth - most certainly a key determinant o f poverty - cannot be assessedusing a static, cross-section model. Second, the analysis is limited by the variables available at the individual and household level from the H I S or the HBS. Other factors- such as social conditions or physical conditions (e.g., variations inclimate or access to markets or social capital or participation in credit/ceoperative organizations) - can not be included due to a lack o f data at this level. Third, although theory holds that many o f the variables captured by the HIS or the HBS and included inthe analysis (occupation, education, physical assets, geographic location, household size, number o f children) do indeed contribute to (cause) poverty, the statistical relationships should be interpreted as correlates not as determinants, since in some instances, causality can run both ways. Finally, in linking the independent variables to the level of consumption per capita, we don't usually know the correct functional form which relates these variables to (the logarithm of) consumption per capita, and it is ingeneral relatively difficult to identify interactions between independent variables. For each year, we estimate three poverty regressions to provide a multivariate picture o f the determinants o f poverty, with the purpose o f checking the robustness of our results to alternative specifications. The models, in effect, represent a form o f a conditional poverty profile, where for example, one can examine the influence o f a location on household consumption, controlling for the education and the occupational status of,he household head. The first model i s estimated with the full 32 sample in each year. Two other models are run separately for the rural and urban areas, to tests if there are significant differences inthe retums to household or location characteristics (the coefficients o f the two regressions) across areas o f residence. All models include as predictors the households demographics (size, age and gender o f the household head, the share o f children), household assets (human capital proxied by the education and the occupation of the household head, and the share o f employed household members in total) and regional characteristics (region). The regressions use robust standard errors to correct for the observed heteroskedasticity, and the survey design information. We restrict our analysis to the period 1997-2002, as in earlier years (1995-96) some o f the predictors were not collected inthe survey. The results are presentedinTable A l - A 2 annexed. Based on the estimated coefficients, and in some cases the mean values o f the predictors, we estimated the elasticities o f real household consumption per adult equivalent to various household characteristics, as reported inTable 12 below. The analysis confirms many o f the insights revealed by the profiling poverty. Table 12. ExpectedChanges inPer Adult Equivalent ConsumptionGiven a Changein One DependentVariable Elasticity / Marginal Effect 1997 1998 1999 2000 2001 2002 Rural -0.037 -0.022 -0.011 0.010 -0.006 -0.014 Region South-East 0.081 0.086 0.141 0.074 0.090 0.053 South 0.152 0.110 0.118 0.092 0.094 0.043 South-West 0.107 0.127 0.159 0.127 0.120 0.069 West 0.175 0.142 0.202 0.130 0.130 0.095 North-West 0.181 0.157 0.178 0.102 0.123 0.123 Centre 0.146 0.133 0.156 0.152 0.140 0.134 Bucharest 0.188 0.192 0.236 0.168 0.184 0.169 Household Size -0.159 -0.136 -0.138 -0.146 -0.132 -0.135 FemaleHousehold Head -0.029 -0.023 -0.029 -0.011 -0.025 -0.052 Shareof Primary Income Earners in :he HH 0.138 0.141 0.119 0.120 0.132 0.137 HHwith a SingleParenl -0.023 -0.025 -0.071 -0.041 -0.047 -0.094 Disabled HHMember -0.055 -0.050 -0.086 -0.079 -0.066 -0.060 Nationality ofHH head Hungarian -0.037 -0.03 1 -0.0 13 -0.042 -0.030 -0.041 Rroma -0.235 -0.242 -0.212 -0.267 -0.171 -0.163 Other 0.010 -0.060 -0.093 -0.010 -0.015 0.064 Occupation of the Household Head Employer 0.437 0.342 0.349 0.309 0.327 0.515 NonFarm Self-employed -0.05 1 -0.113 -0.099 -0.089 -0.058 -0.103 Farm Self-employed -0.032 -0.028 -0.094 -0,050 -0.088 -0.083 Unemployed -0.092 -0.091 -0.096 4.100 -0.114 -0.157 Pensioner 0.045 0.033 0.012 0.013 0.016 0.014 Other -0.059 -0.116 -0.065 -0.070 -0.066 -0.071 Education of the Household Head No Formal School or Primary School (grades 1-4) -0.195 -0.189 -0.181 -0.174 -0.198 -0.184 Middle school, grades 5-8 -0.108 -0.091 -0.089 -0.090 -0.100 -0.079 Highschool 0.086 0.106 0.117 0.133 0.145 0.165 Post-secondaryor Foremen's school 0.200 0.213 0.215 0.212 0.252 0.252 Higher school, short and long term 0.455 0.502 0.506 0.521 0.503 0.542 Marital Status of HHHead LivingTogether -0.142 -0.160 -0.127 -0.149 -0.084 -0.065 Divorced/ Separated -0.064 -0.055 -0.023 -0.073 -0.052 0.036 Widowed 0.013 0.010 0.022 -0.004 0.024 0.050 Unmarried -0.068 -0.028 -0.007 -0.045 0.009 0.027 Average Age ofAdults 15-29 -0.071 -0.071 -0.077 -0.079 -0.085 -0.057 40-49 0.049 0.064 0.087 0.076 0.083 0.081 50-59 0.071 0.123 0.143 0.147 0.139 0.120 1 60-69 0.119 0.149 0.153 0.174 0.174 0.184 70andover -0.002 0.036 0.052 0.084 0.076 0.076 33 The robust and negative relationship between household size and poverty i s confirmed by the model. The average household size is 3.6 members in urban areas and 3.8 members inrural areas. Evaluated at the mean the associated elasticity o f consumption i s between -0.6 to -0.7. This means that an increase inthe household size from 4 to 5 members (with 25%) would reduce household consumption, everything else being constant, by 13-16% (reported in Table 12). The impact o f household size i s relatively constant throughout the period. The returns to education are as predicted by the theory (Figure 18). The coefficients o f all education dummy variables are increasing with the level o f education. One can see that the returns are non-linear, with higher premium eamed by those with post-high-school and university education. For instance, per adult equivalent consumption i s 18% higher for a household headedby a person with hgh-school education than a similar households headedby somebody with vocational occupation. For a similar households headed by a person withhigher-school education (universityor equivalent), the expected increaseinconsumption is almost 60%. Figure18 The Impactof Education and Occupationof the HouseholdHeadon Expected Consumption ) Formal School Occupahon o f the Farm Self-employed Household Head UonFarmSelf- . . - Unemployed 7 -0.300 -0.200 -0.100 0000 0.100 0.200 0.300 0.400 0.500 0.60a The relationship between occupation o f the household head and household consumption (per adult equivalent) gives a ranking o f the "average earning potential" o f various occupations, compared with the reference category "employee", keeping under control other household characteristics (Figure 18). Basically, there are two main differences. On one side, employer-headed household have a 54% hgher expected consumption per adult equivalent than employee-headed households, ceteris paribus. Pensioners and employees are equally well off. On the down side, households headed by farmers or unemployed would loose 10-20% compared with otherwise similar employee-headedhouseholds. Lack o f education has the highest negative effect on consumptiod welfare. T h i s effect i s almost constant over the years (as compared with vocational/ apprentice school) (Figure 19). The category with a constant increasing trend i s households headedby a person with high-school education. 34 Figure19The ImpactofEducationofthe HouseholdHeadon ExpectedConsumptionover Time 0 600 0 500 0 400 0 300 0 200 0 100 0 000 - -0 100 - - e . -0 200 L - - - - - +- --* I --· -0 300 +- 1997 1998 1999 2000 2001 2002 - No FormalSchoolor PnmarySchool(grades 1-4) -Middle school,grades 5-8 -&- - -Highschool X Post-secondaryorForemen'sschool +Higher school,short and longterm Poverty and household welfare have an ethnic dimension (Rroma), which hopefully tend to diminish in recent years. Between two otherwise similar households, one headedby a Gypsy would have a 22% lower expected consumption per adult equivalent than one headedby a Romanianduring 1997-2000, downto 17% lower since 2001 (Figure 20). Such relative improvement inthe status of the Rroma population could be attributed to the policy of the Government in this area (the creation of special committees to promote the interest o fthe Rroma atjudets level, the organization o f specialemployment fairs for the Rromas). Figure20. Changes inHouseholdWelfare Comparedto aReferenceHouseholdwith Similar Endowment 1997 1998 1999 2000 2001 2002 0 000 -% \ I -0 150 - A - - - A . # -0 200 j t L - + A -A. 0 . 4 -0 250 ' I 0 'k -0 300 I" -C-HHwithaSingleParent --Unemployed - A 4 HHHead15-29YearsOld --Rroma 35 Over time, there is decrease o f the negative effect o f age and ethnicity on expected consumption, but an increase o f the negative effect o f being unemployed or single parent (Figure 20). When controlling for primary income eamers, the discrepancies rural/ urban disappear, suggesting that the difference inpoverty is due to concentration o f households with lower economic potential inrural, compared to urban areas. Ceteris paribus (controlling for occupation and household size; such as the number o f children), younger families are worst off. As compared with the age group 30-49, elderly are well of, when controlling for occupation (pensioners) and HHsize (no. o f children). Figure 21. Changes inHouseholdWelfareby the AverageAge ofAdults and Other Characteristics,Comparedto a ReferenceHouseholdwith Similar Endowment Roma 70 and OW HouseholdSlrc 6049 HHwith I Single ~ ~ 4ge or e ~ ~ e Parent AdulU 50-59 (omlled 30-39) Disabled HH Member 4049 FemaleHousehold Head After education, ethnicity and unemployment are the next factors with the largest negative effects (Figure 21). Households headed by farmers and self-employed in non-agricultural activities have significantly lower consumption than households headed by employees. The share o f primary income eamers (employees, employers, and non farm self-employed) has a positive but relatively small elasticity. Inter-regional discrepancies exists more inthe case o f rural areas. Urban areas are quite homogenous. In2002 one can observe less regional disparities as compared withprevious years. 6. Conclusions andPolicyRecommendations The evolution o f poverty in Romania during the second part o f the `90s confirms the robust relationship between poverty and growth, the t w o indicators g oing in different directions. In some years, however, small changes inpoverty are hard to detect with sufficient precision using the current survey (the HIS and the HBS). The elasticity o f total poverty to GDP was found larger than for extreme poverty. Despite the recession that engulfed the country after 1996, or the recovery that started in 2000, the number o f extreme poor changed only modestly. Inpart, this i s due to an active social protection policy that aimed to mitigate the social costs associated with the attempts to stabilize the macro economy and induce structural adjustment. Equally, this result i s due to the success o f the reforms initiated in 1997 that reduced the relative price o f food during the last years o f the `90s. From this analysis, we learned that growth is essential for sustained poverty reduction, and that structural adjustment policies accompanied by effective s ocial safety nets m a y shield the extreme poor from deeper destitution while planting the seeds for a sustainable economic recovery. 36 Frommacroeconomic perspective, the most important factor o f the impoverishment o f Romanians was the economic recessions in early and late `90s. The fall in GDP over 1989-1993 and 1997-1999 was accompanied by a sharp increase in poverty. The growth registered during 1994-96 had reduced poverty. Compared with the GDP decline, the increase in inequality was only a marginal factor o f the process o f impoverishment. After 1997,the first significant dent inpoverty was made only in2001. To alleviate poverty, the implementation o f a set o f macroeconomic policies able to generate sustainable growth i s the key factor. Acknowledging the temporary character o f the poverty in Romania, this analysis stresses the importance o f macro-stabilization, structural and institutional policies for Romania. From a sectoral perspective, the growth inpoverty was associated with the process o f narrowing o f the salaried labor market, to the benefit o f other less onerous eaming opportunities in farming or non- agricultural self-employment, or unemployment. Somehow surprising, Romanian self-employed seems to have chosen this earning alternative not because it promises higher incomes compared to formal employment, but because they do not have access to the salaried wage market. The poverty rates among self-employed or farmers, economically active groups, are close to the rates o f unemployed. Such findings suggest the need o f a thorough investigation o f the labor market, as well as o f the other factor markets. Inthe same time, the findings support the emphasis placed by the current administration on stimulating the geographic and occupational mobility o f the labor force. The promotion o f labor policies aimed at increasing the labor or factors mobility, across economic sectors or from the state to the private sector, contributed to a reduction o f the eaming differential between formal and informal labor markets, and may increase labor productivity, especially in the "poor" sectors. What we have learned about composition o f poverty? By location, there i s higher poverty in rural areas and the North-East. Part o f poverty i s associated with life-cycle factors, such as raising children or gettingolder. Over the life-cycle, some families will cycle in and out o f poverty. Another part o fthe observed poverty, however, has structural roots, such as low education, unemployment or employment in low-retum informal activities. Better education, flexible labor markets and a business climate conducive to entrepreneurship are the rightpolicy responses to the second type o f poverty. We found that most poverty i s not concentrated among easily identifiable groups, a feature that would have made targeting efforts effective and efficient inthe fight against poverty. There are some poverty pockets, such as monoparental families, elderly widows, Romas, and large families (five members o f more) where the breadwinner i s unemployed, works in the informal economy, or lacks schooling. However, these pockets o f poverty account for only a fraction o f the total poor, and even the extreme poor. There are numerous other poor that are hard to detect based on a limited number o f individual or household characteristics. Poverty permeates the Romanian society, and the success in the fight for poverty reduction takes more than curing a limited number o f archetypal poor. Policies designed to address the pockets o fpoverty should be supplemented by an anti-poverty net o f 1ast resort. T he recent reforms o f the minimumincome guarantee program are steps inthe right direction. Education i s essential. Education emerges as the key correlate o f monetary poverty, as well as an indicator o f living standards inits own right. Inspite o f quite highaverage educational attainment o f the population as a whole, major discrepancies exist, particularly insecondary school achievement and above. 37 Bibliography Betti G., Maria Molnar and Filofteia Panduru (2002) "New Equivalence Scales for Romania", Working paper 39, University o f Sienna, Department o f Quantitative Methods Chen, S., and Ravallion, M (1996) "Data in Transition: Assessing Rural Living Standards in Southem China", China Economic Review, Vol. 7, no. 1 Chca, D.and Tesliuc, E.D.(eds), (1999), "From Poverty to RuralDevelopment", NationalCommission for Statistics Deaton, A. (1980) "The Measurement o f Welfare: Theory and Practical Guidelines", LSMS Working Paper 7, Washington D.C., World Bank Deaton, A (1987) "Quality, quantity and the spatial variation in price: estimating price elasticities from cross- sectional data", LSMS Working Paper 30, WashingtonD.C., World Bank Deaton, A and Case, A. (1987) "Analysis o f Household Expenditures", LSMS Working Paper 28, Washington D.C., World Bank Deaton, A (1988) "Quality, quantity and the spatial variation o f price", American Economic Review, 78,418-30 Deaton, A (1997) "The Analysis o f Household Survey: A Microeconometric Approach to Development Policy", The Johns Hopkins University Press Deaton and Paxson 1997, Poverty among Children and Elderly in Developing Countries, mimeo, Research Program inDevelopment Studies, PrincetonUniversity Deaton A,, and Salman Zaidi, 1999 "Guidelines for Constructing Consumption Aggregates for Welfare Analysis", mimeo, Research Program inDevelopment Studies, Princeton University Deaton A., and Grosh, M. (2000) "Consumption" in Margaret Grosh and Paul Gleewe (eds): "Designing Household Survey Questionnaires for Developing Countries: Lessons from 15 Years of Experience", Living Standard Measurement Study, Washington D.C., World Bank Deaton, A., and Tarozzi, (2000) "Prices and Poverty in India", mimeo, Research Program in Development Studies, PrincetonUniversity Dinculescu V. and Chirca, C. (coord.) (1999) "Saracia inRomania: dimensiuni si factori", in"Saracia inRomania 1995- 1998", PNUD, Proiectul de prevenire si combatere a saraciei. Foster, J., Greer, J., and Thorbecke, E. (1984) "A Class o f Decomposable Poverty Measures", Econometrica, 52 Hentschel, J. and Lanjouw., P. (1996) "Constructing and Indicator o f Consumption for the Analysis of Poverty", Working Paper 124, Washington D.C., World Bank Lanjouw, P and Ravallion, M(1994) "Poverty and Household Size", Economic Journal, 105, 1415-34 Lopez, R., and Thomas, T. (1999) "Romanian Integrated Household Survey", mimeo, World Bank Milanovic, B. (1998), "Income, Inequality, and Poverty during the Transition fiom Planned to Market Economy", World Bank,Regionaland Sectoral Studies MinisterulMuncii s i Protectiei Sociale (1998), "Raport privindactivitatea de asistenta sociala inanul2001" Nicholson J. Leonard (1976) "Appraisal o f different methods o f estimating equivalence scales and their results", Review of Income and Wealth, 22, 1-11. OECD, Labor Market and Social Protection Review, 1999 Pollak, Robert A, and Terence J. Wales (1979) "Welfare comparisons and equivalent scales", American Economic Review, 69 (papers and proceedings), 2 16-21. Ravallion, Martin (1994) "Poverty comparisons: a guide to concepts and methods", LSMS Working Paper 88, Washington D.C., World Bank 38 Ravallion, Martin (1998) "Poverty Lines inTheory and Practice", LSMS Working Paper 133, Washington D.C., World Bank Sen, Amartya, (2000) "Development as Freedom", Anchor Books, New York Subbarao, K.et al. (1997) "Social assistance and poverty targeted programs", World Bank Development Series Tesliuc C., L.Pop and E.Tesliuc, "Romania: Social Protection and The Poor" UNDP, Romania -National HumanDevelopment Report, Romania, various issues Wagner, P., C. Chirca , C. Zamfr, C. Molnar, and S. Parciog (eds.) (1998) "Metode s i tehnici de evaluare a saraciei", Ed. Expert Working, Holbrook (1943), "Statistical laws o f family expenditure", Journal o f the American Statistical Association, 38,43-56 World Bank (1997), "Romania: Poverty and Social Policy", Report no 16462-R0, World Bank World Bank (2002), "Romania: BuildingInstitutions for Public Expenditure Management. Reforms, Efficiency and Equity", Report no. 24756-RO. 39 Annex 1. Adult equivalencescale Equivalence scales are usedas crude corrections for the difference inneeds acrosshouseholds with different sizesand demographic composition^'^. They are short-cut methods that acknowledge, inan arbitraryway or endorsing a particular identifylngassumption, that larger families have bigger needs than smaller families, and that the needs of children may be different than those of adults. Other differences in needs between some socio-economic groups may be postulated, but would not be explored in t h s paper. There is no convincing, scientific methodto estimate economies of scale. Onthe contrary, the only establishedproof in the economic literature is the empirical underidentification ofeconomies of scale. In order to estimate such scale, the analystshave to make additionalassumptions. The adult equivalence scale used in thls paper i s based on two parameters: the cost o f children relative to adults and the economies o f scale enjoyed at household level. It uses the formula suggested by the U S National Research Council (NRC). The NRC defmed the number o f adult equivalents as: AJ! = (A+IXK)~, where A is the number o f adults in the household, K is the number o f chldren, a is the cost o f children relative to an adult and 8 i s the economies o f scale parameter. To estimate the cost o f children relative to adults, we will use the assumption attributed to Rothbarth, that households with the same level o f per adult expenditures on goods consumed only by adults are equally well-off. To estimate the economies o f scale parameter, we use the model suggested by Lanjouw and Ravallion (1994) based on the share o f privately and pure publicly consumed goods consumed by the household. There i s circularity in choosing the two parameters, andwe devised aniterative algorithm to find the common solution. Based on the distinction between public and private goods withn the household, Lanjouw and Ravallion (1994) suggest a model where the economies o f scale parameter depends on the share o f private goods in total c onsumption, and the household size. The goods c onsumed by the households are split into pure private goods, excludable inconsumption, andthe public, non-rival inconsumption, goods. Ifwe denoteby x total consumption, by pthe share ofprivate goods", bynthe family size, andby 8 the economies of scale parameter, then the monetary measureofthe average welfare is: x/ne=p*x/n +(l-p)*x. The average welfare is the weighted sum of two parts: (a) average consumption o f private goods, rival in consumption; and (b) total consumption of public goods, non-rival inc onsumption. S olving for 8, the equationbecomes: 8 =-In(1-p+p/n)/ln(n) We augment this specification to take into account the different needs o f children relative to adults. Thus, instead o fusing the family size n, we are using the effective family size, n,= A+aK, where a is determined usingthe Rothbarthmodel. The identification assumption inthe Rothbarth method is that two households with the same level o f adult goods per adult are equally well-off. The AIG is quite precise in identifylng the adult-only goods, versus child-only goods. The consumption o f adult goods includes components fi-om all the sections o f the questionnaire. Fromthe food section, we include alcoholic beverages and coffee. Fromthe non-food section, we include adult fabrics, clothes and shoes, as well as tobacco products. Fromthe section on services, we include shoes and clothes repairs. The sum of these components, or total household expenditures on adult As the AIG collects consumptioninformation only at householdlevel, we cannotlook at intra- householdallocation issues. Fromthe AIG commodity list includedinthe consumptionaggregate, the items that can be classified as publicly-consumed goods are fuels andthe payments for the services provided by utilities. In2000, the median share o f these items intotal consumptionwas 7%.Thus, p = 10%. 40 goods, would be the dependent variable. We counted as children all personsaged 14or less. We assumethat the consumption of adult-goods per adult depends on two main parameters, the real consumption and the demographics o f the household. Real consumption is equal to total consumption per adult equivalent. The demographic characteristics are captured by the household size, and the shares o f adults, elderly and children. To determine the parameter a, we estimate a Working-Leser (1943) equation, where the logarithm of expenditures on adult goodsper adult (AG) depends on the (logarithm of) consumption per adult equivalent, (logarithm o f the) household size, as well as the share of each demographic group (e.g. adults, chldren and elderly, as the omitted category) intotal. ln(AG)=a+p*ln(x/AE) +X*ln(n)+CF*Q/n +e where x/AE i s consumption per adult equivalent, n i s the household size, and ni/nrepresents the share o f adults, elderly andchldren. As these shares would sum up to 1, the equation would be estimatedwithout the term for elderly (the omitted demographc category). For a family with two adults, this reducesto: ln(AG) =a+ p*ln(d2*)+ x*ln(2) +6,*2/2+ 6,*0/2. For a family withtwo adults anda child, this reducesto: ln(AG) = a+ p*ln(~,/(2+a>^~>+~*ln(3)+6]*2/3+ &* 1/3 Using the Rothbarthidentifying assumption, the two families will have the same level o f welfare when the level o f consumption o f adult goods per adult is the same. Thus, we can estimate the ratio ~0 lxl, or the "cost" o fa family withtwo adults anda childrelative to a family withtwo adults only, as: xl /xo= exp[8*(l- X/p)*ln(l.5)+ 6,)/(3*p)] One can recover a from the ratio xo/xl. Say, the cost o f an adult is x, and the cost o f a child i s x, = a*x,. Then: xI /xo = 2xa/(2x,+x,) =2xa/(2x,+ax,) =2/(2+a), or a = 2*( x1/xo 1) - We start by estimating the economies o f scaleparameter 8,under the assumption that a = 1. Then, given 8, we determine the relative cost o f chldren a using Rothbarth model. Then, given a,we reestimate 8. With t h s 8, we reestimate a, and so on. We use an iterative algorithm, and a convergence criterion based on a minimumdifference inthe value of the parameter a of 1%. Our algorithm converges in5 iterations. The estimations made on the basis o f I U H S 1999 and 2000, using expansion factors, indicate that a child cost 50% of an adult, and an economies o f scaleparameter 8 = 0.90. The formula used to determine the number o fadult equivalents is thus: AE =(A+0.5C)A0.9 For each year, households are rankedby the level o freal household consumption per adult equivalent. This variable is usedindistributional analysis throughout the report. 41 StatisticalAppendix Table A1. Rural-to-UrbanPriceIndex Year January February March April May June July August September October November December 1995 0.980 0.967 0.955 0.961 0.950 0.966 0.965 0964 0.972 0.959 0.966 0.952 1996 0.986 0.916 0956 0.955 0.974 0.940 0.971 1.014 0.965 0.950 0.947 0.978 1997 0.953 0.919 0.971 0.971 0.926 0.948 0.933 0.917 0.956 0.941 0.946 0.939 1998 0.931 0.967 0.959 0.930 0.973 0.943 0.951 0 960 0.936 0.942 0.955 0.952 1999 0.944 0.957 0.962 0.938 0.913 0.922 0.958 0 936 0.937 0.935 0.922 0.928 2000 0.905 0.929 0.919 0 934 0.924 0.940 0.941 0 932 0.946 0.940 0.939 0.930 2001 0.926 0.930 0.945 0.932 0.939 0,921 0.923 0 923 0.910 0.930 0.928 0.933 2002 0,932 0,912 0.930 0.939 0.919 0.925 0,922 0 925 0,935 0,929 0.931 0.945 Total 0.945 0.937 0.949 0.945 0.940 0.938 0.946 0 946 0.945 0.941 0.942 0.945 Note: Dividing ruralhouseholdconsumptionby the index will bring it incomparableurbanprices. Source: World Bank estimationsbasedon the ABF for 2001-2002;AIG for 1995-2000 Urban Rural Quintile Poorest 2 3 4 Richest Poorest 2 3 4 Richest January 1.06 1.05 1.02 1.01 0.98 1.06 1.07 1.05 1.07 1.01 February 0.94 0.97 0.98 0.97 0.94 0.90 1.01 0.97 0.99 0.97 March 0.92 0.93 0.96 0.96 0.94 0.97 0.99 0.97 1.oo 0.99 April 0.99 1.01 1.02 1.03 1.02 1.03 1.03 1.02 1.04 1.01 May 0.92 0.93 0.95 0.96 0.99 0.95 0.94 0.98 0.96 0.97 June 0.92 0.92 0.93 0.93 0.93 0.94 0.92 0.92 0.92 0.95 July 0.95 0.94 0.93 0.93 0.96 0.92 0.91 0.95 0.93 0.97 August 0.97 0.96 0.96 0.98 0.99 0.96 0.95 0.95 0.94 1.oo September 1.03 1.03 1.02 0.99 1.05 1.03 0.99 1.oo 0.99 0.98 October 1.02 1.01 1.01 1.oo 1.05 1.02 1.oo 1.01 1.oo 0.99 1.02 November 1.03 1.04 1.01 1.01 1.01 1.oo 1.27 1.02 1.oo 1.oo December 1.23 1.19 1.20 1.I4 1.21 1.17 1.20 1.17 1.13 42 Table A3. FoodCompc I01 3571 00021 2625 20 5 55 00020 IO2 3313 O K 1 9 2632 50 5 02 00018 6 103 12000 0 9083 '53000 3206 35 0 8736 10483 I04 ll000 2 4673 351000 866" 16 23730 26103 I O 5 11667 II5918 2225 00 25791 83 I/1491 1,0072 I06 31800 0 0428 311800 i 3 3 38 00411 1308 I07 31761 0 1514 415000 443 37 0 1456 1625 I08 75000 00254 3203 00 81 49 0 0245 1835 109 22000 02491 313000 929 2% 0 2396 5272 110 16000 04820 351000 1691 96 0 4636 74l8 1 1 1 I5WO 00759 354000 268 81 0 0730 1096 112 32000 00037 3525 00 12 99 0 0035 113 1 I3 85000 0188C 101730 191 23 0 1808 15368 I14 87506 0 5500 2085 60 I147 I O 0 5290 a62S8 115 60000 0 liO9 1798 00 235 37 0 1259 7555 I16 51O"O 0 855, 1317 10 112621 0 8224 45232 117 45000 0 0972 1273 0" 123 78 0 0935 4209 118 60000 0 0071 859 95 6 5% 0 DO74 442 119 181000 0 0230 3175 00 72 88 0 0121 4129 120 83333 04359 3515 00 1532 18 04192 34937 121 70000 02802 2216 00 620 91 02695 18844 I22 72000 0 0222 2032 00 45 06 00213 1536 123 42500 03408 si4 oa 21144 03278 13932 124 40000 00109 1'5875 1587 00105 419 Canncd firh I25 170000 00031 1843 00 5 18 00030 513 MILKCHEESE AIDEGGS Wholc now's and bullalocou'i milk 126 10000 5 2887 780 00 412521 5 0867 50867 127 13000 0 2376 470 00 11166 02285 1970 128 8500 0 0659 900 00 59 29 00634 539 I29 120000 0 0082 4980 00 40 90 0 0079 948 130 30000 0 061% 503 00 32 09 00614 1841 131 16250 0 1755 545 00 95 Ed 0 1688 2743 132 50000 04116 2430 00 100008 a 3958 19712 133 70WO 0 20'4 3050 00 632 61 0 1995 13966 134 io000 0 2626 218000 572 56 0 2526 12630 135 500w 0 1231 2560 00 315 70 0 1186 5931 136 l20WO 0 0068 3000 00 2030 0 0061 781 137 117391 00166 319000 52 97 00160 1875 138 3200 13 9391 85 50 1191 79 13 4067 42901 139 12ow0 0 0077 8060 00 61 99 0 0074 888 140 37167 0 1734 7660 00 132824 0 1668 6198 141 158000 C 0006 8280 DO 4 96 u 0006 91 142 3" 0 9271 8156 00 7937 I 2 0 8922 3,126 la3 48333 0 0300 9270 00 21s 00 00288 '394 I44 38000 0 1321 9270 00 215547 0 2236 8498 I45 12000 0 7201 588 80 423 9% 0 6926 8311 146 16250 0 0586 567 60 33 24 0 0563 915 147 25000 0 0669 183 20 3901 0 0643 1608 148 ,5000 00213 463 50 9 86 0 0201 307 149 30000 0 0078 125 50 3 31 0 0075 225 150 12000 00401 61200 24 53 0 0386 463 151 30000 0 1312 799 00 104 84 0 1262 3786 152 12500 0 0089 542 40 4 82 0 0085 I07 153 ,000 06261 11520 72 I 2 0 6022 602 I54 30WO 00578 2514 00 146 98 0 0556 1667 I55 40000 0 0454 496 40 22 51 0 0436 1745 156 30WO 0 0676 293 70 19 85 0 0650 1950 157 32000 00191 201 50 3 97 00189 606 158 52000 00410 2230 80 91 31 0 0394 2048 I59 20000 0 OM2 601 20 2 53 0 0040 81 160 44500 0 0096 3315 40 32 os 0 0092 410 161 400W 0 3104 266 05 82 17 0 2985 11941 162 6000 0 5499 295 00 I6121 0 5288 3113 163 20000 00116 18600 2 I 5 00111 223 IM 34091 0 8047 212 50 195 I S 0 7740 26386 165 9500 0 1871 202 50 37 89 o 18on 1710 166 44W0 0 2478 229 95 56 99 0 23e4 10189 I67 31779 0 1556 30940 48 I 5 0 1497 1757 168 42231 0 1938 306 72 5943 0 I8M 7870 169 18000 04157 400 20 166 37 0 3998 7197 110 16500 00613 276 00 16 92 0 0590 973 171 54630 O O l N 3220C0 34 18 00102 558 172 13333 0 8903 423 C0 376 59 0 8563 11117 173 a0000 00599 1224 00 73 37 0 0577 2306 174 30000 0 6008 303000 IS20 27 0 5778 17334 175 8000 4 7704 722 40 3446 I 3 4 5882 36105 176 53333 0 0004 370000 I 4 7 0 0004 20 177 25000 I0759 18658 200 73 IO348 25869 178 39000 02831 725 00 205 28 0 2723 10621 Csnnsd wyciblcs 0 1636 492 00 83 iu 0 1631 5018 SUGAR. JAMS, HONEY, CHOCOLATEAND Sup, 0 7722 410000 316606 07427 i i i a 1 Canned fruit 8n '!'up 0 0384 633 60 24 32 0 0369 1431 0 1342 2200 00 295 13 0 1290 6136 183 38W0 0 0044 2750 00 12 17 0 0043 162 184 l0OWO 00130 3360 00 43 58 00125 I247 185 14OWO 00121 5554 00 67 43 00117 1635 I86 75000 00314 3995 00 125 32 0 0302 2263 187 100000 o no75 167000 I247 0 0072 718 188 25000 0 0281 412500 1!596 0 0270 676 189 66667 0 0453 44w 00 19951 0 0436 2907 196 l3OOO" 0 0075 4460 00 33 37 0 0072 935 198 ll000 0 5788 57000 329 91 0 5567 6124 199 46833 00103 264000 27 i s 0 0099 463 200 33333 0 2460 1400 00 344 35 0 2366 7886 201 66923 00578 210000 121 39 0 0556 3721 202 17500 I2138 705 00 85572 1 :674 10430 ... '"3 ..... lifihhl 0 ""7, ..... ..... ldhll"" 3 04 0 0020 214 204 20000 0 3471 390 00 I 3 5 36 03338 6676 J ~ C Ipzc month 80 864 Food p w Ihni: 812.005 ~ lmCI pse dS\ 2.651 lamcnr nct"r 10 2550 F d 0 962 43 Table A4. FoodCompc 101 4004 c 00l0 2625 20 2 57 0 0012 5 102 4000 00013 2632 50 3 48 0 0016 6 103 12ow 07185 3530 00 253628 0 8412 10167 Inr io000 176'0 351000 623227 2083' 20837 105 11667 8 9"RO 2225 00 1982022 10 5041 122548 106 31250 0 0426 311800 132 79 0 0502 I569 107 30000 0 1267 4250 00 s i a 5 8 0 1494 4483 108 60000 00181 3203 00 60 00 0 0221 1325 109 22000 0 2016 3130 00 763 30 02113 5309 110 160W 0 3680 3 5 i o o o 129161 0 1339 6943 111 15w0 0 0630 3540 00 223 07 0 0743 I l l 5 Other 8mA product 112 32000 0 0036 3525 00 1267 0 0012 136 MEAT B d 113 8onoo 0 1595 10i730 16224 o 1881 15045 I14 90000 U l l i l 2085 60 878 I S 0 4965 44685 115 60040 0 0899 1798 00 111 56 0 1060 6357 116 55000 0 6684 1317 10 88036 07882 43350 117 45714 0 0697 1273 00 o o a z 1757 i i a 8S 72 MOO0 0 0055 859 95 4 -4 0 0065 286 119 180WO 0 0208 317500 66 M 0 0245 4415 120 63000 03412 3515 00 119933 0 4023 33394 121 70WO 02388 221600 462 68 0 2462 ,7234 122 75w0 00187 2032 00 37 96 0 0220 1652 123 40625 02472 Sld 00 201 22 02915 11842 124 30wo 0 0068 1458 75 9 88 o ongo 240 125 170wO 0 0028 1843 00 5 20 0 0033 566 126 ,0020 4 1745 780 00 3256 06 4 9225 19225 127 12500 0 2232 470 00 10491 0 2632 3290 I28 85W 00518 9 w OD 46 61 00611 519 129 lllill 0 0063 498000 31 51 0 w75 829 130 26000 o osas 503 00 29 43 0 0690 1932 131 16OW 0 1349 545 00 73 51 0 1590 2545 132 500W 0 3010 243000 741 I5 03597 17983 133 700w 0 1606 301000 489 86 o 1894 13257 134 46250 o0 0987 2057 218000 448 37 0 2421 1121' 135 50000 2560 00 252 73 0 IIM 5B2I 136 I200W 0 0060 3ow00 18 08 oac71 853 137 88077 00139 319000 44 46 O O l M 1448 138 3150 106354 65 50 909 12 125411 39504 139 i2aooo 0 0054 806000 43 29 0 0063 '60 140 37333 0 1378 7660 00 I055 45 0 I625 6066 I41 I580"" 0 ow5 828000 3 83 0 0005 86 I42 350w 07146 855600 6113 10 0 8426 29491 143 48333 00221 927300 204 92 00261 1260 I44 38000 o 1681 927000 1 5 5 ~ 3 3 0 1982 7533 I45 ,2000 0 5997 586 80 353 12 0,072 a4a6 I46 15000 00559 56760 31 71 0 0659 989 147 27500 00406 583 20 23 68 0 0479 1316 148 20000 00172 463 50 7 95 0 0202 405 149 lo000 0 0056 425 50 2 39 0 0066 66 isn ,loo0 00312 61200 19 12 0 0368 442 151 30000 0 ,Ob2 799 00 84 82 0 1212 3755 152 12500 0 0095 542 40 5 16 00112 140 I53 1000 0 4049 I1520 46 64 047-4 177 I54 30000 0 G486 2544 00 I23 56 0 0573 1718 l 5 5 40ow 0 0482 496 40 23 92 0 0568 2273 156 3,WO 0 0609 293 70 1790 ooqia 2227 157 33333 00149 201 50 3 01 001'6 567 158 50000 00282 2230 ao 62 a8 0 0332 1662 159 10000 0 0032 601 20 I95 0 0038 38 I60 60000 DL081 334540 2698 0 0095 570 161 low0 02305 26605 61 32 02718 10872 162 5000 04315 295 00 12730 0 5088 2544 163 20000 0 0094 18600 I 7 5 00111 222 164 34091 0 6003 242 50 145 58 0 7079 24133 i l 5 9500 0 1352 202 50 27 38 0 1595 1515 I66 44W0 0 1910 229 95 43 92 0 2252 9911 167 31779 0 1216 30940 37 6 1 0 1434 4558 168 42231 0 I797 3 0 6 T 55 12 0 2119 8Y19 169 19896 03217 400 20 12873 0 3793 7547 1-0 2oow 00510 276 00 1407 0 0601 1202 171 50000 00085 3120 00 2738 00100 501 172 12667 06611 423 00 279 6d 0 7791 9874 173 40000 00459 1224 00 56 16 0 0541 2164 174 30000 04114 3030 00 136781 0 5323 15969 1'5 8000 3 6965 72240 267037 4 3589 34871 116 90000 00005 37w 00 I70 0 0005 49 177 24000 07883 18658 14709 0 9296 12310 178 40000 02146 725 00 I55 57 0 2530 ,0121 179 30000 0 1270 492 00 62 5 1 0 1498 4494 IFECTlONERY PRODUCTS 180 15000 0 5989 410000 245551 0 7362 10593 181 32458 00348 633 60 22 M 00110 1331 I82 47556 0 1121 2200 00 246 51 01321 6283 I83 38000 00036 2750 00 9 97 0 0013 I62 181 94000 O O l O l 3360 00 3431 0 0120 1132 I85 140000 00122 555400 67 69 0 011-1 2012 186 75000 0 0260 3995 00 103 77 0 0306 1297 Ia7 l00000 0 0074 1670 00 1233 0 0087 871 188 25000 0 0262 412500 10824 0 0309 774 I89 61314 0 0372 4400 00 163 55 0 0438 2687 196 140000 0 0064 4460 00 2s 62 0 0076 1059 198 11111 04'62 570 00 27I44 05615 6239 199 35667 0 0065 2640 00 17 15 0 0077 273 200 33333 0 1905 140000 266 73 0 2247 '189 201 72000 00441 210000 92 i d 0 0520 3741 0 9243 '05 00 651 64 10899 ,9074 00015 1460 00 2 25 00Ol8 194 0 3209 C18 820.171 44 o l r j n ~ 9 o e - - V I 0 0 0 0 0 0 0 c 1 0 W ~ 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 j g N v n o 0 0 X X g0 g0 0 01e 0 x 0 0 9 9 0 0 In * x n e n o 0 0 2 2 0 x m e 0 o m 2 2 0 0 ; - 0 1 c 2v, 22 2: m u 7Oc 9WmOo W. r-r - v 0 m e 0 0 N * -1 u o. e - c m a0gr gmsNgN g g - O 2: 0 0 0 0 0 0 0 e m m c l s g 0 0 3 E m - o r s": 0 0 - 1 9 2 s 0 0 0 0 0 1 0 9 % NV! 0 0 -n 0Z m0" 0 W - t w ~ 0 0 ~ ~ 0 0 0 0 0 0 0 m w m m c " - - v i o w - - - 9 9 9 9 9 0 0 0 0 0 0 vi Tnhlr A9 Currekites uCHouruhold Cun$unlptiuriper Adult Er&iivalenl 1991-2002 DepmdentVmnOle b p e r AdultEp~alcntConiumptlon, 1997 m a 1999 2000 2oor 2602 Hovrcbld SIZE 0 054 c 046 .o 048 0 051 0 o a t 0 047 (l653).* (1394)*- (12 Rvrrl -0 038 0 023 .I25 Fe&Houehold&ad 0 023 .I92 Q 14)' (2 lo)' @SI)** S b O f P ~ J n C O m c EvnrrrinuIIHH 045 0521 0 464 0 483 0 519 @ 529 R 3 Q213)** (2101)" Q367)*' (2299.. HHwiua SmglCPmnt 0 074 .O 042 -0 048 0 099 78 (509'. (509.. .o 09 .O 083 0 Ob2 . . Haad Employer 0 363 0 794 o a59 0 169 0 283 0 415 (I8s)" QV** QP,:;; N P GSclf.employcd ~ -0 052 .O 119 I 6 8 (449.. @ 9 9 * * (328)'. 83 (439.4 FarmSeLf-cmpbyrd 0 099 0 051 n092 .O 087 37 (49u)" (457)'* (39o)** U?employed -0 101 (2 2;; -0 I21 -0 17 (601)** (624)'. (75D" B'E)'* PanElO!le, 0012 0 013 16 0014 091 IOS 19 ow Othar (3%; %; -0 067 -0 073 68 0 073 143 posy. -1 $7 .I36 39 (2 00)" School(gradrr 1-4) -0217 0 11 0 2 0 191 0 22 -0 104 (1637.. (1592);. (1371)** (1223)** (1478)" (1359).* Mid& rrhcol gader5.8 -0 114 0 096 -0 094 -0 095 -0 I05 -0 082 @ 9 9 * * (839.. @%** (I69" Q5S)*. (639.. Highichml 0 083 0 101 0 111 0 I25 0 136 0 153 (720)** (1070)" (1079.. (ll74)** (1422)** (1387)** Q m % s e c m d qor Formen's school 0 183 0 193 0 191 0 192 0 225 0 225 (1357)" (1394)'. (I45Q'. (1265)" (160z)*' (16 05)" Higher school shod mdlmg term 0 375 0 407 0 409 0 419 0 a8 0 433 QdIXj*' ( 2 8 8 v * (2829** (2947)" a82S)** (31481** * O R South.Eas1 0 078 0 083 0 132 0 071 0 086 0 011 (274- (437" (25s)' 089" I 6 5 S O U POT4; 0 104 0 111 0 088 0 09 0 042 (488)** gq*. (387)** (294)- (213.' 146 SOU& West 0 102 0 12 0 147 0 11 0 113 0 067 (328)" (36s).' (417)" (3JZ)** (333)" .l82 Wt.1 0 161 0 133 0 184 0 I22 0 I12 ow (4210. (441).* ( 5 9 v . (438)" (431)'' @94)** Nor& West 0 157 0 146 0 164 0 097 a 115 0116 (539)'. (479)" (478)- (3;s); (409- (378)- Centra 0 136 0 115 0 I45 0 131 0 I I (441)1* (407)** (41s)** (44i)** (40v. (40?)** Bwhamt 0 172 0 175 0212 0 155 (1169 0 156 (5 17)" (512)** (659)" (546)** (573" (544)" M d m l A%tyofHH Hmd LnngTogethcr 0 153 0 174 0 136 0 162 0 087 0 067 (633)** Ql7)** (4 (633.. (431.. (349.. DirmcrdiScpratcd .O 066 -0 056 0 035 13 .1 11 0 022 (3y; @;T2; 0 075 -0 054 W3dowrd 0 049 .I49 -0 3 Unmmrd -0 008 -0 046 .o 35 (2 27)' Auamga Age ofAddh 15.29 .o 074 -0 074 .o 08 -0 082 -0 089 .o OS8 (77Gj** (8701.. (IE7)** (I5S)". (I9r)** (449** 4-49 0 048 0 062 C 084 0 073 08 C 078 (459" @@I)** @89** (I;?;;; @a?;;0 @ O Y ' 50.59 0 069 0 116 0 133 C 114 (J79** p4d)** (109T)** Q8T' Q99". (S64)** 60 69 0 139 0 143 0 161 0 161 0 169 @ g y * egz)-- @7s)** @ 3 g * * (929- 70 mnd D Y C ~ 0 @36 0 051 0 081 0 073 0 073 (215). (aSS)-. (405))" 04T)** (39"' consknt 14487 1 4 4 8 1439 14382 14425 I C 4 4 6 (4886@"* (50014~~. ( 4 3 8 7 5 p (416441" 08542)** (40085)'* Obrarvdmnr 32187 31014 31547 31089 31845 32074 R rquvad 0 37 0 38 0 4 0 4 044 0 4 6 Absolute value o f t stahtlcs mprcnthesrr 'S ~ @ I C m l l & , 5% ** ss&cmt d1930 O m t e d ~ l t e g o n Romamq Gmployrii,V~satl~naVApprlntr..SchooLMarrr3C ~ i 39YcusOld,L~ratidrnNorthEartRie)on 5 1 (I1 11)- (333)" -0 073 (409.. (239. 0 419 0 562 0 496 0 552 0445 0504 0 454 0 235 0 498 057t (15 16)** (1694.. 00oa)- (1489- (11 so- (13 261.. (I8 19- (1346)** (is 74)- (1286)" -0 046 -0 003 -0 03l -0016 012 .DO14 0 om 0 075 0 046 -0053 (204)' 0 12 I13 0 76 0.o11)" -03 092)'. 0-0074 ~ 3 -3 053 -0 C65 -0 061 .o 048 0 112 Q-033'. 009 -0 098 -0 072 @.X 02s). 0 q** P6l)** 054' (5 33)"' 0 59-* (4 33.. 0031.. 06 1 y 081)'. -0 014 0 062 o oza o 047 0 004 -0 049 0 056 0 027 .ooia -0 c53 -0 032 -0048 -1 03 48 .I11 I09 .O 16 .I27 I 9 7 o a5 62 .I39 I 0 8 II1 -0248 .o.I19c -0 146 031 -0 144 0313 0 284 0321 o-0028 -0 218 -0076 -0224 030.' (441)** (463" (6 06)" 023)' (529)'' 001)'. ( 4 9 9 " -0 34 (5in** -075 (383)". 0 089 0 099 0018 0 I51 0 028 0 174 0 02 0 037 0 038 -0 033 0054 0086 .I57 -0 a5 .I72 .o 95 046 071 -0 67 .O 67 -I95 -15 0351 0382 0185 0317 0 x 6 0349 0 221 0 463 0 259 (439" l739)" (749)" (443.. (55q" (471)- (4Jq- (509** (3 46)" 0019 -0112 61 0176 0057 -0159 .o 046 0 142 0 074 043 POT* a5 (399.i I 8 1 0056 -0011 61 -0091 n113018 0 204 -0 16 -0077 053 048 084 75 p93)- -0 25 05v 089- -0 119 -0 066 -0 108 91 -0099 -0 124 0-009.'145 -0 197 -0 139 ( 4 l0001 a " @;'a,; (449.. 003)** (419)" (356)** (576)'. 087)- (7 10)'. (86v. (539.. OOCd 0056 401 0019 0 003 0 021 0 001 0 0 0 0013 7 4 -0 06 -03 -05 I n-1 057 0019 0 0 3 0128 I -066 -0 48 -066 qll).' Schml(grades 1-4) .o 199 -0253 -0 181 -0139 0 194 -0 208 .o 18 .0219 -0 183 .o 13 .om Q 9 9 " " (141zJ." (882)'' (1380]** (848)** (IO 43)'. P 13" (1010)** c,23" (JOT)** (1148'' Middle mhmSgader 5 8 0 I13 0135 0118 0101 0119 0 081 0 094 0 113 0 093 -0056 .O I08 (621)** (767'. (slZ)** (619.. (801)" (5 431' (539.. (6 W"' (6 04)'* 0 2 v * (569** Highrchocl 0 1 0033 0 I2 0047 0 118 0 051 C 15 0 061 (13 19..148 0 016 0151 (lO1)** -176 (1061)** (284)" (IO5l)** 0 37)'' (1221).. 04s)" (1307)'" (71y* 0214 0139 0199 0212 0 12 o 098 0 238 024 0199 (1437))'. 017)**1 ( 1 4 2 v * (613)" (1343.' 0Za)**I (I5 39- (1583)" (599'' 0444 0 136 0 428 0453 0315 @926)** (533J** 0 7 5 3 * - p03.7** (11 11)" 0 097 0011 015 0108 0 053 0 05 0 084 OM2 0023 003086- 9 -164 046)'. (269.. .I36 001031" 9 -1 61 -045 0115 ooaz 0119 P0211' 071 o0 w 091 0 0066 p q * * 019j** 010). o w * I63 -0 96 0 166 -om P2: 0101 oow (z021). 084 0113 0174 0 133 0 101 0 085 0111 -191 000)*" .Ia7 -013 (213* 0 138 0;:;; 0;;;; P z i ; 0 154 0013). 015 0 098 0053 o i a 021)- (234y" 063" 044). (459" 02T' (438)** .II P 94)"" -I65 (200)' 0172 0166 0 161 0134 0 197 0 131 0 14 0 054 0 Ill 0094 0129 (460.. 041)'. (429.. (283** (441)*" 054)" 06l)** 132 0 PSI- 1117)' 090)" 0 IO5 0 195 0 099 0168 0118 0201 0 136 0 16 o 085 OW5 0162 (759** (445)** 033). (414)** 04.5)' (414)'' P19** (3 31)'. 1 9 1 039* PW" 0169 01851 0 176 0108 0211 0153 0 175 0 06 0 136 0126 0203 (444)" I 9 2 (461)** -088 (569.. 172 (522)** 0 16 093)" (379.' poq.. .O I25 4179 .O 143 .O 199 -0 138 -0 128 -0 115 0 183 0 055 0011 .O 136 (393)" (533.' (493)" (653'* (308)" (3lU)** 023.' (5 30)" 0 13)' -042 (5 58)" -0019 C O S .O 04 -0107 0012 0084 0 096 .O 049 .O 046 001 006 .in .. .I71 (1 99. -035 -087 WIdowsd 0 028 01 0001 0041 13 0011 0 os9 0032 0061 -I43 07 006 (20s). 13 .O 63 043. 4 6 033).* 0 -0116 0036 0117 0 036 .o I 0019 0041 .a042 -0 02 (256). -132 011)'' 139 0 21- -0 64 -IS1 -I21 0 043 0 062 0 103 0 085 (553" (47.7'. (5 9 q * * (5 96)- 0 039 0 048 0 1 0 064 0 904091 - (409.. (7 03)- (5 46)" 0 127 0 I51 o i i a (543.. (631)'. P009178 - (5 3s)" 0 129 0 152 0 149 (5 981.. (6 26))'. ( 7 2 y ($27)- 0 026 0 I07 0 315 0 051 I05 (375))" 90)" -1 63 14495 14422 P14313 14504 016 47)'. (32081)** (310 13)'. 063 657.. 15461 14621 15468 15051 0 38 0 41 0 3 0 41 ~EaetRegor 52 m 10 U s a a i Poverty and growth inRomania: 1995-2002* Roberta Gatti INDEX Executive summary 1.Introduction.................................................................................... ..57 1.I Methodological note................................................................ 57 2. Decline and growth in Romania: macroeconomic context and growth patterns for the period 1995-2002.. ......................................................... ..57 3. Poverty and Growth........................................................................... 60 3.1 How much did growth contribute topoverty changes? Resultsfrom thegrowth-redistribution decomposition.................................. 61 3.2 The distribution of gainsfrom growth: growth incidence curves...............66 3.3 Arepoverty changes due to intra sectoral (regional)poverty reduction or topopulation shift? Evidencefrom sectoral and regional decomposition of poverty................................................................................. .71 4. Sources of growth ............................................................................. 75 5. Poverty projections........................................................................... .76 6. Conclusions.................................................................................... ..78 Appendix. ......................................................................................... ..80 * Ithank Nicholas Bumett, Catalin Pauna, Peter Lanjouw, Cem Mete, Lucian Pop: Emil Tesliuc and, in particular, Michael Lokshm for useful conversations. Denis Nikitinprovided outstanding assistance. 55 Executive summary Inthe past seven years the Romanian economy was characterizedby high-income volatility. The country experienced a deep recession between 1997 and 2000, marked by high fiscal and quasi- fiscal deficits that were monetized at the price o f high inflation. However, the past t w o y ears recorded an exceptional recovery, which was the result o f a combination o f factors, including an election-driven expansion indemand, an increase in exports and investment, and disinflation. Over the whole period, growth (decline) rates inthe economy were highly differentiated spatially both across rural and urban areas and across the eight large regions. Urban areas were marginally more hit by the recession throughout the period 1997-2000, but benefited substantially more than rural areas from the recent growth in 2001 and 2002. As the poverty i s more prevalent in rural areas, this trend contributed to perpetuate poverty. The analysis o f the relative contribution o f growth and inequality changes on poverty changes highlights that growth has a strong impact on poverty dynamics. However, movements in redistribution have also been important in determining changes in overall poverty. In the period 1995-2000 (mostly recession years), poverty increased and would have done so even more substantially had the distribution o f income not moved favorably for the poor. Conversely, poverty decreased between 2000 and 2002. However, in these last two years income distribution moved unfavorably for the poor. In particular, in 2000-01, this pattem was driven by the distribution o f growth inrural areas. The occupational structure of the population i s an important dimension along which to assess the impact o f growth on poverty changes. Between 1995 and 2000, people moved progressively to occupational categories that acted as buffer against the recession and were characterized by increasing poverty. These population shifts accounted for a substantial share of poverty changes during the period. Conversely, population shifts across occupations played a very limited role for poverty changes in2000-2002. Although poverty differentials across regions are substantial, we observe very little, if none, net relocation of the population across regions or across urban and rural areas. Although the economic recovery in the past two years has been notable, the extent to which growth i s sustainable in the medium and long run i s a relevant issue. There i s consensus that macroeconomic stability and fiscal discipline are crucial to this end. A t the same time, reforms such as improving governance and the investment climate and revitalizing rural economies can address structural growth determinants and, especially in the case o f rural development, lay the ground for a path o f equitable growth. Projections under different growth scenarios suggest that, provided growth persists, poverty reduction can be achieved consistently. 56 1. Introduction Romania's economic growth recovered in2000 after three years o f consistent recession that were marked by sustained high inflation and macroeconomic instability. In this period, poverty dynamics tracked income quite closely. Inlight of these developments and of the commitment of the current government to poverty reduction, an in-depth analysis o f the role o f growth in poverty reduction in Romania i s particularly relevant. Inorder to provideinsights into these issues, thispaper presents anumberofstatistical exercises based on the rich set o f unit-level data from recent Romanian household surveys (Romanian Integrated Household Survey - R I H S - and Household Budget Survey -HBS). The work i s organized as follows. Firstwe describe dynamics and spatial pattems o f growth. We then quantify and discuss the relative c ontributions o f growth and changes ininequality t o poverty changes (growth-redistribution decomposition) and describe the pattems o f growth across different strata o f the population so as to identify the extent to which growth was pro-poor (growth incidence curves). Since Romania has witnessed over time a substantial shift o f the population across occupation, we also quantify the extent to which intra-sectoral changes inpoverty and population shifts across sectors contributed to overall changes in poverty (sectoral decomposition). In this context we examine how some potentially important sources o f growth for the Romanian economy c ould affect poverty outcomes. Finally, w e discuss poverty predictions for the years 2003-2007based on different assumptions on growth scenarios. 1.1 Methodological note For consistency with existing work on poverty in Romania (see for example Tesliuc et al. 2001) and with other sections o f the report for which this paper i s background work, all o f the results discussed in the following sections are based on a measure o f consumption per adult equivalent. As robustness checks, the analysis was also applied to consumption per capita. The related results are not reported, but divergences in findings are noted. The appendix briefly discusses o f advantages and limitations o f the consumption per adult equivalent measure. For simplicity o f exposition, emphasis is given throughout the report to headcount measures o f poverty for extreme and total poverty lines. However, substantial differences in results based on poverty gap or poverty severity measures are noted. 2. Decline and growth in Romania: macroeconomic context and growth patterns for the period 1995-2002 Throughout the late nineties, the Romanian economy was characterized by cycles o f booms and busts that were the result, amongst other things, of a slow and erratic implementationof reforms (see figure 1). 57 Figure 1 I Real GDP GrowthRates: 1996-2002 6 4 2 0 -2 -4 -6 Source: WorldBank estimates. The expansion o f 1996 - an electoral year - was followed by a substantial output collapse that lasted until the end o f the decade. During that period the restructuring o f many o f the large state owned enterprises (SOEs) proceeded very slowly. SOEs generated substantial losses, which resulted inthe accumulation o f substantial quasi-fiscal deficits. In turn, monetization o f these deficits generated an inflation rate among the highest in transition countries. On the other hand, poor fiscal management combined with controls on bank deposit rates contributed to a substantial credit crunch which brought investment to a halt. Overall, growth performance was disappointing (GDP in 2001 had not yet regained its 1990 value inreal terms), partly due to the slow pace at which structural reforms proceeded inthe country, to macroeconomic instability, and to an unfavorable investment climate characterized by widespread corruption (see discussion later). However, over the last two years, the economy has experienced an exceptional combination o f disinflation and strong growth. In the second half o f 2000, an election-driven expansion in domestic demand and a depreciation-led increase in export (mainly re-export o f processed inputs from EU countries) boosted growth to among the highest levels in the region. Growth was then sustained by an increase ininvestment which benefited from an expansion o f credit. Inflation was also brought down from 36% inmid-2001 to 19% inlate 2002. The dynamics o f income over time i s not the only interesting dimension along which to study growth, as growth rates in the country differ substantially along spatial and occupational dimensions. For example, the consumption decline was relatively less pronounced in rural areas. At the same time, subsequent growth was substantially lower than average in rural areas, suggesting that the rural economy i s to some extent isolated from economic volatility. However, a simple comparison o f relative differences o f growth and decline in urban and rural areas would indicate that rural areas participated less in the growth recovery than they were shielded from recession (figure 2). 58 Figure 2 Growth inmean consumption per adult equivalent s 1996 1997 1998 1999 20w 2wi 2002 Year Durban DRural .Total I Source: Household Budget Survey When disaggregating growth rates at the regional level, the West and South-West regions experienced the strongest overall decline in output. Figure 3 plots cumulative growth rates at the regional level for the sub-periods 1995-2000 and 2000-2002, where regions are sorted by increasing levels o f average consumption in 1995.' Although it is not easy to identify trends with such a limited number of observations, the decline in consumption appeared more pronounced in the regions with lower poverty rates - the West and Bucharest. In the following two years o f growth the substantially lower than average growth inthe South-West i s notable.' A simple growth regression can summarize growth trends across regions over the whole period. When regressing regional growth on time dummies (not reported), average consumption (C), share o f rural population (RURAL) and percentage o f unemployed population (UNEMPLOYED), the results suggest that growth (recession) was higher (lower) inregions with initial lower average consumption (thereby indicating that a standard converge effect was at work) and with a lower share o f rural and unemployedpopulation. Results are reported below (t- statistics inpar en these^).^ Growth~,=90.2-0.0000285*C~,-43 .80*RUFL4Lli,-1 37.6*UNEMPLOYEDit (-5.30) (-5.08) (-2.50) ''Note that the ranking o f regions according to average consumption levels in 1995 and 2000 was the same. A number o fpossible explanations are consistent with this finding.:the ceased severance payments to miners, some convergence effect (the South -West started with relatively lower poverty in 2000), the influence of the embargo on the neighboring Serbia and Montenegro. 3Thecoefficients are estimated with OLS on a small panel o f growth for the sub-periods 1995-97; 1997-00, 2000-02 for the eight macro regions. Standard errors are corrected for clustering atthe regional level. I indices regions and t time. 59 Figure 3 Growth rates in mean consumption per adult I I equivalent I -10% __ . ... . -20% -30% +2000-2002 . ..... . __ . Regionsare sorted by increasing levels of average consumption per adult equivalent Source: HBS. 3. Poverty and Growth The extent to which economic growth contributes to poverty reduction has been the object o f increasing interest in the policy debate. A standard view is that continued economic growth is associated with reductions in poverty rates. Measures o f elasticity of poverty to growth are often quoted as prima facie evidence of the impact to growth on poverty. The evidence for Romania suggests that elasticity o f poverty to growth has been substantial in the past (1996 and 1997) and has decreased since then almost uniformly for both extreme and total poverty. The elasticity o f poverty to growth has been around two (in absolute value) for 1999-2001, a value which i s inline with the impact of growth on poverty in other countries, but decreased somehow in 2002 (see figure 5). 60 Figure 5 Elasticity of poverty (headcount) to growth I 0 2001 2002 -1 - -3 -* 1i w 1+Extreme poverty ,I L--, +Total poverty ' -5 I -6 '1- Source: HBS, author's calculations There are limitations on relying on the calculated measures o f growth elasticity to understand the impact o f growth on poverty. A measure o f elasticity o f poverty to growth captures only the sensitivity o f measured poverty to changes in mean income while the relevant growth rate for poverty reduction i s the growth for poor and, in particular for the headcount ratio, the growth occurring in the neighborhood of the poverty 1ine. Inthis c ontext, it is important to note that certain pattems o f growth might be associated with increased inequality that can offset inpart or even completely the impact o f growth on poverty (El-laithy et al. 2003). Two distinct but related methodologies can address in more detail the impact of growth on poverty. First, the growth-redistribution decomposition introduced by Ravallion and Datt (1992) allows us to quantify the relative importance o f average income growth versus changes inincome distribution for poverty changes. Second, the construction o f growth incidence curves gives a visual representation o f the extent to which economic growth has reached different strata o f the population, including the poor and very poor. These methodologies and the respective results for Romania are described insections 3.1 and 3.2. Section 3.3 discusses results from a decomposition of poverty across occupations where we quantify the relative within- and across- occupation contributions to overall poverty changes. We perform a similar exercise for regions and across urban and rural areas. 3.1 How much did growth contribute to poverty changes? Results from the growth- redistribution decomposition With the sole exception of 1996 - an electoral year - the period between 1995 and 2000 was characterized by a long and persistent recession. 2000 - also an electoral year- marked the end o f the recession with a spectacular 8.4% growth recorded in per-adult equivalent consumption. Figure 6 depicts annual growth rates inconsumption per adult equivalent over the period. Poverty dynamics mirrored these trends quite closely. Poverty, however measured, started increasing in 1996, peaked in 2000, and then decreased (see detailed discussion in Pop and Tesliuc, 2003). Because o f the apparent structural break in growth trends in 2000, it is useful to discuss the relative impact o f growth and redistribution on poverty changes for the sub-periods 1995-2000 and 2000-2002 separately. 61 Figure 6 Growth in mean consumption per adult equivalent I 10.0% 5 0% 0.0% -5.0% -10 0% -15.0% Source: HBS, authors calculations According to the methodology developed by Ravallion and Datt (1992), a change in poverty between two dates can be decomposed into three components: (i) the change inpoverty due to a change in mean income while holding income distribution constant - the growth component; (ii) the change inpoverty due to a change inincome distribution while holding mean income constant - the redistribution component, and (iii)residual capturing the interaction between the growth a and redistribution component. Tables 1 and 2 report the results o f such decomposition for yearly changes inpoverty as well as for the two sub-periods 1995-2000 and 2000-2002. We computed the decomposition for the headcount measure of e xtreme and t otal poverty f o r the country a s a whole and for rural and urban areas. Inthe period 1995-2000 (and especially in 1996-2000), output decline hada substantial impact on poverty increase. However, income distribution moved in favor o f the poor reflecting a compression o f income distribution that i s often observed during recessions, thereby mitigating the adverse effects o f the output decline. 62 The growth-redistribution decomposition methodology (Ravallion and Datt 1992) Consider a poverty measure Pt that can be fully characterized interms of the poverty line (z), the mean income o f the distribution (pt)and the Lorenz curve representing the structure o frelative income inequalities (Lt.).Inthis context, the level o f poverty can change due to a change inmean income relative to the poverty line, or to a change in relative inequalities. Inparticular, the growth component o f a change inthe poverty measure i s defined as the change inpoverty due to a change inmean while holding the Lorenz curve constant at some reference level L,. The redistribution component i s the change inpoverty due to a change inthe Lorenz curve while the mean income is constant at the reference level pr.Formally, a change inpoverty over dates t and (t+n) can be written as P ,Lr )+P(z / p r ,Lt+,)-P(Z / p r ,L,)+R t t n -P =P(z / pt t + n ,L,)-P(z / pt + n This decomposition i s not usually exact. A residual (R) exists whenever the poverty measure innot additively separable between p and L. The sectoral decomposition methodology (Ravallion and Huppi, 1991) Changes in a poverty measure P, can be decomposed into an intra-sectoral effect (measuring the contribution o f poverty changes within sectors controlling for the sectors' population share in the base period, nil); a population shqt effect (capturing how changes inthe population shares of sectors contributed to changes inpoverty) and an interaction effect (which captures the correlation between the two). The decomposition o f poverty changes between period 1andperiod 2 can therefore be written as follows: 63 Table 1. Growth-inequality decompositionof changesin extremepoverty (Headcount) between 1995-2002.Welfare measure:consumptionper adult equivalent Period P(t) P(t+l) DPO DPGRldis DPDISlgr Residual Nationwide 1995-1996 0.094 0.063 -0.031 -0.017 -0.013 -0,001 1996-1997 0.063 0.112 0.050 0.057 -0.003 -0.005 1997-1998 0.112 0.113 0.001 0.004 -0.003 0.000 1998-1999 0.113 0.125 0.012 0.017 -0.006 0.000 1999-2000 0.125 0.138 0.013 0.016 -0.003 0.000 2000-2001 0.138 0.114 -0.024 -0.030 0.003 0.003 2001-2002 0.114 0.109 -0.004 -0.009 0.006 -0,001 1995-2000 0.094 0.138 0.044 0.087 -0.031 -0.012 2000-2002 0.138 0.109 -0.029 -0.04 0.01 0.001 Urban 1995-1996 0.046 0.034 -0.012 -0.008 -0.003 -0,001 1996-1997 0.034 0.064 0.03 0.037 -0.003 -0.005 1997-1998 0.064 0.063 -0.001 0.003 -0.003 -0.001 1998-1999 0.063 0.073 0.01 0.013 -0.002 -0.001 1999-2000 0.073 0.092 0.018 0.013 0.005 0.001 2000-2001 0.092 0.06 -0.032 -0.027 -0.009 0.004 2001-2002 0.06 0.054 -0.006 -0.007 0.002 -0.001 1995-2000 0.046 0.092 0.046 0.068 -0.009 -0.013 2000-2002 0.092 0.054 -0.038 -0.035 -0.006 0.003 Rural 1995-1996 0.151 0.097 -0.054 -0.035 -0.018 -0.001 1996-1997 0.097 0.171 0.074 0.081 -0.001 -0.006 1997-1998 0.171 0.173 0.003 0.002 0.000 0.000 1998-1999 0.173 0.187 0.014 0.021 -0.009 0.002 1999-2000 0.187 0.193 0.006 0.016 -0.011 0.001 2000-2001 0.193 0.178 -0.015 -0.021 0.002 0.003 2001-2002 0.178 0.175 -0.002 -0.008 0.008 -0.002 1995-2000 0.151 0.193 0.042 0.093 -0.040 -0.010 2000-2002 0.193 0.175 -0.018 -0.033 0.010 0.005 Legenda:P(t): povertyat time t; P(t+l): povertyat time tfl; DPO: absolutechange inthe headcount DPGRjdis:change inthe headcount measuredue to growthkeepinginequalityconstant DPDISlgr:change inthe headcount measuredue to inequalitykeepinginequalityconstant. Values are roundedto the third decimal. 64 Table 2. Growth-inequality decompositionof changes intotalpoverty (Headcount) between 1995-2002. Welfare measure:consumptionper adult equivalent. Period P(t) P(t+l) DPO DPGRldis DPDISlgr Residual Nationwide 1995-1996 0.254 0.201 -0.054 -0.035 -0.014 -0.005 1996-1997 0.201 0.303 0.102 0.120 -0.013 -0.005 1997-1998 0.303 0.308 0.005 0.008 -0.003 0.000 1998-1999 0.308 0.332 0.024 0.035 -0.012 0.001 1999-2000 0.332 0.359 0.027 0.031 -0.007 0.003 2000-2001 0.359 0.306 -0.053 -0.061 0.006 0.002 2001-2002 0.306 0.289 -0.017 -0.022 0.005 0.001 1995-2000 0.254 0.359 0.104 0.144 -0.053 0.014 2000-2002 0.359 0.289 -0.07 -0.082 0.009 0.004 Urban 1995-1996 0.152 0.125 -0.028 -0.020 -0.009 0.001 1996-1997 0.125 0.202 0.078 0.092 -0.010 -0.004 1997-1998 0.202 0.206 0.004 0.009 -0.006 0.001 1998-1999 0.206 0.222 0.016 0.031 -0.014 0.000 1999-2000 0.222 0.259 0.036 0.035 0.003 -0.002 2000-2001 0.259 0.188 -0.071 -0.063 -0.004 -0.004 2001-2002 0.188 0.176 -0.011 -0.020 0.012 -0.003 1995-2000 0.152 0.259 0.106 0.135 -0.025 -0.003 2000-2002 0.259 0.176 -0.082 -0.083 0.001 0.000 Rural 1995-1996 0.376 0.292 -0.085 -0.061 -0.012 -0.012 1996-1997 0.292 0.423 0.131 0.153 -0.015 -0.007 1997-1998 0.423 0.43 0.007 0.004 0.002 0.000 1998-1999 0.43 0.463 0.034 0.039 -0.008 0.002 1999-2000 0.463 0.478 0.015 0.024 -0.016 0.006 2000-2001 0.478 0.447 -0.03 1 -0.038 0.005 0.001 2001-2002 0.447 0.424 -0.023 -0.020 -0.008 0.005 1995-2000 0.376 0.478 0.102 0.141 -0.055 0.016 2000-2002 0.478 0.424 -0.054 -0.055 0.000 0.001 Legenda:P(t):povertyat time t; P(t+l): povertyat time t+l; DPO:absolutechange inthe headcount DPGRldis:change inthe headcountmeasuredue to growthkeepinginequalityconstant DPDISlgr:change inthe headcountmeasuredue to inequalitykeepinginequalityconstant. Values areroundedto the thirddecimal. 65 The contribution of growth to poverty changes was also sizeable in 2000-02 - had redistribution kept constant, poverty reduction would have been muchmore substantial on account of economic growth alone. Nonetheless, during this period and, in particular between 2001 and 2002, income distribution moved adversely for the poor, thereby partially hampering the impact o f growth on p~verty.~ Most o f the shift in income redistribution that was adverse for the poor was concentrated in rural areas. Interestingly, income redistribution moved favorably for the poor in urban areas between 2000 and 2001. Although the relative magnitude o f the effects differs somehow, the decomposition o f total poverty changes returns a similar picture. Results o f the decomposition are also similar when we use the poverty gap as the poverty measure. Over the two sub-periods the poverty gap first increased and then decreased and growth and redistribution components moved in the same direction for changes in poverty gap and for changes in the headcount. However, the relative magnitude o f the two components suggests that the shift inthe redistributionwas relatively more effective inreducing the relative distance o f the poor from the poverty line than to get people out ofp ~ v e r t y . ~ The reported trends are robust to using consumption per capita as an alternative measure o f welfare. 3.2 Thedistribution of gainsfrom growth: growth incidence curves Because o f the pattern o f substantial economic decline followed by strong growth inthe past two years, assessing the extent to which the gains from aggregate economic growth and the losses from contraction are distributed across households is especially interesting in the case o f Romania. Growth-incidence curves are useful to this end, as they plot the growth rate in consumption for individuals ranked according to their consumption (Ravallion and Chen, 2002).6 Figure 8 reports growth incidence curves for the sub-periods 1995-2000 and 2000-02. The vertical axis measures cumulative growth (or decline) for the whole period, while people on the horizontal axis are ordered by increasing 1evels o fper-adult equivalent c onsumption. The two vertical lines identify the share o f the population in extreme and total poverty, respectively. Note that a downward sloping incidence curve indicates that people in poorer quantiles o f the populationhave benefited from growth (or lost from output decline) more (less) than the average. By 2000, per adult equivalent consumption was about 80% of consumption in 1995. During this period people in the poorer quantiles o f the income distribution were hurt relatively less by the recession than people inthe higher quantiles. This trend was fairly uniform across urban and rural areas and inyear-by-year dynamic^.^ This evidence is consistent with the evolution of inequality for the period - the Gini coefficient decreased from a value o f 0.31 in 1995 to a value o f 0.28 in The interpretation of the tables i s straightforward. For example, the change inthe headcountfor extreme poverty between2000 and2002 was -0.029 (DPO). Hadincome distribution kept constant,poverty would have decreasedmore substantially,precisely by-0.04 (DPGRldis). Conversely, hadgrowth kept constant, poverty would have increasedby 0.01percentage points (DPDISlgr). Results are not reportedbut are availableupon request. More precisely, Ravallionand Chen(2002) define the rate ofpro-poor growth as lFg(p)dp / H, ,i.e. the integral o f the growth rates by quantile - g(p), where p indexesincome quantiles- up to the headcount index H, weighedby the headcountitself. The authors show that this measure is consistentwith the Watts index ofpoverty. 'Anexceptionwasthe pattern ofgrowth between 1999 and 2000, whenpoor inurbanareas were hitby output declinemore thanpeople inurban areas ranking above the 30" percentileinexpenditure.Year-by- year growth incidencecurves are not reportedbut are availableupon request. 66 2000 - and indicates that the recession was accompanied by a compression in the overall structure o f income distribution. It also lends support to the view that in this period social protection policy was effective in dampening the impact on the poor o f the structural reforms initiated in the late nineties. Finally, it i s useful to note that the patterns o f the growth incidence curves are consistent with the evidence we discussed on the partial effects o f growth and redistributionon changes inpoverty. Between 2000 and 2002 output grew by a cumulative 11.9% but the trend in the distribution o f growth rates across quantiles o f the population reversed when compared to the previous years. People below the 50thpercentile, and inparticular the poor, didnot appear to share the benefits o f economic recovery. Overall this pattern appears to be driven by the distribution o f growth of the population in rural areas, where not only growth was substantially lower than the national average, but the poor seemed to be particularly excluded from it. Conversely, growth in urban areas was substantial (a cumulative average growth o f 15%) and the poor fared relatively better than the rest of the urban residents, as they benefited from growthrates between 15 and 17%. When disaggregating the evidence into yearly growth patterns, the extent to which the poor and the lower quantiles of urban population benefited from growth appears to be driven by the developments in 2000-01, and i s most likely due to increased social spending (pension recorrelation and heating subsidies) and patronage that coincided with the peak o f the electoral campaign and was mainly targeted to urban areas. The pattern o f the growth incidence curve for urban areas in2 000-2002 highlights that g ains from growth inurban areas were concentrated among those below the extreme poverty line while people in the neighborhood of the total poverty line enjoyed less than average growth.8 A further dimension along which one could gain insights into the distribution o f growth across quantiles o f the population i s to disaggregate growth incidence curves by economic sector. When we do so, we find that growth patterns were quite uniform across sectors in 1995-2000 (with the poor being hit less by the recession in all sectors). Conversely, between 2000 and 2002, growth was substantially below the (sectoral) mean for all the pensioners and the unemployed inpoverty and for agricultural workers inextreme poverty. * Thisevidence is consistentwith the results from the growth-inequality decompositionwhich indicatethat in2000-01 (theyear that seems to be driving thispattern) inequality movedreIatively more favorably for the urbanpoor inextreme poverty than for the urbanpoor inpoverty. 67 Figure 8. Growth-Incidence Curves Cumulative growth in consumptionper adult equivalent 1995-2000 2000-2002 -5i I I I O - /" \/ , I, '\ 7 - I 6 - -30 - 2 5 - - ~ 6 - 0 10 20 30 40Percentllel 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 PelOentlleS Nationwide Nationwide -15 - 14 -- I 60 90 100 PelcenbleS 50 70 80 -30 ] ' ' 7 - 8 -~ 0 10 20 30 40 0 10 20 30 40 50 60 70 Bo 90 100 Urban Area6 Percentlles Urban Areas O - I 9 - 8 - 7 - 6 - 5 - 4 - 3 - - 2 - 1 - -30 - O - 0 10 20 30 40Peicenbler 50 60 70 80 90 100 0 10 20 30 40 70 90 100 Percenbles 50 60 60 Rural Areas Rural Areas Notes: the vertical lines indicate the share o fpeople inextreme (leftmost line) poverty and total poverty. 68 Figure 9a. Cumulative growth in consumption per adult equivalent 1995-2000, by sector, 1995-2000 -16 4 6 - -20- -22 -28 -30 0 I O 2Q 30 4bPercentiles 50 60 70 80 90 1W Agriculture ;.I2- -11 ------- -22 - -24 - , -22 .21 -26 ,' -28 - ~ -26 .28-- -30 - ' -30- I 0 10 20 30 40 5) Bo 70 80 90 100 Unemployedand others Pers.nllle* -28 1 .30 -- i 0 10 2C 30 40 50 W 70 80 SO 100 Persemes Services 69 Figure9b. Cumulativegrowth in consumptionper adult equivalent 2000-2002 by sector, 2000-2002 10 - 9 - a - 7 - 6 - 5 - 1 - 3 - 2 - 4 - 1 I - o - Pensioners PWCo"llc. 0' 10 20 30 40 50 W 70 M 90 100 Agriculture PWCcnblcs I 7 ~ I 16 - b ' ____ 30 10 Sd 7-- 60 70 BO 90 100 Percenlis Unemployedand others 0 10 20 d 40 5) W 70 80 90 IO0 Services PRCRlUIU 70 3.3 Are poverty changes due to intra sectoral (regional) poverty reduction or to population shifts? Evidencefrom sectoral and regional decomposition ofpoverty The structure o f occupation in Romania has changed substantially over the past s even years.g Following the structural adjustment policies initiated in 1997, the share o f employees dropped significantly in 1998 and decreased consistently since then. The flow out o f employment translated partly inhigher unemployment (which peaked in 2000) and in an increased number o f pensioners, and partly in a higher share o f agricultural and non-agricultural self-employment that acted as a buffer for the effects o f the recession. The reported number o f self-employed can be taken to be a goodproxy for the share o f the population engaged inthe informal economy. Within this category, the increase o f the share o f people employed in agricultural was particularly significant." The rapid growth in the number o f pensioners has been a c onstant throughout the nineties. I t reflected both the increase in the number o f disability pensions which, thanks to lax rules, were often used as coping mechanisms for the unemployed, as well as a sharp increase in retirees in 2001 - the transition period before the implementation of the more restrictive Law on Pension Reform (World Bank, 2002). Figure 9 depicts the shares of the population in different occupational categories for 1995, 2000, and 2002." Figure 9 Trends in occupationaI structurer+Employee ~ I+Self-empl. and I employer 4 Self-employed agriculture -beL-Unemployed ~ ( t Pensioner I 1995 2000 2002 ' -a- Other I Although, for simplicity, we use the term "sectoral" throughout this section, we computed the decompositionacross occupational categories (employees, employers, self-employedin agriculture, etc.). This approach is more likely to allow us to capture movements of workers from the formal to the informal sector. 10 The share o f people in agriculture increased substantially in 2000 and then marginally in the following two years o f growth (2001 and 2002), partly as a result o f the continued process o f land restitution initiated with the landreforminthe early 1990s. "Notethat, becauseofthestructure ofthehouseholdbudgetsurvey, whichcollectsinformationonlyatthe household level, the occupational stnicture o f the population i s constructed by attributing the same ' occupation o f the head o f the household to all family members. 71 Figure 10 depicts consumption growth by occupational category for the sub-period 1995- 2000 and 2000-02. Figure 10 Growth in mean consumption per adult equivalent 400% - 30 0% - I I 10 0% OSelf-employed non-ag .Self-employed ag $ 00% OUnemployed -10 0% aPensioner 'bother -20 0% ,, OTotal -30 0% 1 4 0 0% - 1995-2000 2000-2002 Not surprisingly, self-employed in non-agricultural activities (a category which also includes employers) were subject to substantial volatility - self-employed were hit more strongly than average by the recession and benefited more than average from the recovery. This trend is reversed among agricultural workers, who were relatively more isolated from the recession It is interesting to couple this information with changes in the structure o f poverty by occupational categories. The share of employees in extreme poverty has declined substantially since 1995, so that in 2002 less than 10% o f the poor were employees. Conversely, the share of agricultural workers in extreme poverty increased from less than 25% in 1995 to around 32% in 2002. Finally, the share o fpensioners among the poor increased marginally, but steadily, reaching 33% in 2002 (see figure 11). 72 Figure 11 Occupational structure of the poor (extreme poverty), %of I all poor .... . .....-..........-._. . .. -. .-. - . . -. .. ._ 1995 2000 2002 ~ Year I ~ 0unercployed pensioner other L These trends lend support to the interpretation that people have progressively moved towards occupational categories with increasing poverty. Inthis context, it i s interesting to quantify with precision the extent to which sectoral growth and inter-sectoral shifts o f population contributed to poverty reduction (increase). The sectoral decomposition introduced by Ravallion and Huppi (1991) allows us to disentangle the intra-sectoral effects from the effects o f population shifts across sectors. The related evidence i s reported in table 3 for changes in the headcount for extreme and total poverty. Results for the sectoral decomposition o f changes in the poverty gap and poverty severity (not reported) retum a similar picture. In the period 1995-2000, extreme poverty (as measured by the headcount) increased. A substantial part o f this increase (around 60%) was due to increases in poverty within s ectors, particularly among pensioners and agricultural workers. Shifts o f population across activities accounted for 30% o f t he poverty increase. This reflected the strong decrease in the share o f employees, which resulted in an increase in the share o f agricultural worker, pensioners, and unemployed. The sign o f the population-shift effect indicate that people moved towards sectors with increasing poverty over the period. Inthe following two years (2000-02), almost all of poverty reduction was due to intra-sectoral effects and was concentrated among pensioners (who benefited from the recorrelation exercise), employees, and workers in the informal, non-agricultural sector. Population shifts across occupational categories indicate that people moved towards occupations were poverty was decreasing at a lower rate, but this effect was o f limitedmagnitude. Results for the sectoral decomposition o f total poverty are similar, although it is notable that in 1995-2000 an increase in total poverty among employees contributed substantially to the overall change inthe indicator. A similar exercise can be camed over to investigate whether population shifts across regions and urban and rural areas have significantly contributed to changes in poverty rates. The evidence suggests that most poverty reduction (or increase, as it i s the case for the period 1995-2000) was due to intra-regional effects. Moreover, no significant contribution to poverty changes was due to 73 movements of people across urban and rural areas. Most of the poverty change in both sub- periods was due to changes inpoverty inurban areas. Results are similar for decomposition o f the headcount ratio as calculated with the total poverty line as well as for the poverty gap for extreme and total poverty. Table 3. Sectoral and regional decomposition of changes in poverty EXTREM OVERTY TOTAL WERTY Period 1995-2000 2000-2002 1995-2000 2000-2002 Poverty inyear 1 1.0938 1.1379 12544 0.35857 Poverty in year 2 1.1379 1.1091 1.3586 0.28897 Absolute change 0.044 0.029 1.1041 -0.0696 Yochange 46.9 -20.8 40 9 -19.4 Pop. % of Pop. % of Pop. % of Pop. % of share, Abs. change ir ,hare, Abs. change in share, Abs. change ir share, Abs. change in yearl Change' pov. yearl change pov. yearl Change pov. yearl change pov. Activity Emp1oyees 49.54 0.00384 8.72 38.33 -0,0095 33.14 49.54 0.0286 27.46 38.33 -0.0323 46.38 Self-employed non-agr. 3.35 0.0037 8.39 4.68 -0.0063 21.99 3.35 0.00564 5.41 4.68 -0.0076 10.89 Self-employed agr 8.56 0.0062 14.09 10.23 -0.0034 11.96 8.56 0.00714 6.86 10.23 -0.0046 6.6 Unemployed 5.1 0.00159 3.61 8.63 -0.0009 3.19 5.1 0.00374 3.59 8.63 -0.0045 6.51 Pensioner 31.84 0.00891 20.23 36.83 -0.0094 32.56 31.84 0.02815 27.03 36.83 -0.0233 33.49 Other 1.61 0.00271 6.15 1.31 -0.0007 2.42 1.61 0.00196 1.88 1.31 -0.0003 0.43 Infra-sectoral effect 0.027 61.18 -003 105.26 0.0752 7223 -0.073 104.29 Population-shqt effect # 0.0133 30.22 0.0021 -7.31 0.0251 24.13 0.0026 -3.67 Interaction effect 0.0038 8.6 -0.0006 2.04 0.0038 3 64 0.0004 -0.63 Region North-East 16.86 0.01125 25.54 17.12 -0.0052 18.06 16.86 0.01845 17.72 17.12 -0.0102 14.6 South-East 13.07 0.00925 21 13.13 -0.0047 16.31 13.07 0.01585 15.22 13.13 -0,0081 11.65 South 15.6 0.00428 9.71 15.52 -0.002 6.87 15.6 0.01928 18.51 15.52 -0.0107 15.41 South-West 10.8 00028 6.36 10.75 -0.001 3.39 10.8 0.00653 6.27 10.75 -0.0022 3.1 West 9.09 0 00541 12.29 902 -00041 14.38 9.09 0.0111 10.66 9.02 -0.0071 10.19 Norfh-West 12.71 0.00469 10.65 12.67 -0.0046 16.15 12.71 0.01554 1492 12.67 -0.0145 20.79 Center 11.75 0.00394 8.96 11.76 -0.005 17.47 11.75 0.00883 8.48 11.76 -0.0094 13.49 Bucharest 10.12 0.00207 4.7 10.04 -0.0022 7.59 10.12 0.00803 7.71 10.04 -0.0076 10.88 Intra-sectoral effect 0.0437 99.2 -0.029 100.21 0.1036 99.5 -0.07 100.11 Population-shfi effect 0.0002 0.55 0.00007 -0.26 0.0005 0.47 0.00008 -0.11 Interaction effect 0.0001 0.25 -0.00001 0.04 0.00003 0.03 0.00000 -0.01 U r b a f l u r a l Urban Areas 54.4 0.02488 56.49 54.37 -0.0206 71.63 54.4 0.05773 55.45 54.37 -0.0447 64.27 Rural Areas 45.6 0.01914 43.46 45.63 -0.0081 28.09 45.6 0.04634 44.5 45.63 -0.0247 35.5 Intra-sectoral effect 0.0440 99.94 -0.029 99.72 0.1041 99.95 -0.069 99.76 Population-shgt efecr 0.0000 0.06 -0.0001 0.24 0.00005 005 -0.000I5 0.21 Interaction effect 0 0 0.0000 0.05 0 0 0.0000 0.03 Total change in poverty 100 0.04405 100 100 -0.0287 100 100 010413 100 100 -0.0696 100 *This column reoorts solute changes in poverty in each sector. These changes sum up to the intra-sectoral effect. JIndicate poverty increased by 4.4 percentage points. Of these, 2.7 percentage points were due to increase in poverty within sectors (61.18%), while 1.3 percentage points (30.25%) were due to shifts of the population towards activities were poverty increased. 74 4. Sources of growth The recovery o f the Romanian economy that started in the late 2000 was impressive. In this context, an important issue i s the extent to which growth i s sustainable in the medium run.There i s consensus on the fact that macroeconomic stability and responsible fiscal policies are crucial preconditions for sustained the growth over time. However, structural development in a number o f specific areas, such as improvement o f the investment climate and the revitalization o fthe rural economy, can lay the foundations for equitable growth inthe long run. The cross-country literature as well as microeconomic studies highlight the importance of good investment climate as an engine o f growth. Many studies have shown that measures o f good governance, enforced property rights, and low corruption are causally associated across countries with higher growth inthe short run (see for example, Knack and Keefer, 1995 and Mauro, 1995) and with higher income per capita in the long run (see Kaufman et al., 1999, and Hall and Jones, 1999). Results from microeconomic evidence confirm the strength o f this association. For example, using firm-level data for five countries, Dollar et al. (2003) find that indicators o f investment climate (such as number of inspections per year, number o f days to clear customs or to obtain a phone line) are important determinants o f firm productivity and growth. Moreover, good institutions are believed to promote FDIflows (see World Bank, 2003) Strengthening institutions i s a particularly relevant issue inthe Romanian economy. According to Transparency International (TI), in 1998 Romania was among the top third most corrupted countries in the world (ranking 6lStout o f 85 countries inthe TI index o f lack o f corruption, with a Corruption Perception Index o f 3/10 -where 10 indicates absence o f corruption). To the extent that comparison o f the index over time are informative, Romania's corruption index had fallen by 2002 to 2.6, with the country ranking 77thout o f 102 countries." Findings from a very detailed World Bank report that surveyed households, enterprises, and public officials also indicates perceptions o f a highlevel o f corruption surrounding the implementation and design o f laws and regulations. Firms reported corruptionto be a substantial obstacle to their business. An analysis o f the patterns o f informal payments revealed that corruption has important distributional consequences: poorer households appeared paying larger share o f their income than richer families inthe form o f bribes and the data suggested that access to health care might be precluded to poor households who cannot afford to make unofficial payments. (World Bank, 2000). Rationalizingthe regulatory burdencan also be highly beneficial, especially inan economy that i s witnessing a progressive move o f the work force from the formal to the informal sector. This need for institution building has long been recognized and governance reforms are at the core o f the World Bank support to the private sector development and to the strengthening of accountability inthe public sector (see CAS, World Bank, 2002). Poverty in Romania has an important regional dimension and is particularly more entrenched in rural areas (see table 4). Rural poverty i s a reflection, amongst others, o f the collapse o f agricultural production and o f the absorption o f displaced industrial workers. Substantially slower growth rates (and only marginally lower decline rates o f output during recession times) account for the persistence o f poverty in rural areas. In this context, devising strategies to revitalize the rural economy can accomplish the twofold objective o f tackling lack o f growth for a large share o f the population as well as addressing an important dimension o f inequality within the country. The World Bank Country Assistance Strategy has moved along these lines and targeted a large l2 For information onhow the TI index is constructed see httu://www.transparencv.org. 75 share o f its interventions to rural areas by, amongst other things, financing capacity building and infrastructure activities (see CAS, World Bank, 2002). Table 4. Urban and rural poverty and growth rates 5. Poverty projections Inthis section we discuss resultsfromprojections for the headcountratios for extreme poverty for the years 2003-2007 using two altemative methodologies. First, we compute poverty using predicted future consumption obtained by applying a uniform growth rate to the whole structure o f current consumption (base projections). Second, w e assume a balanced growth path across occupational categories and, while keeping sector shares constant at the existing level (2002), we apply to each sector the growth rate that i s compatible with the assumed overall growth rate. On this basis we thencompute povertyrates (sectoral growth).. We first implement a simple historical validation o f the two procedures. In particular, (i) we apply to the 2001 consumption structure the average growth rate between 2001 and 2002 to predict consumption in 2002 and then calculate poverty (base); (ii)using 2001 sector shares in output and 2000-2001 ratios o f growth by occupational categories, we predict consumption in 2002 and calculate headcount ratios (sectoral projections). We then compare the poverty measures obtained in (i) and (ii) actual poverty measures. to Inthe base projection, the source offorecast error lies in applying to every individual inthe distribution the same growth rate (while we know from growth incidence curves that in this period the poor benefited from growth less than richer people). The base projections turn out to be very close to the actual poverty rate, although, as one would expect, they slightly underestimates poverty. Insectoralprojections bothhypotheses of fixedsector shares and balanced growth are potential sources of forecast error. However, by diversifying the growth rate by occupation, we expect these projections to be more accurate than base projections (see table 5). Table 5. Historical validation for base and balanced sectoralgrowth projections, 2001-2002 Extreme Poverty Headcountin Total Poverty Headcountin2002 2002 Predicted Actual Error Predicted Actual Error Per capitaconsumptiongrowth rate 3.3 3.3 3.3 3.3 Base (3.3%) 0.104 0.109 4% 0.283 0.289 2% Sectoral(3.3%) 0.106 0.109 3% 0.284 0.289 2% 76 The results o f this exercise suggest that, in the short term, these methodologies might produce reasonably good projections o f poverty rates. However, as the horizon lengthens, projections are bound to becomes less informative, not last because the assumption o f constant sector share and balanced growth across sectors might be less adequate. Projections are presented for growth in consumption per adult equivalent (see discussion in the appendix for the choice o fwelfare measure) and for a number o f different growth scenarios. The benchmark growth rate for 2003-2007 is taken from IMF projections on per capita consumption g r 0 ~ t h . IThese growth projections are constructed using consumption from the ~ National Accounts. It i s useful to note that there exists a substantial drift between consumption data reported by national accounts and the data (see table 6). This divergence has been noted in other contexts and has been linked to a possible poor representativity o f the survey andor methodological discrepancies in the construction o f the consumption aggregate (see for example Tesliuc, 2001). Table 6. DriftbetweenNational Accounts and HouseholdBudget Survey consumptionaggregates Household consumption 1996 1997 1998 1999 2000 2001 2002 National Accounts per capita % annual change, real terms 8.1 -3.6 0.7 -2.4 -0.7 6.4 3.3 Household Budget Survey per capita % annual change, realterms 6.6 17.2 -1.2 -4.9 -4.2 8.4 3.3 Source: National Accounts @A) and Household Budget Survey. Per capita consumption growthfrom NA is calculated based on the hypothesis of annualpopulation growth of -0.1 %. Alternative assumptions are also adopted, corresponding to an optimistic scenario (sustained 5% annual growth over the whole period), two middle ground scenarios (one where growth is sustained at the rate o f 4%, the other where growth continues at the current annual 3.3%), and a low growth scenario (where growth i s at annual rate o f 2.5% throughout). Table 7 reports the projections for the headcount ratio for both extreme and total poverty. l3 As discussed inthe appendix, consumption growth o fper capita andper adult equivalent consumption track each other very closely. 77 Table 7. Poverty Projections, 2003-2007 Growth rate Scenarios Extreme Poverty Total Povero 2003 2004 2005 2006 2007 2003 2004 2005 2006 2007 IMF'sProjections of p.c. growth rates* Growthprojections 4.9 4.3 4.1 4.1 3.9 4.9 4.3 4.1 4.1 3.9 Base 9.36% 8.22% 7.17% 6.20% 5.42% 25.82% 23.17% 20.73% 18.82% 16.77% Sectoral 9.60% 8.72% 7.95% 7.35% 6.78% 25.98% 23.84% 22.00% 20.01% 18.56% 2002 growth rate Growth projections 3.3% 3.3% 3.3% 3.3% 3.3% 3.3% 3.3% 3.3% 3.3% 3.3% Base 9.91% 8.84% 7.99% 7.18% 6.36% 26.73% 24.75% 22.74% 20.75% 19.25% Sectoral 10.16% 9.19% 8.59% 7.95% 7.44% 26.77% 25.09% 23.53% 22.01% 20.34% 5 Percent Growthprojections 5% 5% 5% 5% 5% 5% 5% 5% 5% 5% Base 9.35% 7.98% 6.84% 5.63% 4.72% 25.75% 22.73% 19.88% 17.51% 15.31% Sectoral 9.57% 8.59% 7.77% 7.01% 6.59% 25.95% 23.54% 21.12% 19.14% 17.41% 4.0 Percenta Growthprojections 4% 4% 4% 4% 4% 4% 4% 4% 4% 4% Base 9.64% 8.55% 7.48% 6.49% 5.58% 26.35% 23.82% 21.43% 19.45% 17.49% Sectoral 9.86% 8.94% 8.21% 7.57% 6.97% 26.42% 24.42% 22.61% 20.56% 19.08% 2.5 Percent Growthprojections 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% Base 10.15% 9.34% 8.67% 7.93% 7.33% 27.28% 25.72% 24.14% 22.64% 21.13% Sectoral 10.30% 9.46% 8.82% 8.43% 7.93% 127.27% 25.42% 23.99% 22.71% 21.62% Notes: actual extreme poverty rate in 2002 = 10.91%; actual total poverty rate in 2002 = 28.89%. *Per capita growth rates are computed assuming a constant annual population growth of -0.1%. "The 2001 CAS projected per-capita consumptiongrowthat 4% for 2003. 6. Conclusions Inthis work we analyzedthe role of economic growth for poverty reductionfor the years 1995- 2002 in Romania. During this period the c ountry experienced a long recession followed by a substantial economic recovery. Poverty dynamics mirrored mean income developments quite closely. However, we have shown that changes in income inequality were important determinants o f changes in poverty. In particular, during the penod 1995-2000 (mainly o f output decline), income distribution moved favorably for the poor, thereby hampering the impact o f the recession on poverty. Conversely, in the following period the poor and the extreme poor benefited from growth less than richer strata o f the population. This pattem appeared to be partly driven by income dynamics inrural areas, where growth was substantially lower than the mean and skewed towards people above the 50% percentile inthe consumption distribution. 78 During the recession, poverty increase was due not only to increase inpoverty indifferent activity categories (especially among pensioners and agricultural workers) but also to people moving out o f employment into early retirement and work in the informal sector. Inter-sectoral shifts were not as substantial during the economic recovery. An important issue to tackle is how to sustain the important economic recovery the country has experienced int he past two y ears. There is consensus that macroeconomic s tability and fiscal responsibility are crucial preconditions for sustained growth. However, structural developments can lay the foundations for growth in the medium and long run.In this context, an improvement o f the investment climate and govemance could on one hand maintain the recent investment expansion and address a number o f distributional issues, such as improving access o f the poor to social services. Moreover, a revitalization o f the rural economy could provide an engine for equitable growth. Projections under different scenarios o f growth suggest that if the economic recovery i s sustained inthe nextfive years, poverty canbecurtailed substantially. 79 Appendix Adult equivalence (AE) scales allow to correct per capita measures o f consumption for the cost o f children relative to adults and for economies o f scale enjoyed at the household level. Although most s cholars a gree that measuring c onsumption in adult e quivalent terms is c onceptually the correct approach, the debate on how to implement practically such correction is longstanding (see, for example, Lanjouw and Ravallion, 1995). In general, poverty appears to be over- estimated when comparing poverty rates based on per capita consumption to poverty rates based on per AE consumption. However, since the AE adjustments include elements o f arbitrariness, the actual extent o fthe over-estimation o fpoverty is hard to gauge. Inthiswork, the case for using consumption per adult equivalent (CAE) c an be made o n the ground that it is consistent with recent poverty work in Romania (se Tesliuc et al.) and with the rest o f this report (for a description o f how the consumption measure i s constructed, see the poverty profile by Pop and Tesliuc (2003)). However, it i s useful to highlight the main differences inthe dynamics o f poverty when measured with respect to CAE and per capita consumption (CPC). Figure 12 depicts the evolution o f the headcount ratio for CAE and CPC for extreme poverty. When based on per capita calculations, consumption i s on average 20% lower than CAE and poverty rate. Growth rates in CAE and CPC track each other almost perfectly (figure 13). 80 Figure 12 I Evolution of extreme poverty rates, 1995-2002 25%- 20%- I I 10%- 5% 0% Year 1995 1996 1997 1998 1999 2000 2001 1 Year I ,+Peradult I equivalent -- I -1 Figure 13 Growth in mean consumptionper adult equivalent and per I capita, 1995-2002 I -20% 1995-1996 1996-1997 1997-1998 1998-1999 1999-2000 2000-2001 2001-2002 1 Period I 81 References Dollar, David, Mary Hallward-Driemeier and Taye Mengistae (2003). "Investment Climate and Firmperformance inDevelopingEconomies," Manuscript, World Bank. El-laithy, Heba ,Michael Lokshin and Arup Banerjee (2003). "Poverty and economic growth in Egypt, 1995-2000," Manuscript, World Bank. Hall, Robert and Charles Jones (1999). "Why D o Some Countries Produce So Much More Output Per Worker Than Others?," Quarterly Journal ofEconomics. 114(1), pp.83-116. Kaufmann, Daniel, Aart Kraay, and Pablo Zoido-Lobaton (1999). "Governance Matters," Manuscript, The World Bank. Knack, Steve and Phil Keefer (1995). "Institutions and Economic Performance," Economics and Politics, 1, pp. 207-27. Lanjouw, Peter and Martin Ravallion (1995). "Poverty and Household Size," TheEconomic Journal, Vol. 105, no. 433. pp. 1415-1434 Mauro, Paolo (1995). "Corruption and Growth," Quarterly Journal of Economics, August, no.110, pp. 681-712. Ravallion, Martin and Shaoua Chen (2002). "Measuring pro-poor growth," World Bank Poverty Research Working Paper, no. 2666. Ravallion, Martin and Gaurav Datt (1992). "Growth and redistributioncomponents o f changes in poverty measures : a decomposition with applications to Brazil and India inthe 198Os," Journal ofDevelopment Economics no. 38, pp. 275-295. Ravallion, Martin and Monika Huppi (1991). "Measuring changes inpoverty: A methodological case study of Indonesia during an adjustment period," WorldBank Economic Review no.5, pp. 57-84. Pop, Lucian and EmilTesliuc (2003). "Poverty Profile," background paper for the Romania Poverty Assessment 2003. Tesliuc, Cornelia, LucianPop and EmilTesliuc (2001). Saracia si sistemul deprotectie sociala, POLIROM, Bucarest. World Bank (2000). Diagnostic Surveys of Corruption Romania. World Bank (2002). Romania: County Assistance Strategy, Reportno. 22180-RO. World Bank (2002). Romania, Building Institutions for Public Expenditure Management: Reform, Eflciency and Equity, Public Expenditure and Institution Review, August. World Bank (2003). Global Economic Prospects. 82 Labor Force Participation.Unemployment andthe Poor Cem Mete INDEX 1 Introduction................................................................................ . 84 2 Poverty. employment. andUnemployment.......................................... . 88 3. The unemployed: who are they. who are the most vulnerable amongthem? The characteristics of the unemployed.................................................. 90 Vulnerability to becoming unemployed................................................. 92 Moving out of unemployment -.. term unemployment.......................... long 92 More on thejob-search process.......................................................... 93 4 The individuallevel determinants of employment andwages.,...................95 . 96 Incomefrom Employment............................................................... Employment................................................................................. 98 5 Conclusions.................................................................................. . 99 83 1. Introduction The end o f planned economy era in Romania had significant implications for the labor market. Overall characteristics are shared by many other transition countnes: emergence o f open unemployment, reductions in employment rates, being stuck with a large industry sector (and loss making state-owned enterprises), a wage structure with little if any connection to productivity, an underdeveloped financial sector not capable o f responding to the needs o f the private sector that i s supposed to flourish under such adverse conditions. Sometimes changes over time are impressive though, while at other times much needed reforms are only slowly implemented with mixed results. This section provides an overview o fthe labor market trends inRomania, to provide a context for the empirical analysis that follows. An attempt is made to discuss time trends when comparable data are available (figures based on Household Budget Surveys and Labor Force Surveys go back till 1995 and 1996 respectively). The chapter emphasizes poverty linkages for obvious reasons, and it further limits coverage to micro issues because the Country Economic Memorandum (under preparation) focuses on macro topics with an emphasis on Romania's integrationwith the EU. Employment rate o f those who are 15 years and older steadily decreased from slightly over 60 percent in 1996 to less than 59 percent in2001 (Figure l).' The male-female gap inemployment i s decreasing over time, although not because o f increases in female employment rates but because of decreases in male employment rates. The Roma are much less likely to be employed, but they are catching up with the rest of the population with employment rates rising from 39 percent in 1996 to 48 percent in2001.2 ' Ifone focuses on individuals aged 15 to 64, the employment percentages are 72 percent in 1995 and 68.5 percent in 2001 (slightly below the EU average o f 69.2 percent). Figure 1 is based on the employment definition used by the Romanian Statistics Institute, which i s "all people aged 15 years and over, who have carried out an economic or social activity producing goods or services, with a duration o f 1 hour at least during the reference period (one week), with a view to achieve certain incomes inform o f salaries, inkind remuneration or other benefits." The 2002 employment estimate is not included in Figure 1 because the relevant survey questionhas changed inthat year and thus it may not be comparable to the other years. According to various I N S surveys, the share o f Roma in the population varies between 2 percent and 2.5 percent. This is believed to be an underestimation, partly due to individuals not revealing themselves as Roma. The under representation o f Roma may or may not be a problem, depending on the objective o f the analysis. For sample size purposes, in our case this i s not a major issue: information about thousands o f Roma exist inboth the Household Budget Surveys and the Labor Force Surveys. The measurement error i s not necessarily a problem if it stems from individuals not identifying themselves as Roma and if one i s willing to focus attention on individuals who do identify themselves as Roma. The other major suspect in the underestimation is that the areas that are heavily dominated by Roma are less likely to be represented in the survey sample units. While the P A team's discussions with the I N S officials did not produce support for this possibility, it can not be ruled out at this stage. Finally, note that the extent to which this underreporting results in misleading insights in multivariate analysis depends on the extent to which excluded Roma differ from the included Roma. 84 Figure 1. Employment by ethnicity and gender, 1996-2001 (World Bankstaffcalculationsbased on LFS) 70 - 65 - 60 1 QJ 5 5 - *3t 5 50 p" z 45 - 40 - 35 - 30 ~ 1996 1997 1998 1999 2000 2001 Year 14-All +Romanian+Hungarian-8-Roma+Other +Female + Male The employed population is aging: the share o f individuals aged 65 and over in total employed population increased from 8.2 percent in 1996 to about 10 percent in 2001.3 One implication o f this trend which is emphasized by the Government of Romania and European Commission Joint Assessment o f Employment Priorities in Romania (October 2002) i s that the elderly tend to be low-skilled compared to the rest o f the population. Having said that, it is important to make the urbadrural distinction to better understand these figures, since about 90 percent o f the employed elderly (defined to be ages 65 and over here) reside inrural areas -making up almost 18 percent o f the rural employment. The transition to market economy is underway, with a sizable shift in recent years from state to private sector employment (Figure 2). But the restructuring o f state enterprises has been slower than originally intended, with large losses especially in the manufacturing sector. The layoffs routinely trigger protests (most recently protests took place to oppose layoffs inthe steel plants, in the mining sector, and tractor & truck manufacturing plants). The unions negotiate severance payments with the government on a case bask4 Romanianpopulation has been declining since 1990, with an estimateddecrease of 1.8millionuntil2020. The structure of the "dependent population" is changing, becoming more skewed towards the elderly as opposedto the children. This is primarily because of smaller-than-replacementtotal fertility rates at around 1.3 in recent years, but in the coming years it might be reinforced by increases in life expectancy -life expectancy at birth is quite stable over time in Romania, which hit 69 in 1973 and since then recorded values between69 and 70. A large scale agreement inApril was 20-24 average monthly wages to be paid to dismissed employees, in addition to two average wages payable at the time of`dismissal. Furthermore, micro-credit programs for entrepreneurs are being launchedinmining regions hit hardest by the restructuring. 85 The success stories in privatization tend to involve relatively small enterprises.j The recent record is mixed, with some movement (e.g., Romtelecom, Tepro pipe maker) but many pending operations with failed attempts to privatize (most visible one being Petrom, the national oil company). Reasons for the slow progress are varied, in some cases having to do with the Government's desire to keep the controlling share or the requirementsto retain a sizable portion o f the employees after privatization. Also it has proven to be difficult to attract private investors for certain companies. Employment protection legislation o f Romania is considered to be more restrictive than the OECD average but less restrictive than many other transition economies6 Having said that a labor code which became effective on March 2003 introduced further restrictions on employers when it comes to hiringand firing, and strengthened the role o f unions. Within private sector employment a sizable portion is informal work (Figure 3). This i s perhaps not surprising, given the high payroll taxes and social security contributions which are discussed below. Since 1996, the share o f self-employed and unpaid family workers remained stable (if anything, increased slightly). Gender differences are visible, males being much more likely to be self-employed and females being much more likely to be unpaid family workers. Trends by urban versus rural residence (not shown in the graph) reveal significant but expected differences: in rural areas being an employee ( 26 percent versus 90 percent in urban) or employer (0.5 percent versus 2.1 percent in urban) is less likely while being self-employed (39 percent versus 6.1 percent in urban) or unpaid family worker (34 percent versus 2.2 percent in urban) i s more likely. Figure2. State Versus Private Sector Employment by gender, 1996-2002 Figure3. ProfessionalStatus of those who worked, 1996 and 2002 Wudd B m k ~ t ~ ~ ~ ~mnLFS)l ~ m ~ h ~ t (Wodd Bsnkap=islrul.amibsvdm LEE) 801 70 60 50 *$ 40 $ 30 20 1996dl 2002 all 1990 male 2002malc 1996iemalc 2002 fcmslc 1996all 2002 all 19% male 2002male 1996iemale 2002 female 1 IStateIPrivate0Other Entrepreneurshipand EnterpriseDevelopment. Romania.March2002. OECD and EBRD. See John Haltiwanger, Stefan0 Scarpetta, and Milan Vodopivec. 2003. How Institutions Affect Labor Market Outcomes: Evidence from Transition Countries. Manuscript, the World Bank; also on regulatory framework see Labour Market and Social Policies inRomania.2000. OECD. 86 Inthe second halfof 199Os, the private sector's share (added value) inagriculture has been more than 90 percent followed by construction at around 80 percent, services at around 70 percent and industry less than 60 percent (although a significant increase from less than 40 percent in 1995).' Access to finance i s problematic for SMEs. For the large enterprises, the main sources o f financing are internal-fundsiretained-earnings (52 percent) and local commercial banks (23 percent), while for SMEs the main sources o f financing are intemal-fundshetained-earnings (65 percent) and familyifriends (15 percent) with local commercial banks a distant third at slightly over 5 percent.' The relative contribution of services to GDP increased significantly over time, from 42.9 percent in 1995 to 56 percent in 2001. On the contrary, both agriculture and industry lost ground.' Interestingly though, the distribution o f employment favored agriculture during this period (maybe in part due to a move towards subsistence agriculture): percentage o f employed in agriculture/forestry/fishing increased from 34.4 percent to 41.4 percent; but decreased from 28.6 to 23.2 inindustry and from 37 to 35.4 inservices. Such a trend i s quite disappointing since even in 1997 one o f the structuralreform priorities was re-deployment o f labor t o non-agricultural jobs. Inrecent years the unemployment rate is tends to be close to, but less than 10percent (depending on the definition o f unemployment). While 10 percent unemployment certainly deserves attention, among transition countries this figure is on the low side." The registered unemployment rate was 9.5 in 1995; with some fluctuations inbetween it became 8.6 in2001.I' Both sizable unemployment and high share o f informal sector workers among the employed can be linkedto a labor market that is not flexible. Payrolltaxes remain high at 52 percent of salaries, despite a slight decrease in January 2003 - prior to that, the taxes made up 57 percent o f salaries." ''EBRDiWorld See the Romanian Government's National Action Plan for Employment (2002) for more on this. Bank Business Environment and Enterprise Perfonnance Survey, 1999. The sector contributions to GDP in 2001 were 15 percent by agricultureiforestryifishing, 29 percent by industry and 56 percent by services. For 1995, these percentages were 21,5; 35.6 and 42.9 in the same order. lo Vodopivec, Worgotter and Raju (2003) calculated unemployment rates based on comparable labor force survey data for a number o f transition countries. In2000, Romania with an unemployment rate calculated as 7.7 percent fared better than Bulgaria (18.7 percent), Estonia (13.5 percent), Latvia (14.4 percent), Lithuania (15.9 percent), Poland (16.6 percent), and Slovakia (19.1 percent). The only countries in authors' list with comparable or less unemployment to Romania were Czech Republic (8.8 percent), Hungary (6.6 percent) and Slovenia (7.1 percent). Milan Vodopivec, Andreas Worgotter and Dhushyanth Raju. 2003. "Unemployment Benefit System in Central and Eastem Europe: A Review o f the 1990s". Social ProtectionDiscussionPaper Series No. 0310. The World Bank. *'Some policies (such as requiring registrationwith unemployment office as a prerequisite for qualification for MIG benefits) may result in changes in registered unemployment rate but not in the ILO definition o f unemployment rate or in the percentage o f individuals who consider themselves unemployed. Indeed the ILOunemployment rate increased slowly but steadily from 6.7 percent in 1996 to 7.1 percent in2000. 12 The current rates are distributed as follows: pensions (9.5 percent employee and 24.5 percent employer), health (6.5 percent employee and 7 percent employer) and unemployment fund (1 percent employee and 3.5 percent employer). 87 2. Poverty,employment, and unemployment The linkage between poverty, type o f employment and unemployment may seem trivial at first, but the strength of the relationship could vary. For example, in cases where unemployment is high for younger adults and not for others, one could expect a relatively weak relationship since unemployment is an "individual level" event and poverty is measured at the household level. Or ifthere is a substantial exchange offunds across households (e.g., through transfers from abroad), then unemployment may not necessarily lead to poverty - which i s a consumption based measure as defined in this study. Indeed, many policy makers in Romania seem to suspect a rather weak relationship between unemployment and p~verty.'~ I t i s useful to illustrate the relationship between unemployment and poverty via two questions, with obvious "yes" or ``no'' answers but not so obvious magnitudes that go with them. Unemployed adults: Are they more likely to bepoor? The 2002 household budget survey data suggest that 44.9 percent o f unemployed adults o f ages 15 to 64 are poor, as opposed to 25.8 percent o f adults o f the same age group who are either working or who are not in the labor force.l4 Interestingly, self-employed adults in agriculture are more likely to be poor than the unemployed: 55.6 percent o f this group are poor." Poor adults: Are they more likely to be unemployed? About 14 percent o f poor adults reveal themselves to be unemployed, as opposed to 6.6 percent for the non-poor.'6 Once again self- employment in agriculture comes into play: 29 percent o f poor adults are inthis group." At the household level, the linkage between share of unemployed individuals ina household and consumption per adult equivalent i s shown by Figure 4 (Quartic kernel, with a bandwidth o f 0.25 i s used to obtain the graph). There i s a clear negative relationship between share o f unemployed individuals and household consumption (correlation coefficient i s -0. l), the function being l3 Based on discussions between the PA team and senior policy makers. Occasionally this issue enjoys direct or indirect coverage in newspapers, under titles "false unemployment...2 billion USD entered Romaniainone year only...". l4 Similarly, 19.9 percent of unemployed adults are below the extreme poverty line as opposed to 9.4 for the remainder adult population. In this case, since we are using the Household Budget Survey data (with consumption information to define poverty), unemployment figures are not based on the ILO definition of unemploymentbut rather whether individuals reveal themselves to be unemployed. 15 A more general question is what is the relationship between occupational status and poverty? The least likely to be poor individuals were employers (2.5 percent, 0 percent extreme poor) and employees (11.2 percent, 2.1 percent extreme poor); followed by pensioners (20.5 percent, 5.6 percent extreme poor) and students (25.1 percent, 8.3 percentextremepoor); and finally housewives(39.2 percent, 18 percentextreme poor), self-employednon-agriculture (40.9 percent, 17.7percent extremepoor), unemployed (44.9 percent, 19.9 percent extreme poor), the category "other (dependent, military service etc.)" (55.4 percent, 27.6 percentextremepoor), and self-employedagriculture (55.6 percent,24.6 percentextremepoor) follow. 16 For the extremepoor this figure increases to 16.7 percent. 17 And the more general question is how does the occupational status of poor adults differ from the remainder ofthe adult population? The poor adults are less likely to be employees (14.6 percent versus 43.9 percent), less likely to be employers (0.05 percent versus 0.68 percent), more likely to be self- employed non-agriculture (5.05 percent versus 2.76 percent), more likely to be self-employed agriculture (29 percentversus 8.8 percent), more likely to be unemployed (14 percentversus 6.6 percent), less likely to be pensioner(12.3 percentversus 18.2 percent), roughly as likely to be a student (10.4 percentversus 11.8 percent), andmore likely to be a housewife (10.7 percent versus 6.3 percent). The percentagesfor the extreme poor are, in the same order, 7.4; 5.8; 34.2; 16.7; 8.9; 9.l;'and 13.1. In each case the remainder responses fall into the other category. steeper at first followed by a more modest negative relationship. While the expected negative relationship is there, unemployment alone is not that good o f a predictor for poverty -which i s consistent with the informal sector findings at the individual level. Remembering that poverty line used inthis study i s 1,535,370 Lei, only when around 75 percent o f the household members are unemployed the household is predictedto fall into poverty." Figure 4. Household Consumption and Unemployment 2.6e+06 -, I -s2 24et06 I! a ; 22e+06 0 , 2 2 Oe+06 E ,/ / 4 w -5 18e+06 - U ~ 2 U , 16e+06 - 1.4et06 0I Share of Unemployed.4Household Members .2 .6 I .8 The main message that comes out o f this set o f descriptives is that while there is a relationship between unemployment and poverty, the informal-employment and poverty relationship i s as strong (see also the predictors o f household consumption regression reported in the poverty profile section). The remainder o f the paper is organized around these findings. The next section elaborates more on the unemployed (and transitions in and out of unemployment), and the final section turns to employment and wages by paying special attention to informal sector workers. 3. The unemployed: who are they, who are the mostvulnerable among them? "We don' t j n d employmentanywhere... Often whenyou try to get ajob, they look a little closer at you and they see you, and ifthey know whether you are a Roma or a Gypsy,as they say, they won't takeyou.'' (Low-incomerespondents,Roma, Alunis, rural. 2003) "The state should help thepoor, not those who are capable offinding employment, who arephysically healthy but refuse to work. Why should the state help these individuals? But no one wants to hire thesepeople anyway, especially if they see that they are Gypsy. There areplenty of other people, who are 40-50 years old, and who cannotjnd employment." (Averageincome respondents,Alunis, rural.2003) "No one wants to receiveyou or listen toyou, anywhereyou go to inquire about (services orjobs) no one receivesyou, because Iam over 45 years old. Do Inot have the right to live, do Inot have the right to work, even though Iam over 45 years old?" (Low incomerespondent,Roma; Targu Mures, urban. 2003) 89 Identifying the losers and winners o f the transition period i s a difficult task. A starting point might be those who have stopped working. Excluding retirees (59.7 percent), the main reasons for stopping work in 2001 were "fired or down-sizing o f personnel" (34.5 percent o f non-retiree responses), end o f temporary activity (26.3 percent o f non-retiree responses) and sickness or invalidity (22.8 percent o f non-retiree response^).'^ Inrural areas "end o f temporary activity" i s much more common and individuals are less likely to be fireddownsized; and Roma are more likely to be cite "end o f temporary activity" and "fired or downsizing o f personnel" as reasons. Since open unemployment did not exist prior to the transition in Romania, it seems plausible to count discouraged workers (difficult to distinguish quantitatively from those who do not want to work), younger individuals who have not worked previously but are looking for work, those who face a higher risk o f unemployment and those who are more likely to experience long-term unemployment among the losers. This section aims to provide some new insights by focusing on the characteristics ofthe unemployed and predictors ofthe durationo funemployment and also by identifying those who are not unemployed but face a higher risk o f being unemployed. Firstbasic descriptives are presented, a very brief summary of empirical findings based on longitudinal data from Labor Force Surveys follows. The characteristics of the unemployed The Roma are much more likely to be unemployed than the rest o f the population. Even if the enormous gap between the Romanians (and Hungarians) and Roma in 1996 has closed over time, the Roma are still more than twice as likely to be unemployed (Figure 5). Females were slightly more likely to be unemployed prior to 1998, after that they are slightly less likely to be unemployed (Figure 6). Inrural areas unemployment i s lower and declining over time, in urban areas it i s highand does not display a declining trend (also Figure 6). One could come up with competing stories to explain "unemployment by school attainment graphs" by speculating on reservation wage differences etc., here it i s sufficient to highlight two key findings. Figure 7 shows that the unemployment rates are lowest among primary school graduates and among those who have completed high school (or more) - there is a need to interpret the former trendjointly with low rural unemployment rates discussed previously though. The most important issue to note here is that since 1999, the graduates o f "professional, complementary or for apprentices" schools are the most likely to be unemployed. There is increasing concern about the prevalence o f vocational schools in the region", and the idea that these schools provide their graduates' better labor market opportunities does not hold up against empirical scrutiny (see also the employment/wage analysis inthe next section). Figure 8 shows that older individuals are much less likely to be unemployed compared to the young. Based on these trends, the need to deal with unemployment among the Roma (e.g., through making sure active and passive labor market programs reach Roma communities) seem to be immediately justified. Simple unemployment rates do not make the case for prioritizing l9Basedon the 2001 Labor Force Survey data. 2oFor Romania, a World Bank Education Policy Note (October 2002, ECA Human Development Sector Unit) recommends (i)improving the match between the courses offered by vocational and technical education and the existing and prospectivedemand of labor markets, (ii) developing a rolling program for updating equipment used in technical and vocational schools, and (iii) developing an information disseminationprogram to ensure that there is aproper balancebetweenstudents with academicbackground and students with technical andvocational skills. 90 women and elderly, but duration of unemployment and vulnerability to becoming unemployed needto be considered(which is done next). Even if women and elderly do not emerge as higher risk groups, it could be that even small and short duration unemployment among these groups is socially undesirable. Thus once can argue the need to target women and elderly regardlessof risk status, but it is important not to put these individuals in the same category with the younger individuals and Roma: effective policy interventionsare likely to differ widely betweenthese groups simply because of the differences in the target group size and geographicaldistribution. Figure5. Unemployment by Etbnicity, 1996-2001 Figure6. Unemploymentby gender and urbanirural residence,1996-2001 world Bnnhrtlcddidoru bsxdon LFS) (Horld BmkruBcddiUmrbswdmlFSJ 25 z \ B i z:IS E 8 10 5 - 0 19% 1997 1998 1999 2000 2w1 1996 199: 1998 1939 2wo 2001 Year Year I+ All +Romanian +- Hunganan*Roma -Other' 14All+Female +Male +Rural +Urban1 ~ ~~ Figure7, Unemploymentby schooling attainment,1996-2001 Figure8. Unemployment by age group, 1996-2001 (HorldBanlilaiiCilrul~llonshsDdonlF~ ( V d d Bml. stsflcdrulslionsb i d on LFS) 04 1 0 ........ - ._......._ ~ L .... ----t--+---t-l 19% 1997 1998 1999 2W0 2001 1996 1997 1998 1999 2000 2001 Year Year +So schooling 4All +Ages 15-24 t 4ges21.34 +Ages 35.44 +-Middle -Ages 45-54 +Ages 55-64 .3- 4ges 65 and over 91 Vulnerability to becomingunemployed The Labor Force Surveys (quarterly) contacted individuals o f ages 15 and over four times. The second survey took place three months after the first one, another follow-up nine months later that, and a final follow-up three months later. The empirical analysis reported in this section relies on survey data from 2000,2001 and 2002. The sample sizes are large, with around 140,000 observations per year. Benefiting from such large surveys, it i s possible to estimate the probability o f becoming unemployed (at least once) during the period under observation. One benefit o f relying on an analysis o f unemployment that occurred while under observation i s that one can then include individual-status information from wave 1 (Le., while not-unemployed) in the model.21 Separate Probit models are estimated for individuals aged 15 to 23 and others (because for the former group broader school attainment indicators are needed since some o f these individuals are still attending school). About 4.7 percent o f the 15 to 23 year olds became unemployed while under observation. The results described here are based o n specification 2 reported in Table 2. When all explanatory variables are at their mean values, the marginal effects on unemployment are as follows. Males are 1percent more likely to be unemployed. Marriage reduces chances o f unemployment by 2.3 percent. Schooling reduces the likelihood o f unemployment but the estimated coefficients are not statistically significant at 10 percent level and marginal effects are small. Relative to the "other" category, those who were working or who were students at the time o f the first survey are less likely to become unemployed. And finally, the urbadrural residence does not have a statistically significant effect on unemployment for this age group. Table 3 (specification 2) reports the estimates for those who are 24 years or older. While under observation, 2.3 percent o f this group became unemployed. After controlling for other explanatory variables the elderly are less likely to be unemployed, probability o f unemployment i s higher for males (0.7 percent), and for those who were students when they were first surveyed (0.7 percent). Marriage reduces chances o f unemployment (by 0.8 percent), so does working at the time o f first wave ,survey (0.5 percent). While schooling reduces the probability o f unemployment, effects are statistically significant only for high school and higher education, and the magnitude o f the effect i s around 1 percent (relative to those with n o schooling). In an alternative specification, "type o f employment (public, private or other) at the time o f first-wave survey'' information is also captured, which shows that unemployment probability i s higher by about 0.3 percent for those who were employed inthe private sector compared to those who were employed in the public sector. Thus despite restructuring in the public sector, the probability o f falling into unemployment is still significantly higher inthe private sector. Moving out of unemployment ---long term unemployment Inrecentyears, about 60percent o fthe unemployedare estimated to be inthis situation for over 9 months. The unemployment duration exceeded 12 months for slightly less than 50 percent o f the 21 Probit models are used for estimatingpredictorsof becomingunemployed. Since some individuals were under observation longer than others, and because the value of the dependent variable depends on this duration (if an individual is observed longer, he or she will be more likely to be recordedas unemployed), each model includes "months under observation" as an explanatory variable. Not surprisingly in each modelthis variablehas a positive and statistically significant effect onunemployment. 92 unemployed. Such a range i s typical o f transition countries.22 Not much else is known about the transition into and out-of unemployment, however. In particular gender, age, schooling and ethnicity differentials inthe duration o f unemployment are not w e l l understood. T he interval regression estimates o f the duration o f unemployment (months) also use the longitudinal labor force surveys, years 2000 to 2002.~' The key findings are as follows, based on a number o f alternative specifications reported in Tables 4 and 5. The set o f explanatory variables included inthe models is identical to that used in the previous section. Although separate models are estimated for those who are between ages 15 - 23 and others, the directions of the estimated coefficients are identical. The duration o f unemployment i s longest for males and for those who live inurban areas. Notice that previously we have shown that urbadrural residence i s not a statistically significant predictor o f becoming unemployed, but residence comes in strongly when duration i s considered. Interestingly marriage, schooling and status at the time o f the first wave survey (employed, student or other) do not have statistically significant influences on the duration o f unemployment -yet the previous section showed that (with the exception o f schooling) these are key determinants o f whether one becomes unemployed. More on thejob-search process The Social Protection chapter o f the Poverty Assessment has shown that unemployment benefits are progressive - although it is important to recognize that the main objective o f unemployment benefits i s not poverty reduction but easing transition between jobs. This section adds to that discussion. Unemployment benefits are administered by the National Agency for Employment (NAE). The benefits are paid for a maximum duration o f 12 months (depending on the duration o f contribution to the unemployment fund) at 75 percent o f minimum wage. The unemployment benefits are also paid new graduates without jobs and to those who have completed military service and could not find employment -for a six month period, the monthly benefit being set at 50 percent o f the minimumwage. The resources come from the unemployment insurance fund, although adjustments (in the form o f loans from Treasury) have occurred in the past. The fund experienced surpluses in 2001 and 2002. The NAE also administers the active measures program -whichincludescounselingservices,vocationaltraining,jobsearchassistance,jobfairs, 22 Milan Vodopivec, Andreas Worgotter and DhushyanthRaju. 2003. "Unemployment Benefit Systems in Central and EasternEurope:A Review of the 1990s". SocialProtectionDiscussionPaper No. 0310. The World Bank. 23 Three key methodologicalissues are the following. First, similar to the analysis of "predictors of falling into unemployment", the estimationsample is limited to those individuals who becameunemployedduring the period under observation. The surveys do ask about "unemployment duration" to the individuals, but inclusion of those who entered the survey as unemployedwould be still problematicbecause then one has to deal with selectionbias (the sample would over represent the long-termunemployed). Second, with the exceptiono f the 2002 questionnaire,the 1abor force surveys didnot inquire about "the time o f s tarting employment" for those individuals who were unemployed a t the time of the previous survey but found employment sometime after that (but before the follow-up survey that they revealed themselves as employed). Thus, hazardrate models that require precise information on the timing of transitions are not applicable. Here intervalregressioni s utilized for estimationinstead. Finally, the individuals who became unemployed while under observation but who were still unemployed at the time of the last follow-up survey contributeinformation to the maximumlikelihood function as observationswith upper-censoring. 93 temporary employment in public works etc. The spending on active programs tend to be small, representing about 2.5% o f the unemployment fundbudget inrecent years.24 Some insights on the effectiveness o f active labor programs in Romania arise from a quasi- experimental evaluation c onducted in2 002, which focused ont he impact o f four active 1abor market programs: training and retraining (TR), small business consultancy and assistance (SB), public works community job creation (PW), employment and relocation (ER).25The findings include the positive impact o f TR, SB and ER on participants' employment outcomes; and the positive impact o f SB and ER on participants' monthly earnings. The P W was not associated with positive outcomes, if anything the study revealed some negative impact on likelihood o f future employment which is not easy to justify (this i s likely to be due to unobserved characteristics o f the program participants versus the characteristics o f the control group which was constructed with some assumptions). The finding that public works programs do not "work" is not new.26 Even for the programs that the above study found positive impact, the cost-benefit ratio is not known. Thus available evidence raises questions about the payoffs to active labor programs in general and in public works in particular. In the Romanian context, pubic works programs are believed to be especially difficult to implement inrural areas. The NAE identifies "young people, women, long-term unemployed, disabled persons, over 45- year old persons, Roma and unemployed from other disadvantaged categories" as target groups for the active measures pr~gram.~'This is a comprehensive list, maybe too comprehensive inthat essentially only the middle aged males are excluded from the targeting scheme. Inpractice, such a targeting scheme means there i s no targeting. For example, while the Roma can benefit from the employment office functions, special activities towards Roma consist o f rarely heldjob fairs. Same goes for women: once or twice a year job fairs for women are unlikely to have much o f an impact. The lack o f a functioning targeting scheme is not necessarily a problem for all the groups listed above. Given the need to prioritize, some o f these individuals should not be targeted to start with. For the younger individuals, the main intervention is the payments to employers - for 12 months, a monthly payment equal to the minimum wage -for hiringnew graduates with open ended contracts. The employers have to keep these employees for at least three years, otherwise monetary reimbursement with interest is required. Wage subsidy programs are also suspect when it comes to their effectiveness: as emphasizedby the previous World Bank Poverty Assessment for Romania (1997), the firms might respond to incentives for hiring new graduates not by increasing overall number o f employees but by substituting the cheaper labor for more expensive labor. 24 For more information, see ; Joint Assessment o f Employment Priorities in Romania, October 2002, Ministryof Labour and Social Solidarity andEU. Basedon informationprovidedby the National Statistics Institute, active measures expenditures in 2002 were as follows in billions of lei: qualificatiodrequalification programs (4l),graduates hiring payments (201), hiring within the unemployment benefit period (1 18), bonuses for labor force mobility (16), encouraging employers to hire persons from disadvantagedcategories (62), public works expenditures (118), credits (1000). 25 Impact of Active Labor Market Programs inRomania. 2002. Abt Associates Inc. Manuscript prepared for the Ministry ofLabor and Social Protection. 26 More generally, a review of studies which evaluate the effectiveness of active-labor-market-programsin a number o f developed and developing countries reveals that in many cases cost-effectiveness of such ''rograms are disappointing (Dar and Tzannatos, 1999). National Strategy for Employment for 2002-2004. October 2001. Ministiy of Labour and Social Solidarity, National Agency for Employment. 94 The investigation o f job search methods in Romania, the role o f employment agencies and the characteristics o f those who benefit from those services also show that there i s room for improving targeting. For those who are unemployed and looking for a job in 2001 (ages 15 to 64), the leading methods usedfor "finding ajob or for a supplementary work to the present job --- in the last 4 weeks" is appeal to friends/relatives/colleagues/trade-unions (62.2 percent), direct approach to employers (54.7 percent), registering to the agencies for employment and vocational training (45.9 percent), answering to advertisements (22.7 percent), and publishing advertisements (7.7 percent). Each o f the following categories had less than 3 percent o f responses: made arrangements for a self-employment activity, registering at privatejob agencies, and other. Only 34.4 percent of these individuals used one single method for job search, and only 19 percent o f those who were registered to the agencies for employment used this as the sole job search method. The women are more likely to mention using employment agency in the job search process (49 percent as opposed to 45.9 percent o f all unemployed adults), so are those who are between ages 45 and 64 (52.6 percent). The younger unemployed adults (ages 15 to 23) are slightly less likely to mention the employment agency (43.1 percent). But, the Roma unemployed are significantly less likely to mention using employment agency in thej o b search process: Only 11.9 percent o f Roma revealed using employment agency during job search (as opposed to 45.9 percent o f all unemployed adults). Those who are estimated to be poor are also less likely to mention registering to the employment agency as ajob search method (41.2percent versus 50percent). 28 The other differences between the poor and the non-poor are minor, the poor being slightly more likely to rely on direct approach to employers and appeal to friends, relatives and less likely to respond to advertisements etc. The calculations for the poor are rough estimates, since poverty status i s estimated with the help o f HBS data. Having said that, even if there i s a large error margin, clearly the poor unemployed are not more likely to utilize employment agency as part of theirjob searchprocess. 4. The individuallevel determinantsof employment and wages The analysis o f employment and wages in Romania has direct implications for poverty. First, as shown previously, unemployment i s a rather weak predictor o f poverty: those who work in the informal sector are also likely to be poor, one question that comes to mindis "do informal sector participants earn less even afler controlling for their observed characteristics such as schooling attainment?". Other predictors o f employment and wages include non-income dimensions o f poverty (inparticular health and schooling). 28Since Labor Force Surveys do not containconsumptioninformation, those who are poor are estimated as follows. First, for those who are unemployed and between ages 15 and 64, probability of being poor is estimated (via probit model) using 2002 HouseholdBudget Survey data. The explanatoryvariables are age, agesquared, gender, marital status, household size, schooling, ethnicity and urbadrural residence indicator. Then, the estimated coefficients are used to predict probability of beingpoor in the 2002 Labor Force Survey data. Since about 45 percent ofunemployed adults are found to be poor in the HBS, 45 percent of those with highestprobability o fbeingpoor are designatedas "estimated poor" inthe LFS. 95 Using the 2002 Romania Living Conditions Survey, this section investigates the determinants o f being employed as well as determinants o f income from employment conditional on employment status.29 Only the key findings will be highlighted below. Employment The determinants o f employment is investigated by a number o f alternative models, reported in Tables 7 to 11. The basic set o f explanatory variables are: potential experience (and experience squared), gender, marital status, schooling attainment, ethnicity and urbadrural residence. Variations on this framework include additions o f health status indicators and estimating separate models for males and females as well as for those who reside in urban and rural areas. Probit model is used for estimation, where the dependent variable takes the value 1 to indicate employment and 0 otherwise. Probability o f employment increases by age, although the paste o f increase declines as an individual gets older. Marriage increases the probability o f employment by 9 percent, when all other explanatory variables are at their mean values (first column o f Table 7). T h i s i s a common finding in the literature, and it can be explained by both the selection argument (i.e., the individuals with desirable labor market characteristics are more likely to be married) or by the "beneficial causal effects o f marriage" argument. Schooling significantly increases the probability o f employment with the highest schooling category increasing employment chances by 43 percent (relative to no schooling). Ethnicity variables do not have statistically significant effects o n employment (with the exception o f c ategory other" which i s the residual category " after identifying major ethnic groups). Finally, urban residence decreases probability o f employment by 8 percent. Further insights can be gained by expanding on this reduced form specification. Note that there are other interesting points that can be made by comparing different models reported in Tables 7 to 11, the discussion here i s very selective. When separate models are estimated for urban and rural areas, we find that schooling have a much larger influence on probability o f employment in urban areas compared to rural areas (Tables 8 and 9). Marriage increases the probability o f employment o f males in urban and rural areas, but i t decreases probability o f employment o f females (by 14 percent) in rural areas - the effect is not statistically significant for urban females. A key finding emerges when an indicator or reporting chronic illness i s included in the model (Table lo). Even affer controlling for other factors, those reporting chronic illness are 30 percent less likely to work. The effect is larger for males at 34 percent, and for females at 25 percent. This finding supports the responses to direct questions on reasons for stopping work, and the magnitudes are quite tartl ling.^'The health-work relationship is further confirmed by models that use "self evaluatedhealth" inplace o f "chronic illness" variable (Table 11). A further look at the predictors o f chronic illness reveal that compared to Romanian ethnicity, the Roma are 8 percent more likely to report chronic illness and Hungarians are 4 percent more likely to *'It might be beneficial to estimate wage equations by taking into account employment-selection via Heckmantype models. Finding valid instrumental variables for the first-stage employment equation i s the challenge inthat case. 30 Inan employment equation, health indicators are endogenous and thus the causation i s open to question. But it seems unlikely that if health is treated as an endogenous variable such a sizable correlation disappears altogether. 96 report chronic illness (Table 6). Other parameter estimates reveal patterns that are similar to those observed in other countries: the elderly report more chronic illness, males report less, married report less, schooling attainment has a mixed story and reporting chronic illness is more likely in urban areas. Probably the last two findings are in part due to the fact that better educated individuals and those residing in urban areas have better access to health information and health services (thus conditional on having a chronic illness -which we do not observe -they may be more likely to report chronic illness). Among those who have chronic illness, the educatedare less likely to report that the illness limits their ability to work or affect daily activities (percent reporting "very much" as limitation steadily declines with schooling from 62 percent for uneducated to 26 for those who had more-than-high-school education). Two possible explanations for this association is (i)the work and daily activities of the uneducated (who are much more likely to be poor) are more challenging, so having a disease has a more catastrophic effect on their life, and (ii)the educated are more likely to have effective treatment for the chronic diseases that they report, so for the educated the restrictiveness of diseases is minimized. Future research should attempt to distinguish between these two explanations, since policy implications differ depending on the dominating effect. If the first explanation is "the" explanation, then the policy interventions might focus on improving worWlife conditions of the poor etc. Ifthe secondexplanation is more important, then one may need to focus on improvingthe access to quality care ofthe uneducated/poorindividuals. "Everyone, everyone goes to workfor food as daily laborers. WeJindemployment,day by day... "(Low income respondents, Alunis, rural. 2003) The payroll tax increases starting from 1998 can be expected to result inmore unemployment and more informal sector work, since labor becomes more expensive in the formal sector. Figure 9 shows that as payroll taxes increase, percentage o f employees decrease and percent self-employed & unemployed increase. This finding is verified by multivariate regressions at the individual level which control for changes inthe characteristics o f adult population over time (not reported). Figure9. PayrollTaxes, Employees, and Self-EmploymentNnemployment 1995-1002,Ages 15-64 (World Bank staff calculationsbased on HBS) 60 - 55 - 45 - * I 35 0 I 30 _. 1995 1996 1997 1998 1999 2000 2001 2002 Year '+- Payrolltax +Employee *Allselfemployed&unemployed 97 Incomefrom employment As shown by IMF (2003), the economy-wide average gross wage inRomania is low compared to other transition economies (at around US$ 140 in 2001), which strengthens Romania's competitiveness in labor intensive ind~stries.~' Untilrecently minimumwage influenced a small portion o f the labor force: in 2002 it was 1,750,000 lei (gross), which corresponded roughly to US$ 52. But in2003 gross minimumwage became 2,500,000 lei (or US$ 75). The predictors o f earnings are discussed inthe remainder o f the document from a poverty perspective, recognizing that the CEM will include a more comprehensive analysis o f wage setting policies. Various documents observe that real wages are higher inpublic (44 percent higher in 2002) than in the private sector.32 It seems evident that wages in long-term loss-making large public enterprises are not linked to productivity, but what additional insights can one obtain? For example, while it is known that wages are higher in the public sector, to what extent i s it because public sector employees have higher qualifications? Inother words, does the wage differential between public and private sectors remain after taking into account schooling and other characteristics o f individuals? For brevity, this discussion will focus on the direction o f effects (magnitudes can be derived from the estimated coefficients reported inTables 12 to 15). The basic insights from the baseline income regression (using the same set o f variables used for the baseline employment model described previously) are that males earn more than females, increased schooling has a large and consistently improving effect on wages, and those residing inurban areas eam more. The only ethnicity variable that is statistically significant is the identifier for Hungarians, who earn more than Romanians after controlling for other variables in the model. The Roma eam slightly less (as always, after controlling for other variables such as schooling etc.) but the effect i s not statistically significant. When public and private sector employment indicators are added to the model, the Roma are found to eam more (first c olumn o f T able 15). S econd and third columns o f the same table report estimation results separately for those who are employed in the public sector and for those who are employed inthe private sector. These estimates show that the Roma who have private sector employment eam more than others who report private sector employment, and in the public sector employment regression the Roma ethnicity dummy i s not statistically significant (but negative). The male-female wage differential is robust to alternative specifications -not only in models that control for schooling etc. but even after taking into account sector (publidprivate) and type o f employment (employee/employer/self-employed).33In fact, the gender gap prevails even if separate regressions are run for public sector and private sector (admittedly the gender gap in wages is smaller in the public sector). On the contrary, after controlling for schooling etc., the Roma earn as much as others do. Inother words, there is no evidence o f ethnic discrimination on 3' Romania: Selected Issues and Statistical Appendix. January 2003. IMFCountry ReportNo. 03/12. 32 See, for example, Monthly Statistical Bulletin, No 112002, NIS; Joint Assessment o f Employment Priorities inRomania, October 2002, Ministryo f Labour and Social Solidarity and EU. 33 Making the full-time versus part-time employment distinction also does not influence the findings. In 2002, 89 percent o f males and 87 percent of females considered their work as full-time. 98 the eamings front: if the health and schooling o f Roma were to improve, they might move out o f poverty without having to deal with the discrimination inlabor markets environment. The previous section had shown that the healthier are much more likely to be employed. Now, Tables 13 and 14, which add chronic illness and self-evaluated-health-status to the model, show that healthier individuals also eam significantly more than others even after controlling for other explanatory variables. Finally when state versus private employment indicators are added and the distinction i s made between employees, employers and self-employed, we find that controlling for all the variables in the model those who work inprivate sector eam less than their state counterparts, employers eam more than employees and self-employed eam less than employees. Reasons for the dismal eamings o f the self-employed (even after controlling for other variables) might include the lack of excess to resources (e.g., land or required capital investment), the inability to obtain credit, and lack o f access to markets and information. The publiciprivate differential in eamings remain even after controlling for schooling and other individual level characteristics. This has several implications. First, those who are employed in the public sector are unlikely to find private sector jobs with similar pay (at least inthe short run. Inthe long run the scaling down of inefficient loss-making public sector might give the private sector a boost through better use o f resources economy wide leading to increase in productivity and thus, under standard assumptions, increase in wages). Second, given that the public sector is scaling down, the relevant rates o f returns to schooling for many o f the younger individuals are those that prevail inthe private sector which i s underminedby the excessively large public sector. Such unnecessarily/unnaturally low retums to schooling might reduce parents' incentives to invest in their children's schooling, in tum jeopardizing long-term economic growth. Conclusions 1. From a poverty perspective, this chapter showed that informal sector workers require attention as muchas the unemployedrequire attention. These individuals are outside the scope o f labor legislation, without access to much needed credit markets etc. Regulations and formalization o f this sector would certainly be beneficial, especially if accompanied with significant reductions inpayroll taxes to encourage formalization. 2. The payroll taxes are extremely high inRomania. Such high taxation o f the formal sector workers and employers alike result in a n inflexible 1abor market 1eading to inefficiencies and contributing to high unemployment levels. This chapter showed the clear association between increases in payroll taxes and simultaneously occurring increases in the numbers of self- employed and unemployed individuals. It i s useful t o note that even if one simple-mindedly considers only the total amount collected, reductions in payroll taxes would not lead to decrease o f revenue at the same rate, because reductions inpayroll taxes would inpart be compensated by the expansion o f tax base - through shifts from informal to formal work and reduced unemployment; as well as through increased compliance in terms o f accurate wage reporting and tax payments. The Romanian Govemment i s aware of the problem, and in 2003 payroll taxes were reduced from 57 to 52 percent. But suchrelatively small decreases will not make-up for the large increases inprevious years (in 1997,payroll taxes were 35 percent o f gross wage). 99 3. The analysis of longitudinal survey data showed that private sector workers have much higher likelihood o f falling into unemployment than public sector employees. Even those workers who are laid-off as part o f public-sector restructuring process are not the worse-off segments o f the Romanian population since they benefited from relatively high public sector wages previously (incompanies implicitly subsidized by the rest o f the population) and since they often benefit from two year severance pay package deals. The question that i s critical is, what awaits these workers (and others who are often more vulnerable) inthe medium term? It could be a dynamic economy with a strong private sector that emerged through the downsizing o f the public sector and favorableiflexible labor market environment for business development. Or alternatively, it could be social protection programs in a weak economy with public sector restructuring that failed to be comprehensive or timely. 4. An increase inminimumwage i s often presentedas a pro-poor policy, even though it can lead to well known negative outcomes such as contribution to labor market rigidity, more expensive labor input, higher unemployment rates and higher informal sector participation. Inthe Romanian case, most o f the poor would not directly benefit from minimum wage increases because many o f them are self-employed - even though the poor who are unemployed would benefit from increases in minimum wages because o f the linkage between minimum wage and unemployment benefits, a (flat) unemployment benefit system can be implemented without indexing benefits to minimum wage. Thus, in this context, it i s difficult to support minimum wage increases on the grounds that they are pro-poor. 5. The priorities of theNational Employment Agency need to be established clearly, both in terms o f what programs to implement and to whom. As for active labor market programs, available evidence only helps to make the case to think twice before implementing them (especially for public works). Until and if favorable cost-benefit analysis results emerge, the National Employment Agency resources would be well advised to focus on administration o f the unemployment fund and on counseling services geared towards well defined target groups. New active labor market programs, or any modifications to the existing programs, should be designed as a pilot program with an experimental design which would allow one to have a legitimate control group and over-time comparisons. 6. The groups that are targeted for active labor market programs (to include job counseling, matching etc.) are too broad to be useful, and not always easily justified. Inparticular, empirical evidence suggests that it may be appropriate to identify "Roma" as a target group, but groupings such as "women" or "elderly" are too broad to be useful since many of the individuals who belong to these latter groups do not rank poorly in terms o f vulnerability to becoming unemployed or staying unemployed for a long duration o f time (and in terms o f poverty, as shown by the poverty profile). Ifanything, males are more likely to become unemployed and the duration o f unemployment i s also longer for them. I t would be useful to move away from defining "unemployed women" as a target group and not targeting them in practice, to defining groups like "women household heads who are unemployed" (since poverty profile suggests higher likelihood o f poverty for them) or "single mothers who are unemployed" (in fact more generally marriage reduces the likelihood o f becoming unemployed for both genders) etc. 7. Some o f the targeted groups for job search assistance and counseling (in particular younger individuals and Roma, as well as individuals who are estimated to be poor) are found to be significantly less likely to mention utilizing employment agency in their job search process. Thus there i s room for improvements inmaking NAEservices available to these groups. 100 8. This paper showed that predictors of the vulnerability to becoming unemployed are very different from predictors of the duration of unemployment. The long-term unemployment is an urban issue, being especially relevant for males. Other individual characteristics do not have a statistically significant impact on the duration o f unemployment. Thus, active labor programs aiming to reduce long-term unemployment should concentrate on urban centers. 9. High unemployment rates prevail for the youth. The key thing to note here i s that even though the Romanian Government implements a special wage subsidy program for new- graduates, the unemployment rates are still high and thus the program is not a big success. Yet it i s another inefficient labor market distortion: many new graduates would have found employment inthe absenceofthe subsidy, andrequirements ofkeeping these employees for years addto labor market rigidities. 10. If one leaves out those who are retired, it is striking that about 23 percent of the remaining individuals stopped work because o f sickness or invalidity. Consistent with this, multivariate analyses showed that health status is a key determinant o f employment and earnings. Furthermore, the uneducated (who are more likely to be poor) are more likely to reveal that illness has a detrimental effect on their ability to work. Obviously investments in health are important and justifiable for their own right, but there are also clear economic benefits to healthiness, probably more so for the poor than the non-poor. 11. A similar conclusion emerges for the education sector, although with at least two complications. First, while eaming regressions document robust increasing returns to schooling, unemployment rate is highest among the vocational school graduates. There seems to be a mismatch between the demand for and supply o f vocational school graduates. This imbalance can be remedied by either reducing number o f vocational school graduates or by modifying the curriculum t o increase the demand for them. T here are other possibilities as well, at 1east in theory it could be that students who enroll in vocational schools have unobserved (individual and/or household) characteristics that are in part responsible for the unfavorable 1abor market outcomes. This analysis flagged and provided some insights o n the issue, w hich needs t o be further explored. Second, the differential rates o f returns to schooling inthe public sector and in the private sector (which does most o f the hiring, but with lower pay) may reduce poor households' incentives to invest intheir children's schooling if they do not consider public sector employment as an option upon graduation (this could happen ifpoor households lack connections for public sector openings that are becoming scarce, or if poor households are more risk averse and thus even if the probability o f public sector employment is similar for poor children and others, poor parents may choose not to take the risk). Demand side interventions may be strengthened to prevent this from happening, although such schemes need to focus on the poor only (as opposed to universal child subsidies that are easier to implement but highly ineffective in transferring resources to the poor inreturn for children's school attendance). 12. The gender inequality in eamings cannot be explained by observed individual characteristics such as schooling, part-time versus full-time employment, or sector and type o f employment. Thus gender discrimination emerges as a real possibility inRomania, the pathways to which remain to be uncovered. 13. Finally, the data and monitoring suggestions are as follows. (i) Labor Force Surveys do not collect information on anything that can be usedas a proxy for household wealth. While it seems unrealistic to add a consumption module to this survey, the addition o f a section on household possessions would. be beneficial. 101 (ii) Labor Force Surveys are being implemented since 1996, but only once wage information was collected as part o f these surveys. Lack o f wage data in a labor force survey i s a major shortcoming, and i s relatively easy to fix. (iii)Labor Force Surveys do not contain any question on health status. While the inclusion of a detailed set o f questions on health may not be feasible, selected few health questions (as in HBS) would significantly increase the value of these surveys. (iv) All household surveys fielded by the NIS (HBS, LFS, AMIGO etc.): It would be beneficial to include some questions on parental schooling, subjective evaluation o f parental health (and parental mortalitylage-of-death), and parental wealth. The addition o f such questions would serve two purposes. First, they would enable one to have a rough idea about social mobility trends. Second, such variables can be used to single-out causal relationships since they are the standard instrumental variables to explain eamings, health etc. (v) To our knowledge, none of the surveys fielded by the NIS contain questions on migration. At the least retrospective questions can be added to the household surveys, inquiringabout where the individual was bom, where helshe lived at age 12, and where helshe lived 5 years ago etc. (vi) There would be significant payoffs to achieving "mergeability" across various different NIS surveys. Such a suggestion might be unrealistic for many developing countries, but Romania is an exception: the Romanian Statistics Institute i s extremely capable and much o f the required infrastructure (e.g., unified sampling unit codes, survey personnel who are permanent inthe field etc.) are already inplace. 102 Table 1. Occupation of those between ages 15 to 64, as revealed by themselves (source: HBS surveys 1995 to 2002). As percentageo f individuals inthe 15-64 age group Year Employee Employer Self-employed non-agriculture Self-employedagriculture Unemployed' Other (pensioner, student, housewife, dependent, 33.58 34.48 35.20 34.93 35.04 36.00 36.93 37.23 military service etc.) Total 100 100 100 100 100 100 100 100 *Note that the unemploymenr that. 103 Table 2. Predictorsof becoming unemployed, for those between ages 15 and 23. Probitmodel is estimatedwhere the dependent variable takes the value 1 ifthe individualbecame unempl red during the period under observation, 0 otherwise. The reported figures are marginal effects, and the Itl-statis :s are in parentheses. Specification 1 Specification 2 I Gender (1 ifmale, 0 otherwise) ,013 ,010 (4.30) (3.44) Marriage (1 ifmarried, 0 otherwise) -.018 -.023 (2.94) (5.01) No schooling _____ _ _ _ _ _ Primary -.022 -.004 (2.37) (0.38) More than Primary -.032 -.002 (2.21) (0.23) Status at wave 1 I Other (military, housewife or other) Working -.021 (5.54) Student -.061 (10.64) Urban residence(1 ifurban, 0 otherwise) -.009 2.7 * 10'~ (2.92) (0.01) Time under observation(months) ,005 (18.3) Log likelihood -1881.1 -1822.9 Number of observations 11200 11200 104 Table 3. Predictors of becomingunemployed, for those betweenages 24 and 64. Probitmodel is estimatedwhere the deoendent variable takes the value 1ifthe individualbecameunemnlovedduringthe oeriod under observation, 0 - . otherwise. The reported figures a marginaleffects, andthe Itl-statisticsare in parentheses. I I Specification I I Specification2 ISpecification3 I I ,002 ,003 ,003 (7.03) (7.93) (8.32) Age squared (1100) -.004 -.004 -.004 (8.96) (9.83) (10.2) Gender (1 ifmale, 0 otherwise) ,006 ,007 ,007 (7.17) (7.97) (8.13) Marriage(1 ifmarried, 0 -.008 -.008 -.007 otherwise) (7.71) (7.20) (7.02) Schooling No schooling Primary -.003 -.002 -.003 (1.01) (0.70) (0.75) Middle -.006 -.005 -.005 (1.95) (1.62) (1.63) Professional -.006 -.005 -.004 (2.02) (1.61) (1.45) Highschool -.009 -.008 -.007 (2.90) (2.53) (2.32) Postsecondary / foremen -.011 -.010 -.009 (4.78) (4.36) (3.86) Higher education -.011 -.010 -.009 (4.97) (4.41) (3.91) Status at wave 1 Other (military,housewife or ___-- other) ----- I ----- -.005 I Working -.010 (4.36) (4.46) Student ,008 ,007 (1.90) (1.86) Type of enterpriseworked Other -____ _____ Public ___-_ __ __ __ __ __ -.002 (0.96) Private _____ _____ ,005 (3.10) Urbanresidence(1 ifurban, 0 ,0006 -.0001 ,0007 otherwise) (0.67) (0.12) (0.76) Time under observation ,0017 ,0018 ,0018 (months) (22.4) (22.6) (22.7) Loglikelihood -4260.0 -4245.8 -4228.2 Number of observations 45231 45231 45231 105 Table 4. Predictors of unemployment duration, for those between ages 15 and 23. Interval regression estimates are presented, where the dependent variable indicates number of months unemployed. The ltl- statistics are inparentheses. Specification 1 Specification 2 Gender (1 ifmale, 0 otherwise) ,803 ,838 (1.77) (1.79) Marriage (1 ifmarried, 0 otherwise) -1.21 -1.09 (1.03) (0.91) Schooling No schooling _____ _ _ _ _ _ Primary -2.66 -2.72 (1.58) (1.59) More than Primary -1.74 -1.78 (1.16) (1.17) Status at wave 1 Other (military, housewife or other) Working -.268 (0.44) Student ,158 (0.25) Urban residence(1 ifurban, 0 otherwise) 1.46 1.36 (3.24) (2.87) Constant 12.0 12.2 (7.89) (7.75) Log likelihood -274.1 -213.7 Number of observations: total 524 524 right-censored 389 389 interval 135 135 106 Table 5. Predictors o f unemployment duration, for those between ages 24 and 64. Intervalregression estimates are presented, where the dependent variable indicates number o f months unemployed. The ltl- statistics are inparentheses. Specification 1 Specification 2 Specification 3 ,182 ,191 ,196 (1.23) (1.27) (1.30) Age squared(1100) -.246 -.257 - 262 (1.32) (1.35) (1.38) Gender (1 ifmale, 0 otherwise) ,884 .a79 ,860 (2.57) (2.51) (2.45) Marriage (1 ifmarried, 0 ,531 .536 ,556 otherwise) (1.54) (1.54) (1.60) Schooling No schooling _ _ _ _ _ _____ _--__ Primary -.379 -.357 -.344 (0.31) (0.29) (0.28) Middle 1.38 1.39 1.41 (1.36) (1.36) (1.39) Professional ,880 ,877 ,911 (0.89) (0.88) (0.92) High school ,542 ,533 ,584 (0.55) (0.54) (0.59) Post secondary/ foremen 1.03 1.03 1.09 (0.68) (0.68) (0.71) Higher education ,209 ,173 ,303 (0.17) (0.14) (0.25) Status at wave 1 Other (military, housewife or _____ _____ other) Working ,093 -.523 (0.22) (0.66) Student ,491 ,485 (0.42) (0.41) Type of enterprise worked Other -____ Public ,619 (0.73) Private ,676 (0.93) Urbanresidence (1 if urban, 0 1.45 1.46 1.48 otherwise) (4.27) (4.18) (4.23) Constant 5.46 5.17 5.02 (1.84) (1.70) (1.65) Log likelihood -572.2 -572.1 -571.7 Number of observations: total 1020 1020 1020 right-censored 702 702 702 interval 318 318 318 107 Table 6. Predictors o f reporting chronic disease for ages 40 and older. Probitmodel is estimatedwhere the dependentvariable takes the value 1 ifthe individual reporteda chronic disease, 0 otherwiseand the deuendentvariable for the secondmodel takes the value 1 ifthe individual had an acute symptomin eported figures are marginal effects, and the Itl-statisticsare parentheses. Chronic Disease Means and Standard Deviations All Female .028 5 9 . 3 ( 7 . 4 8 ) ( 3 . 9 2 ) ( 6 . 6 2 ) L12.11 Age squared / 100 - ,011 - , 0 0 6 - ,015 ~ ~~~~ Gender (1 ifmale, 0 otherwise) - , 0 9 6 .457 ( 1 0 . 3 6 ) r.4981 Marital statusof householdhead - .056 - ,039 - .067 .715 (1ifmamed, 0 otherwise) ( 5 . 3 7 ) r.4511 Schooling attainment None _ _ _ _ _ I ----- r . 1851 ,036 Primary . 0 1 1 - ,039 , 0 2 3 .242 ( 0 . 4 5 ) ( 0 . 8 6 ) 0.791 [ ,4281 Gymnasia1 ,046 - ,028 ,074 ,270 ( 1 . 7 9 ) (0.60) 2 . 3 9 ) 1.4441 Professional , 0 6 9 - .002 .111 .174 ( 2 . 4 7 ) ( 0 . 0 3 ) 3 . 0 2 ) [ ,3791 High School ,020 - ,027 , 0 3 1 ,167 ( 0 . 7 3 ) ( 0 . 5 6 ) 0 . 9 1 ) [ . 3741 Post-Secondary/ foremen ,034 - ,018 ,050 ,048 ( 1 . 0 5 ) ( 0 . 3 6 ) ( 1 . 0 7 ) [ . 2131 Higher - , 0 7 1 - ,092 - ,067 ,063 ( 2 . 3 2 ) ( 1 . 9 0 ) ( 4 . 8 6 ) [ . 2431 Ethnicity Romanian ,909 r.2871 Hungarian ,039 .oa7 - .004 .066 ( 2 . 2 1 ) ( 3 . 4 3 ) ( 0 . 1 5 ) [.2481 Roma ,082 , 1 2 0 ,037 .010 ( 1 . 7 7 ) ( 1 . 8 4 ) ( 0 . 5 7 ) [ . 0981 Other - ,043 - ,020 - 0 . 0 6 ,015 ( 1 . 1 8 ) ( 0 . 3 8 ) ( 1 . 1 9 ) [.1201 Urbanlruralresidence ,132 .465 (1 if urban, 0 otherwise) ( 1 2 . 9 2 ) ( 7 . 5 0 ) ( 1 0 . 3 8 ) r.4991 Log-likelihood -8048.5 -3544.4 -4493.2 Number o f observations 13435 13435 108 Table 7. Predictors of employment for ages between 15 and 64. Probamodel is estimatedwhere the dependent variable takes the value 1 ifthe individualis working, 0 otherwise The .ndthe ltl-statisticsare inparentheses. Employment Means and Standard Deviations All Male Female Potential Experience ,042 . 0 4 1 .040 2 3 . 1 f32.1) ( 2 2 . 9 9 ) ( 2 1 . 9 6 ) Potential Experience2 / 100 - ,088 - , 0 8 8 -.084 ( 2 5 . 1 ) ( 2 3 . 0 ) Gender (1 ifmale, 0 otherwise) ,192 _ _ _ _ _ ,489 (22.26) [ ,5001 Marital status of household head , 0 9 0 ,226 - ,026 ,632 (1 ifmamed, 0 otherwise) ( 8 . 5 9 ) ( 1 4 . 5 ) ( 1 . 8 5 ) L.4821 Schoolingattainment None ,012 [ . 1101 Primary ,189 , 2 4 1 ,104 ,085 ( 4 . 5 4 ) ( 4 . 6 8 ) (1.78) L.2781 Gymnasia1 - ,003 ,084 - ,090 ,268 ( 0 . 0 7 ) ( 1 . 4 7 ) ( - 1 . 6 3 ) L.4431 Professional ,179 ,200 ,111 .214 ( 4 . 3 0 ) ( 3 . 5 9 ) ( 1 . 9 0 ) [ ,4101 High School .240 ,244 ,185 (5.80) ( 4 . 4 6 ) ( 3 . 2 4 ) r .,298 4571 Post-secondary/ foreman ,260 ,185 .305 .045 ( 6 . 0 6 ) ( 3 . 2 6 ) ( 4 . 8 9 ) [ . 2071 Higher ,429 ,360 ,454 , 0 7 7 ( 1 1 . 7 9 ) ( 8 . 0 4 ) ( 8 . 0 3 ) [ ,2671 Ethnicity Romanian .__._ _ _ _ _ _ ----_ ,915 [ ,2791 Hungarian ,027 ,005 .054 ,059 ( 1 . 4 8 ) (0.21) ( 2 . 2 2 ) L.2371 Roma -.038 - ,017 - ,068 ,015 ( 1 . 0 4 ) ( 0 . 3 5 ) ( 1 . 7 4 ) [.1201 Other - ,119 - ,123 - ,103 , 0 1 1 ( 2 . 8 6 ) ( 2 . 1 2 ) ( 1 . 8 3 ) [ ,1051 Urbanlrural residence - ,083 - . 1 4 1 - ,030 ,547 ( 8 . 8 8 ) ( 1 0 . 9 3 ) ( 2 . 3 1 ) I . 4 9 8 1 Log-likelihood -9273.3 - 4 3 4 1 . 8 - 4 7 6 2 . 0 Number of observations 16437 8039 8398 16437 109 Table 8. PredictorsofEmploymentbetweenages of 15 and64 (Urban). 1 Probit model is estimated where the dependent variable takes the value 1 if the individual i s working, 0 otherwise. The reported figures are marginal effects ndthe ltl-statistics are inparentheses. Employment Means and Standard Deviations Male Female Potential Experience ,072 ,072 , 0 6 9 2 1 . 1 ( 3 1 . 7 4 ) ( 2 2 . 5 ) ( 2 2 . 0 7 ) L13.61 Potential Experience2 / 100 - ,176 - ,176 - ,171 (34.45) ( 2 5 . 1 ) ( 2 3 . 6 ) Gender (1 ifmale, 0 otherwise) ,133 __... ,482 ( 1 0 . 6 3 ) [.5001 Marital status of householdhead , 0 9 0 .172 ,020 ,625 (1ifmarried, 0 otherwise) ( 5 . 7 0 ) ( 7 . 0 1 ) ( 0 . 9 7 ) 1.4841 Schooling attainment None .... ,006 [.0801 Primary . z o o ,174 ,209 .030 ( 2 . 0 0 ) ( 1 . 3 1 ) ( 1 . 4 1 ) [ . 1701 Gymnasia1 ,259 ,197 ,318 ,187 ( 2 . 8 2 ) ( 1 . 5 7 ) ( 2 . 3 9 ) [ . 3901 Professional . 4 3 1 ,360 , 4 7 9 ,204 ( 5 . 0 9 ) ( 3 . 0 5 ) ( 3 . 8 8 ) L.4031 High School , 4 7 1 , 3 8 1 ,529 ,379 ( 5 . 3 4 ) ( 3 . 1 5 ) ( 4 . 2 4 ) 1.4851 Post-Secondary/ foremen .449 ,330 ,547 ,068 ( 5 . 9 7 ) ( 3 . 0 6 ) ( 5 . 1 3 ) 1.2511 Higher , 5 5 6 ,459 ,633 ,126 ( 8 . 3 1 ) ( 4 . 9 6 ) ( 6 . 4 7 ) [ ,3321 Ethnicity Romanian _ _ _ _ _ ____. ._.._ ,929 [. 2571 Hungarian , 0 3 1 ,004 ,067 ,055 ( 1 . 1 3 ) (0.11) ( 1 . 7 2 ) 1.2271 Roma - ,026 ,004 - ,085 , 0 0 8 ( 0 . 3 4 ) ( 0 . 0 4 ) ( 0 . 7 6 ) [.0911 Other , 0 6 6 .029 ,123 , 0 0 8 ( 0 . 9 0 ) ( 0 . 2 9 ) ( 1 . 1 6 ) [ ,0891 Log-likelihood -4359.8 - 2 0 0 6 . 2 -2322.2 Number o f observations 8983 4328 4655 8983 110 Table 9. Predictorsof Employmentbetweenages of 15 and 64 (Rural). Probit modelis estimated where the dependentvariable takes the value 1ifthe individual is working, 0 otherwise.The reportedfigures i :maiginal effects, andthe :tl-statisticsare inparenthes Employment Means and Standard Deviations Male Female Potential Experience ,036 .035 , 0 3 2 2 5 . 5 119.7) ( 1 4 . 7) ( 1 3 . 4 ) 115.81 ~ Potential Expenence2/ 100 - ,064 - , 0 6 6 - .059 ( 1 4 . 5 ) (12 . 7 ) Gender (1 ifmale, 0 otherwise) ,274 ,498 ( 2 1 . 8 ) [ .5001 Marital status of household head ,039 .214 - ,136 .640 (1 ifmamed, 0 otherwise) ( 2 . 5 8 ) ( 1 0 . 0 ) ( 6 . 8 8 ) [ ,4801 Schooling attainment None ,019 r.1381 Primary ,153 ,218 ,053 , 1 5 1 ( 3 . 1 6 ) ( 3 . 6 3 ( 0 . 8 6 ) L.3581 Gymnasia1 ,053 ,126 - . 032 .366 ( 1 . 0 9 ) ( 1 . 9 5 ( 0 . 5 3 ) [ .482] Professional , 2 1 7 ,227 .14 8 ,227 ( 4 . 4 2 ) ( 3 . 5 7 ( 2 . 2 1 ) [ ,4191 High School ,273 ,239 .236 ,200 ( 5 . 6 2 ) ( 3 . 8 8 ( 3 . 5 9 ) ,4001 Post-Secondary/ foremen , 3 1 0 .135 ,463 , 0 1 7 ( 5 . 0 0 ) ( 1 . 6 7 ) ( 5 . 0 3 ) [ ,1311 Higher , 4 1 9 ,320 .449 ,019 ( 7 . 4 0 ) ( 4 . 9 3 ) ( 5 . 0 7 ) [ .1361 Ethnicity Romanian .____ _ _ _ _ _ ,897 L.3041 Hungarian , 0 4 9 ,016 , 0 6 9 ,065 ( 1 . 9 7 ) ( 0 . 4 8 ) ( 2 . 1 4 ) [ ,2471 Roma ,025 .023 -.016 ,022 ( 0 . 5 7 ) ( 0 . 4 1 ) ( 0 . 2 8 ) [ .1471 Other - , 1 9 5 - , 2 0 7 - ,179 , 0 1 5 ( 3 . 7 7 ) ( 2 . 8 1 ) ( 2 . 8 1 ) [ .1231 Log-likelihood -4386.3 - 2 0 9 7 . 1 - 2 1 7 5 . 0 Numberof observations 7454 3 7 1 1 3743 7454 111 Table 10. Predictors o f employment for ages between 15 and 64. Probit modelis estimated where the dependentvariable takes the value 1ifthe individual is worlung, 0 1 othernise. The reportedfigures are marginaleffects, and the t -statistics are inparenthest Employment Means and Standard Deviations Male Female Potential Experience ,045 ,043 2 3 . 1 ( 2 4 . 6 ) ( 2 3 . 5 ) L14.81 Potential Experience' / 100 - ,089 - ,090 - ,085 Chronic Illness (1 ifyes, 0 otherwise) I ( 2 5 . 3 ) ( 2 3 . 0 ) - . 2 9 5 - ,342 - ,250 .184 ( 2 0 . 1 ) ( 1 7 . 0 ) [.3881 Gender (1 ifmale, 0 otherwise) I ,185 _ _ _ _ _ _ _ _ _ _ , 4 8 9 ( 2 1 . 1 ) L ,5001 Marital status of household head ,068 ,212 - ,050 ,632 (1 ifmamed, 0 otherwise) ( 6 . 3 5 ) ( 1 3 . 3 ) ( 3 . 5 3 ) L.4821 Schooling attainment None .012 [ ,1101 Primary , 1 3 1 ,196 ,040 ,085 ( 3 . 0 2 ) ( 3 . 5 2 ( 0 . 6 8 ) [.27a1 Gymnasia1 - ,053 ,028 - ,133 ,268 ( 1 . 2 4 ) ( 0 . 4 6 ( 2 . 4 0 ) [ ,4431 Professional ,144 . 1 5 5 ,084 . 2 1 4 ( 3 . 3 3 ) ( 2 . 6 3 ( 1 . 4 3 ) L ,4101 High School ,196 ,197 .143 ,298 ( 4 . 5 6 ) ( 3 . 3 9 ( 2 . 4 8 ) [ ,4571 Post-secondary1foreman ,239 ,156 , 2 9 1 ,045 ( 5 . 3 8 ) ( 2 . 5 8 ) ( 4 . 5 6 ) L.2071 Higher ,407 ,337 ,426 ,077 ( 1 0 . 6 ) ( 6 . 9 7 ) ( 7 . 2 7 ) L.2671 Ethnicity Romanian .____ ,915 L.2791 Hungarian ,037 ,019 .063 ,059 ( 2 . 0 3 ) ( 0 . 7 6 ) ( 2 . 5 3 ) f.2371 Roma - ,043 - ,018 - ,097 ,015 ( 1 . 1 6 ) ( 0 . 3 7 ) ( 1 . 9 2 ) [ ,1201 Other - ,142 - ,161 - , 1 1 3 , 0 1 1 ( 3 . 3 7 ) ( 2 . 7 0 ) ( 2 . 0 0 ) [ . 1051 U r b a n h r a l residence - ,066 - ,127 - ,011 ,547 ( 9 . 6 8 ) ( 0 . 8 7 ) [.,4981. Log-likelihood -8928.3 - 4 1 3 6 . 3 - 4 6 1 2 . 1 8039 8398 16437 112 Table 11. Predictors of employment for ages between 15 and 64. 1 Probit model is estimatedwhere the dependentvariable takes the value 1 ifthe individual is working, 0 otherwise. The reported figures are marginal effects ndthe Itl-statisticsare inparentheses. Employment Means and Standard Deviations All Male Female Potential Experience .045 ,045 ,042 2 3 . 1 ( 3 3 . 4 ) ( 2 2 . 6 ) L14.81 Potential Experience2I 100 - ,089 - ,091 - ,084 ( 3 3 . 5 ) ( 2 5 . 3 ) ( 2 2 . 6 ) Self-Evahated Health ,285 Very Good [ ,4511 ,019 , 0 3 6 ,006 ,450 Good ( 3 3 . 4 ) ( 2 . 3 6 ) ( 0 . 3 5 ) L.4971 -.126 - .136 - ,107 ,182 Satisfactory ( 8 . 4 4 ) ( 6 . 3 7 ) ( 5 . 3 6 ) L.3861 - . 3 7 5 - . 4 5 5 - . 3 0 0 .083 Bad I Very Bad ( 2 1 . 2 ) ( 1 7 . 1 ) ( 1 3 . 2 ) [.2761 Gender (1 ifmale, 0 otherwise) , 1 8 5 ,489 [ .SO01 Marital status of householdhead , 0 6 8 ,210 - ,046 , 6 3 2 (1 ifmarried, 0 othemlse) ( 6 . 2 8 ) ( 1 3 . 1 ) ( 3 . 2 6 ) L.4821 Schooling attainment None ,012 [ . I 1 0 1 Primary ,123 ,169 ,050 ,085 ( 2 . 8 0 ) ( 2 . 9 5 ) ( 0 . 8 5 ) [ . 2781 Gymnasia1 - ,068 - ,003 - ,134 ,268 ( 1 . 5 7 ) ( 0 . 0 4 ) ( 2 . 4 1 ) L.4431 Professional , 1 2 6 ,124 ,075 ,214 ( 2 . 9 0 ) ( 2 . 0 8 ) ( 1 . 2 6 ) [ ,4101 High School ,174 , 1 6 0 ,136 , 2 9 8 ( 4 . 0 2 ) ( 2 . 7 1 ) ( 2 . 3 2 ) [ .457] Post-secondary1foreman .212 , 1 0 8 ,276 ,045 ( 4 . 6 6 ) ( 1 . 7 3 ) ( 4 . 2 9 ) L.2071 Higher ,392 ,315 .416 .077 ( 9 . 9 2 ) ( 6 . 2 3 ) ( 7 . 0 0 ) [ ,2671 Ethnicity Romanian _ _ _ _ _ __... __.._ .915 ,2791 Hungarian , 0 2 9 ,008 ,057 ,059 ( 1 . 5 9 ) ( 0 . 3 0 ) ( 2 . 2 8 ) .2371 Roma - .050 - ,038 - ,093 ,015 ( 1 . 3 4 ) ( 0 . 7 6 ) ( 1 . 8 3 ) .1201 Other - ,136 - ,154 - ,111 , 0 1 1 ( 3 . 2 4 ) ( 2 . 5 9 ) ( 1 . 9 6 ) .1051 Urbanlruralresidence - ,068 - ,129 - ,015 ,547 (1 ifurban, 0 otherwise) ( 7 . 0 9 ) ( 9 . 7 5 ) (1.11) L.4981 Log-likelihood -8320.8 -4110.4 - 4 6 2 8 . 1 Number of observations 8039 8398 16437 113 Table 12. Predictors o f (log) income from employment (ages o f 15 and 64). Itl-statistics are in \ parentheses. ~~ Log (Income from Salanes, Agnculture & Self-Employment Activities, In-kindIncome) Means and StandardDeviations All Male Female PotentialExperience , 0 4 0 ,037 .044 22.0 ( 1 3 . 8 ) ( 9 . 2 2 ) ( 1 0 . 1 ) L11.31 Potential Experience2/ 100 - ,084 - , 0 8 1 - ,088 ( 9 . 8 1 ) ( 8 . 8 1 ) Gender (1 ifmale, 0 otherwise) .208 _ _ _ _ _ _ _ _ _ _ ,587 ( 1 2 . 6 ) i.4921 Marital status of householdhead .124 ,214 , 0 0 8 ,725 (Iifmarried,0otherwise) ( 6 . 3 3 ) ( 7 . 6 0 ) ( 0 . 2 9 ) [ ,4461 Schoolingattainment None _ _ _ _ _ ,004 [. 0631 Primary ,189 ,205 ,082 . 0 4 1 ( 1 . 4 2 ) ( 1 . 1 8 ) ( 0 . 4 0 ) [ ,1991 Gymnasia1 , 5 2 0 , 3 8 8 , 7 5 1 ,137 ( 3 . 9 9 ) ( 2 . 2 9 ) ( 3 . 7 0 ) [ .344] Professional ,907 , 7 6 7 1 . 1 4 ,257 ( 6 . 9 3 ) (4 .51) ( 5 . 5 7 ) [ ,4371 High School 1 . 0 3 ,868 1 . 2 9 ,362 ( 7 . 8 5 ) ( 5 . 0 9 ) ( 6 . 3 0 ) [ .481] Post-Secondary1foremen 1 . 2 4 1 . 0 8 1 . 5 0 .063 ( 9 . 2 3 ) ( 6 . 1 5 ) ( 7 . 2 1 ) [ ,2431 Higher 1 . 4 8 1 . 3 3 1 . 7 4 ,135 ( 1 1 . 2 ) ( 7 . 6 9 ) ( 8 . 4 5 ) [ . 3421 Ethnicity Romanian --___ _ _ _ _ _ _____ ,919 [ . 2731 Hungarian ,135 ,128 ,152 .062 ( 4 . 1 1 ) ( 2 . 7 9 ) (3.34) [ .242] Roma - .039 - ,076 ,002 ,010 ( 0 . 4 6 ) ( 0 . 7 5 ) ( 0 . 0 1 ) [ ,0991 Other , 1 2 9 ,148 ,104 .009 (1.56) ( 1 . 3 9 ) ( 0 . 7 9 ) [ ,0951 Urbadrural residence , 4 1 3 ,403 ,428 , 6 1 4 (1 ifurban, 0 otherwise) ( 2 2 . 7 ) ( 1 6 . 5 ) ( 1 5 . 8 ) [ ,4871 Number of observations 7740 4544 3196 7740 114 1Table 13. Predictors o f (log) income from employment (ages o f 15 and 64). It[-statistics are in parentheses. Log(Income from Salaries, Agriculture & Self-Employment Activities, In-kind Income) Means and Standard Deviations All Male Female Potential Experience , 0 4 1 .037 ,044 2 2 . 0 ( 1 3 . 9 ) ( 9 . 2 3 ) ( 1 0 . 1 ) L11.31 ~ Potential Expenence2/ 100 - .083 - ,080 - ,087 ( 1 3 . 3 ) ( 9 . 7 2 ) ( 8 . 7 2 ) Chronic Illness (1 ifyes, 0 otherwise) - ,092 - . 080 - ,115 ,098 (3.40) ( 2 . 0 8 ) ( 3 . 1 0 ) L.2981 Gender (1 ifmale, 0 otherwise) ,205 ,587 (12.41) L.4921 Marital status of householdhead , 1 2 1 ,214 ,003 ,725 (1 ifmarried, 0 otherwise) 16.221 ( 7 . 6 0 ) ( 0 . 1 1 ) [ ,4461 Schooling attainment None _.._. ,004 [. 0631 Primary ,185 ,208 ,062 , 0 4 1 ( 1 . 3 9 ) ( 1 . 2 0 ) ( 0 . 3 0 ) [ ,1931 Gymnasia1 ,520 ,393 , 7 3 9 , 1 3 7 ( 3 . 9 9 ) ( 2 . 3 2 ) ( 3 . 6 4 ) L.3441 Professional ,908 ,772 1 . 1 3 ,257 ( 6 . 9 4 ) ( 4 . 5 4 ) ( 5 . 5 3 ) [ ,4371 High School 1 . 0 3 ,872 1 . 2 8 ,362 ( 7 . 8 5 ) (5.12) ( 6 . 2 5 ) [ ,4811 Post-Secondary/ foremen 1 . 2 4 1 . 0 8 1 . 4 9 ,063 ( 9 . 2 6 ) ( 6 . 2 0 ) ( 7 . 1 8 ) [ ,2431 Higher 1 . 4 8 1 . 3 3 1 . 7 3 ,135 ( 1 1 . 2 ) ( 7 . 7 2 ) ( 8 . 4 1 ) [.3421 Ethnicity Romanian __.__ _ _ _ _ _ _ _ _ _ _ ,919 [.2731 Hungarian ,135 ,127 ,154 ,062 ( 4 . 1 2 ) ( 2 . 7 9 ) ( 3 . 3 9 ) L.2421 Roma - .035 - , 0 7 1 ,004 ,010 ( 0 . 4 2 ) ( 0 . 7 0 ) ( 0 . 0 3 ) [ . 0991 Other ,123 , 1 4 1 , 1 0 1 ,009 ( 1 . 4 8 ) ( 1 . 3 2 ) ( 0 . 7 6 ) [ ,0951 ~ Urbadrural residence ,417 ,405 ,432 .614 (1 ifurban, 0 otherwise) (22.9) (16.6) ( 1 5 . 9 ) [.4871 Number of observations 4544 3196 7740 115 Table 14. Predictors o f (log) income from employment (ages 15 to 64). Itl-statistics are in \ parentheses. Log (Income from Salaries, AgriCukure & Self-Employment Activities, In-kind Income) Means and Standard Deviations All Male Female Potential Experience ,042 , 0 3 9 ,046 22,0 ( 1 4 . 3 ) ( 9 . 6 3 ) (10.31 111.31 Potential Experience2/ 100 - ,084 - ,081 - ,087 ( 1 3 . 4 ) ( 9 . 8 1 ) ( 8 . 7 3 ) Self-Evahated Health Very Good , 2 8 7 [ ,4531 Good - ,056 - ,048 - ,063 , 5 4 1 ( 2 . 8 7 ) ( 1 . 8 7 ) ( 2 . 1 9 ) L.4981 Satisfactory - ,128 - ,148 - ,103 .14 5 ( 4 . 6 5 ) ( 3 . 9 2 ) ( 2 . 6 2 ) L.3521 Bad I Very Bad - ,398 - ,432 - ,352 , 0 2 7 ( 7 . 7 3 ) ( 5 . 9 1 ) ( 4 . 9 8 ) [ ,1621 Gender (1 ifmale, 0 otherwise) ,196 ,587 ( 1 1 . 9 ) L.4921 Marital status of householdhead ,120 . 2 1 4 ,003 ,725 (1 ifmarried, 0 otherwise) ( 6 . 1 6 ) ( 7 . 6 3 ) ( 0 . 1 2 ) [ ,4461 Schoolingattainment None ,004 L . 0631 Primary ,163 ,178 ,058 , 0 4 1 ( 1 . 2 3 ) ( 1 . 0 3 ( 0 . 2 8 ) L .1991 Gymnasia1 ,504 .373 ,734 , 1 3 7 ( 3 . 8 9 ) ( 2 . 2 1 ( 3 . 6 2 ) L ,3441 Professional ,894 ,754 1 . 1 3 ,257 ( 6 . 8 5 ) ( 4 . 4 5 ( 5 . 5 2 ) [ ,4371 High School 1 . 0 1 ,847 1 . 2 7 .362 ( 7 . 7 2 ) ( 4 . 9 9 ) (6.21) L.4811 Post-Secondary/ foremen 1 . 2 2 1 . 0 6 1 . 4 8 ,063 ( 9 . 1 5 ) ( 6 . 0 9 ) ( 7 . 1 4 ) [ ,2431 Higher 1 . 4 6 1 . 3 0 1 . 7 2 ,135 (11.1) ( 7 . 5 8 ) ( 8 . 3 7 ) L.3421 Ethnicity Romanian ___._ _._._ .919 L ,2731 Hungarian ,133 .123 ,155 , 0 6 2 ( 4 . 0 9 ) ( 2 . 7 0 ) ( 3 . 4 0 ) L.2421 Roma - ,024 - ,061 ,019 .010 ( 0 . 2 9 ) ( 0 . 6 0 ) ( 0 . 1 2 ) [ .0991 Other ,122 ,136 ,102 , 0 0 9 ( 1 . 4 7 ) ( 1 . 2 8 ) ( 0 . 7 7 ) 1.0951 Urbanirural residence , 4 2 7 , 4 1 9 ,438 ,614 (1 ifurban, 0 otherwise) ( 2 3 . 4 ) ( 1 7 . 1 ) (16.1) [ ,4871 Number o f observations 4544 3196 7740 116 Table 15. Predictors of (log) income from employment (ages o f 15 and 64). Itl-statistics are in parentheses. A11 Public P r i v a t e Standard sector sector Deviations Potential Experience I 032 .030 11233) ( 6 . 3 8 ) ( 9 . 6 6 ) [11.31 Potential Experience* / 100 -.059 - ,053 - , 0 6 1 ( 4 . 8 7 ) Gender (1 ifmale. 0 otherwise) .215 ,299 ,587 ( 1 0 . 6 ) ( 1 3 . 9 ) L.4921 Marital status of householdhead (1 ifmamed, 0 otherwise) ,080 (4.48) ( 3 . 1 8 ) [ .446] Schoolingattainment - ,145 ,004 ( 1 . 0 7 ) [ . 0631 Primary I ,147 ____. , 0 4 1 (1.24) [ .199] Gymnasia1 1 ,309 - ,012 ,168 ,137 II (2.65) ( 0 . 0 9 ) ( 3 . 1 4 ) r.3441 Professional ,471 ,173 . 3 3 1 , 2 5 7 (4.01) ( 1 . 2 7 ) ( 5 . 7 3 ) L.4371 High School ,566 ,290 ,408 ,362 (4.81) ( 2 . 1 3 ) ( 6 . 9 1 ) [ .481] Post-Secondary1 foremen :709 ,446 ,522 .063 (5.89) ( 3 . 2 2 ) ( 6 . 8 0 ) [ ,2431 Higher I .,964. , 6 3 1 ,893 ,135 (8.12) ( 4 . 6 0 ) ( 1 3 . 3 ) [ ,3421 Occupation I Employee - - _ _ _ - _ - _ _ Employer ,209 ,192 (4.14) ( 3 . 3 4 ) Self-employed 1.921' - .904 (38.78) ( - 3 1 . 5 ) Type of Enterprise Worked State - - - - - Private II -.122 (6.881 Other :.005 (0.18) Ethnicity Romanian II _ _ _ _ _ ___.. .919 - - - - - [ ,2731 Hungarian ,102 .002 , 1 6 6 ,062 (3.49) ( 0 . 0 4 ) ( 4 . 0 4 ) [ ,2421 Roma ,178 - .240 , 2 0 1 ,010 (2.37) ( 0 . 7 3 ) ( 2 . 3 1 ) [ ,0391 Other ,048 - .004 ,073 ,009 I (0.65) ( 0 . 0 4 ) ( 0 . 6 8 ) [ ,0951 Urbanlrural residence -"" (Iifurban,0otherwise) ,203 ,092 (1 1.92) ( 4 . 0 5 ) Constant 9.06 9.48 (75.49) ( 6 7 . 9 ) (146.7) Number of Observations 7740 2216 117 Protecting the Poor and Vulnerable' Emil Tesliuc. Lucian Pop and Richard Florescu INDEX I.OverviewoftheSocialProtectionSystem................................................................................... 119 I1 The Impact o f Social ProtectionPrograms on IncomeDistribution ........................................... . 124 I11. The Impact o f Social ProtectionPrograms onPovertyReduction............................................. 126 B.Distribution of SocialProtection Benefitsacross groups andBenefitAdequacy.............................. A.Coverageof Social ProtectionPrograms ...................................................................................... 126 126 C. Marginal Benefit Incidence Analysis............................................................................................ 128 D.CostBenefit Analysis ................................................................................................................. 129 E.Overall Effectivenessof Social Protection ProgramsinReducingPoverty...................................... 130 F.ImprovingSocial AssistanceAdministration. ................................................................................ 133 IV. Summary of Key Issues andPolicy Recommendations............................................................. 136 Bibliography..................................................................................................................................... 138 Annex 1.Main Social Assistance Cash Transfers during2000 - 2002............................................ Statistical Annex .............................................................................................................................. 139 150 156 Annex 3. Social Insurance Benefits ................................................................................................. Annex 2.Unemployment Benefits 2002 .......................................................................................... 157 1 The authors are indebted to Gordon Betcherman. Nicholas Bumett. Margaret Grosh. Cem Mete. Stefan0 Scarpetta and Quentin Wodon for their guidance and advice. W e thank the Quality Team o f the Social Protection Anchor. Randa el-Rashidi in particular. for organizing a Quality Enhancement Review during the early stages o f the paper. in our benefit. The institutional review o f the main social protection programs. annexed to the paper. was prepared by team o f experts from the Ministry o f Labor and Social Solidarity. the National Pension Authority. and the National Employment Agency. We thank Mr. Petre Ciotlos. Mr. Dumit~u Calinoiu and Mr. IoanCindrea 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 I. OverviewoftheSocialProtectionSystem 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 o f public interventions that assist individuals, households and communities to manage risk better and provide support for the critically poor (Holzmann and Jorgensen, 2001). InRomania, the protection o f the poor and vulnerable was a constant priority of successive governments, as witnessed by the large share o f resources mobilized through the system in the last eight year, and its anti- cyclical role. The Romanian state redistributes about 10% of GDP via social protection programs. During 1995-2002, the share o f social protection transfers in GDP was remarkably stable, between 9% and 11% of GDP (Table 1). During the recession o f 1997-1999, the Romanian social protection system acted like a social-shock absorber, smoothing the social costs o f economic transformation. T h i s 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 inparallel with a process o f decentralization o f social assistance activities. The share o f social assistance activities (all hnds except for social insurance and unemployment) financed from the local budget increased from 16%in 1995 to 31% in 2002. Table 1. Share of Social ProtectioninGDP, and Social ProtectionFinancing b y Source Year Social Protection Source of Funds for Social Protection Program Total Unemployment % GDP State Budget Local Budgets Social Insurance Fund 1995 9.3 15 3 73 9 100 1996 8.9 15 3 76 7 100 1997 9.6 19 2 67 13 100 1998 10.5 19 3 65 13 100 1999 10.8 18 3 65 13 100 2000 9.8 16 5 69 10 100 2001 10.1 20 6 69 6 100 2002 10.0 18 8 69 5 100 Source: ConsolidatedGovernment Budget, Ministry of Finance, Bucharest The main social protection programs or policies can be grouped into: (i) insurance; and (ii)social 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 o f benefits and services for contnbutors, 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% o f the GDP. From social risk management perspective, the social insurance system mitigates a number o f social risks, such as the risk o f unemployment, accident, disability or lack o f earning capacity duringold-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% o f the GDP). The deficit i s the result o f (i)very a low dependency rate caused by population aging and a shrinking number of employees, and (ii) past early retirement policies. T o maintain the fiscal balance o f the system, the administrators opted for l o w replacement rates for pensioners (the ratio o f the average pension to the average wage i s 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 o f pensioners 119 which retired with significantly different pension levels for similar contribution terms. Table 2 presents the structure o f contributory pension's benefits, number o f beneficiaries and the associated costs for 2002. Annex 3 reviews the legislative and institutional aspects o f the system. Table 2. PensionBenefitsin2002 Number of beneficiaries (th ou) GDP share (YO) Total Social Insurance benefits 7.09 Social Insurance Pensions, o f which 6,212.3 6.46 State social insurance pensions 4,535.3 5.99 Farmers pensions 1,677.0 0.46 Social Aid Pension 6.2 0.00 W a r veterans pensions 30.2 0.04 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 i s set for 75% o f the minimumgross wage and is granted for a period o f 6 to 12months depending onthe length o f 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 o f service. Table 3 presents the structure o f the unemployment benefits, number o f beneficiaries and the associated costs for 2002. Beside the cash benefits, the unemployment fund finances a wide range o f active labor market measures, including job counseling, public works and micro-credit programs. Annex 2 reviews the legislative and institutional aspects o f the system. Table 3. UnemploymentBenefitsin 2002 N u m b e r o f b en eficia r i e s G D P share (YO) U n e m p l o y m e n t b e n e f i t 2 2 6 , 9 4 9 1 m o n t h 0.2 9 S u p p o r t a l l o w a n c e 2 1 7 , 3 7 9 l m o n t h 0 .o 7 G r a d u a t e s a l l o w a n c e f o r v o c a t i o n a l i n t e g r a t i o n 5 2 , 2 7 5 l m o n t h 0.0 3 S e v e r a n c e p a y m e n t s 1 9 , 0 0 0 1 y e a r 0.0 8 Source: Government of Romania, ANOFP, Bucharest The social assistance system-includes a number o f 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 o f implementation they were split into two separate programs. Together, the two programs channeled about 0.35 % o f the GDP in 2002 (compared with a budgetary allocation o f 0.40%). Annex 1reviews the legislative and institutional aspects o f the system. 0 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 implementationbecame ineffective. Eligibility ' for the MIG is gained for applicants that pass an income and asset test. The income threshold is function 120 o f family income and size. The MIGbenefit 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 c ost o f 0.28% o f the GDP. At the end o f 2 002 the number o f families benefiting o f MIG was about 380 thousands, or about 5.4% o f 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 l o w income families during the cold season (November to March), the size o f the benefit depending on the aggregate income level of the family and the source / type o f 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 inmonthly 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 o f this program, covering 13.5% o f the country's population, for a total cost o f 0.1% o f 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 o f the population. By far, the biggest share o f the social assistance transfers, are represented by the State Child Allowance and the Supdementarv 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 o f 0.68% o f GDP in 2002. The State Child Allowance is an universal benefit, granted monthly for each child up to the age of 16(18 ifenrolled inregular secondary education system), providedthose over the age ofseven are attendingthe school classes on a regular basis. Beginning with January 2003, the level o f the benefit is set for 210,000 ROL/month. The level o f 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 BeneJit.The level o f 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 1997. The supplementary allowance was introduced in an attempt to improve the targeting o f the program to the poor, knowing that families with more children face higher risk o f poverty. However, there were two inconsistencies between this objective o f 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 n o extra benefits for children rank five or higher. Thus, the program failed to cover the marginal income gap for families at higher risk o f poverty, despite the low cost o f 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 o f the social assistance benefits, number o f beneficiaries and the associated costs for 2002. 121 Table 4. M a i n Social Assistance Benefits in 2002 N u m b e r o f G D P S h a r e B e n e f i c i a r i e s ( % ) F A M IL Y A L L O W A N C E S S t a t e c h i l d a l l o w a n c e 4,8 3 5 , 6 0 6 0 .6 3 S u p p l e m e n t a ry a l l o w a n c e f o r f a m Hies w ith c h i l d r e n 1 ,o 2 2 , 9 0 0 0 .o 5 N e w b o r n g r a n t 1 5 0 , O l 1 0.0 1 S O C I A L A S S I S T A N C E S o c i a l a s s i s t a n c e b e n e f i t ( M I G ) 1 ,O 9 8 , 4 5 3 0 .2 8 R e s i d e n t i a l h e a t i n g a l l o w a n c e s a n d f a c i l i t i e s p r o v i d e d to h o u s e h o l d s t o 3,O 2 3 , 0 4 8 0 .o 7 d e f r a y h e a t i n g c o s t s A l l o w a n c e f o r f o s t e r c a r e a n d f a m ily c a r e 4 3 , l 1 4 0 .o 2 E m e r g e n c y b e n e f i t 4 6 8 0 .o 0 S u b s i d i e s to a s s o c i a t i o n s a n d f o u n d a t i o n s 5 , 8 6 2 0 .o 0 G r a n t s 2 , 2 9 5 0 .o 0 B E N E F I T S T O P E O P L E W I T H D I S A B I L I T I E S S o c i a l a s s i s t a n c e a l l o w a n c e t o p e o p l e w i t h s e v e r e o r s i g n i f i c a n t s i g h t N A 0 .O 8 im p a i r m e n t M o n t h ly c o m p e n s a t i o n N A 0 .o 8 P e r s o n a I a s s i s t a n t N A N A Source: Government of Romania, Ministry of Labor and Social Solidarity, Bucharest The coverage of the social protection system i s extensive. Overall, 87% o f the population i s covered by at least one social protection transfer, directly or indirectly (as household members, through income and consumption sharing) (see Table 5, based on ABF22002). 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 AJ3F confirms the ranking o f social protection transfers, the largest programs being pensions (82% o f reported s ocial protection s pending), 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 ina narrow range, with l o w coefficients o f variation. For comparability, Table 5 presents similar statistics for private transfers - remittances received from other households, in cash or inkind. Overall, the benefits collected from private transfers' reach 41% o f the population and represent one quarter o f total social protection benefits - they are quite important. These remittances tend to occur between members o f 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 o f transactions and value). 2 ABF i s the RomanianAcronym for the HouseholdBudget'Survey. See backgroundpaper 1 for a description o f the survey. 122 Table 5. Receipt of Social Protection Benefits: Household-level Descriptive Statistics, H B S 2002 Coef. Share of SP `rogram # of cases % covered Mean benefit Variation benefits iocial Protection, o.w.: 26817 87% 1655568 0.81 100.0% Social Insurance pension - old-age 12256 29% 2550989 0.29 62.3% pension - disability 2496 9% 1637341 0.24 9.3% pension - survivorship 3034 6% 825814 0.22 5.0% pension - farmers 5954 12% 476746 0.22 4.9% pension - war veterans 145 0% 841876 0.73 0.2% pension - social assistance 71 0% 992842 0.19 0.1% unemployment benefit 1512 7% 1085030 0.35 4.2% redundancy payments 27 0% 3668775 0.32 0.3% Social Assistance child allowance 9544 56% 306286 0.40 8.5% scholarship 123 1% 983709 0.39 0.3% support for people with disabilities 614 2% 930544 0.35 1.2% allowance for war veterans 1246 2% 398234 1.03 0.9% allowance for victims o fpolitical persecution 121 0% 1630869 0.82 0.4% social assistance provided by mayor's office (MIG in 2002) 808 4% 991932 0.79 2.1% other social assistance benefits 171 1% 747077 1.72 0.3% temittances (Private Transfers) 12923 41% 798769 2.92 23.9% cash remittances 3107 9% 1398926 1.63 10.4% in-kindremittances 11342 37% 517726 3.27 13.5%1 mrce: WB StafEstimations based on ABF 2002 During 1995-2002, the coverage o f the social insurance system expanded slightly, due to the increase in the number o f households benefiting from pensions (Table 6). The coverage o f the unemployment benefit program was highly anti-cyclical, with the program expanding during the 1997-1999 recession, and contacting thereafter. Most social protection programs maintained their coverage, with the notable exception o f the MIG- the program which inherited the Social Aid in2002, 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 o f a weakly targeted "bread allowance" that accompanied the liberalization o f the sector for six months. The bread allowance temporarily increased the coverage of the social assistanceprograms by 8-9 percentage points. 123 Table 6. Changes inthe Coveragewith Social ProtectionPrograms, 1995-2002 1995 1996 1997 1998 1999 2000 2001 2002 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 - socialassistance 1.0% 1.1% 0.6% 0.2% 0.2% 0.2% 0.2% 0.2% unemploymentbenefit 11.8% 7.5% 7.4% 10.0% 12.3% 10.9% 9.0% 7.0% redundancypayments 0.3% 0.1% and child care leave 1.6% 1.1% 1.1% 14% 1.3% 1.0% 1.0% 0.8% socialassistance/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% SocialAid / MIG in2002 0.7% 0.5% 0.4% 0.5% 0.5% 3.9% other socialassistancebenefits (includesMIG 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% allowancefor 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% 86.8% Source: WB StaffEstimations based on AIG 1995-2000 and ABF 2001-2002 About 84% of households benefit from at least one social protection program. The gap in coverage with social protection program i s low: about 7% o f the poor and 24% o f the non poor do not receive at least one type of transfer (Table 7). Incontrast, one infive households receives more than one social protectionbenefit (23% for the poor, versus 15% for the non-poor in2002). Table 7. Gaps andDuplicationsinthe Coveragewith SocialProtectionProgramsby Poverty Status # Social Protectionbenefits / HHs 1995 1996 1997 1998 1999 2000 2001 2002 TotalPopulation No benefit 19 20 17 17 16 16 16 16 Onebenefit 61 62 55 64 64 65 66 65 Multiple benefits 19 18 28 19 19 18 19 19 Non-poor No benefit 25 26 23 24 24 24 22 24 Onebenefit 59 60 54 61 61 62 62 61 Multiplebenefits 15 15 23 16 15 15 16 15 Poor No benefit 9 9 9 9 8 9 8 7 Onebenefit 65 67 57 68 67 69 70 70 Multiple benefits 26 24 34 24 24 23 22 23 Total 100 100 100 100 100 100 100 100 Source: WB StaffEstimations based on AIG 1995-2000 and ABF 2001-2002 11. The Impactof SocialProtectionProgramson IncomeDistribution The previous section highlighted the large redistributive effort occurring in Romania. Both social insurance and social assistance spending are designed to be redistributive: to transfer resources from current to past workers in the case o f pensions, and from taxpayers to the poorest families in the case o f targeted social assistance. How successful is this redistribution in reducing inequality? H o w are program benefits being 124 distributed? Who are more likely to capture the benefits o f 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 o f the population, one should identify the welfare status o f a households in the absence o f the government intervention, or the counterfactual per adult equivalent consumption. Many benefit incidence studies subtract the entire amount o f the transfer from householdincome or consumption to approximate pre- intervention welfare, and then rank the population into quintiles. Such approximation assumes: in fact, that there is no replacement o f the lost income source through savings, increase labor effort, remittances or other behavioral responses. This assumption i s implausible. Other studies take the post-transfer or observed consumption, as the welfare indicator. This altemative 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 o fthe replacement income would be around 50% o fthe value o f the transfer (Ravallion (2000), v an de W alle (2001;2 002)). Inorder t o e stimate the inequality- reducing (and, in the next section, the poverty-reduction) impact o f social protection transfers, we net out 50% o f the transfers from household consumption, and rank households based on this counterfactual consumption. While this estimate i s not precise, the results presented in the paper are robust to the choice o f a different share for the replacement income, especially for transfers that are small compared to household income. Thus, the assessment o f the inequality- or poverty-reducing potential o f the MIG and other social assistanceprograms would be robust to the choice o f this parameter, while will tend to differ in the case o f pensions. We tested for the sensitivity o f the findings to the choice o f this parameter, and we found very little divergence inthe qualitative findings. Three related concepts are used to identify the inequality-reducing or redistributive impact o f social protection transfers. A transfer is: Regressive: if the poorest groups receive a smaller share o f program benefits than the share o f the group in total consumption. If the share of the poorest quintile in total consumption is 4%, and the group captures only 3% o f the program benefit, the programis regressive. Mildly Promessive, 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 o f the group in total population. If the share of the poorest quintile intotal consumption is 4% (and intotal population i s 20%), and the group captures 10% o f the program benefit, the programi s mildlyprogressive. Highly Prouessive, ifthe poorest groups receive a larger share of programbenefits than the share of the group in total population. If the share o f the poorest quintile in total consumption i s 4% (and in total population is 20%), and the group captures 21% or 60% o f the program benefits, the program i s highly progressive. Progressive transfers contnbute to inequality reduction, and may contribute to povertyreduction as well A transfer can be characterized as progressive or regressive given the distribution o f 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 o f a program: Statically: What share o f program benefits accrue to a particular income group or quintile? and Marginally: Given a small increase (decrease) in the size o f t he program, which income groups will capture (loose) the extra benefits? 125 111. The Impact of Social Protection Programs on Poverty Reduction The previous s ection assessed the impact o f t he s ocial protection programs inreducing inequality. This section takes a narrower look, at the poverty reduction impact of these programs. Using three related concepts o f coverage, absolute targeting incidence and adequacy - share o f program benefits in the consumption o f the beneficiary - the last part of this section highlightthat the good targeting performance o f the MIG i s accompanied by weak coverage and benefit adequacy. All other programs are weaker poverty- reduction instruments. A. Coverageof Social Protection Programs Social protection programs reach directly (program participants) or indirectly (the household members o f program participants) about 87% o f the population, including 96% o f the poorest quintile (Table 8). The MIG, program that represents a social safety net o f last resort in fighting poverty, while very well targeted, covers only 11.7% o f the poorest 20% o f the population. Table 8. Program Participation Rate by Quintile, 2002 Program Poorest 2 3 4 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% 1% 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 - socialassistance 0.2% 0.2% 0.1% 0.1% 0.1% 0.2% unemploymentbenefit 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 support for people with disabilities allowance for war veterans allowance for victims ofpolitical persecution social assistance provided by mayor's office (MIG in2002) other social assistance benefits Source: WBStaffEstimations based on ABF 2002 B. Distribution of Social ProtectionBenefits acrossgroups andBenefit Adequacy While the previous section examined the coverage o f programs (interms o f "people" or beneficiaries), this section examines the distribution o f benefits (the target "incidence") o f social protection programs usingdata from the ABF 2002. Two concepts are used: (a) absolute target incidence, which measures average benefits receivedby any particular group as a share o f total benefits (or the targeting outcomes o f 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 o f the benefit inprotecting 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%) o f 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 are relatively more important to the poor than the non-poor (Table lo). Specifically, public transfers represent 80% of total consumption for the poorest quintile (net o f 50% o f social protection transfers), as compared with 6.8% for the top quintile. Table 9. Distribution of SocialProtectionBenefits,by Quintile-2002 Program // Consumption Quintile Poorest 2 3 4 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 o f political persecution 18% 14% 26% 31% 12% 100% social assistanceprovidedby mayor's office (MIG in2002) 62% 21% 12% 4% 2% 100% other social assistancebenefits 33% 24% 19% 11% 13% 100% MemorandumItems Share intotal population 20% 20% 20% 20% 20% 100% Share intotal consumption 7% 12% 17% 23% 40% 100% A number of social assistanceprograms, suchas the MIG, supportfor 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 Table 10. Benefit Adequacy by Quintile-2002 Ratio ofprogram benefits in the consumption of the group Program // Consumption Quintile Poorest 2 3 4 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 o f political persecution 0.2% 0.1% 0.1% 0.1% 0.0% social assistance provided by mayor's office (MIG in2002) 4.1% 0.8% 0.3% 0.1% 0.0% other social assistance benefits 0.4% 0.2% 0.1% 0.0% 0.0% Source: WB StaffEstimations based on ABF 2002 C. MarginalBenejit Incidence Analysis The current distribution o f program participants, however, is a poor indicator o f how marginal benefits will be distributed if the size o f the program will change, as discussed in Lanjouw and Ravallion (1999). If the socioeconomic composition o f a program changes as programs expands or contracts, the distribution o f the new program participants will be different than the current ones. One can distinguish the case o f early capture, when the poor benefit from the program in the initial stages, but as program expands, the non poor will capture most o f the gains from increasedparticipation. In contrast, some programs may exhibit late capture by the poor. In the initial stages o f program implementation, the non-poor get most benefits. However, as the program expands, the poor get most o f the gains from increased participation. When reforming existing social protection programs, by rationalizing some and expanding others, it is marginal distribution o f 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 f o r the MIG. Compared t o the earlier program ( Social A id), the c overage o f t he MIGincreased substantially in its first year o f application, thanks to more generous funding and adequate eligibility criteria - the income threshold for the program was substantially increased. However, the program coverage o f the extreme poor i s still deficient: only three out o f ten extreme poor benefit from the program. It is important to assess if, under the current implementationarrangements, an increase inthe MIG will entail a reduction or an increase inthe targeting performance o f the program. An increase in the size of the MIG can be implementedwithout changing the targeting performanceof the program. Table 11 presents estimates o f average and marginal program participation rate, and the average and marginal distribution o f program beneficiaries by quintiles. The average share of program beneficiaries illustrates what proportion o f the MIG beneficiaries are from a given quintile. The marginal share o f program beneficiaries presents, for a given increase o f decrease in program size, what proportion o f the new beneficiaries will come from a given quintile. The MIG i s a case o f "early capture" program. From 128 100 new program participants, 58 will be from the poorest quintile, and only 2 will come from the richest one. Compared to the current distribution o f program participants, the distribution o f 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-poor3. We suggest, in section III.F, that beneficial changes to program administration may be implemented to further improve its targeting performance. Table 11. BenefitAdequacyby Quintile-2002 ProgramParticiDationRate Share of Program B e n e f i c i a r i e s Q u i n t i l e s Average Marginal Average Marginal 1 11.7% 10.5% 60.4% 58.1% 2 4.5% 4.2% 23.O% 23.3% 3 1.9% 1.9% 9.8% 10.8% 4 1.O% 1.1% 5.1o/o 5.9% 5 0.3% 0.3% 1.7% 1.9% National - _. -__._ ._- 3.9%- 3 .fi0/n 100.00A - - - .- . - 1- 00.00! - - .- .- Source: WB StaffEstimations based on ABF 2002 D. CostBenefit Analysis Different programs contribute toward the poverty reduction goal proportionally with the amount o f resources channeled through the program (transferred to the beneficiaries) and the degree o f targeting. One way to focus on the targeting performance o f various programs i s to estimate the amount o f resources spent in order to reduce the poverty gap o f the program beneficiaries by 1ROL. In this section, we use a simplified cost- benefit analysis (CBA) framework to assess the contributions o f various social protection programs in reducing poverty. For each program, we computed a costhenefit ratio: (i) benefits o f the social The protection (SP) programs were judged against their impact on reducing poverty, that i s how much they contribute t o the reduction inthe poverty g ap? and (ii)Program c osts include the benefits transferred to programparticipants (without taking into account administrative costs, or other intake-cost for participants). The estimation proceeds inthree simple steps: First, we compute estimates o f the current and counterfactual poverty gap4. To estimate the counterfactual poverty gap, we use the results reported in section 3, that the level o f consumption in the absence o f a welfare program equals current consumption minus half o f the welfare benefit. Second, we estimate the contribution o f 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 inthe poverty gap due to the program bytotal volume ofbenefits derived from that program. The validity of our inferences i s limited to marginal changes inprogram participation, say, an increase by 1% to 10%. These estimators will be a poor guide for a major reform o f the program. 4 The counterfactual poverty gap takes into account important incentive effects, such as the change in their behavior ifpublic transfers were discontinued (supplying more work, tapping private transfers, etc.). The implicit social welfare function used inthe CBA penalizes all spending inexcess o f the household poverty gap. Such criterion fails to take into account the fact that (i) some social programs have multiple objectives; (ii) i s not it feasible to "fillthe poverty gap" by passive transfer policies (Le., 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 programbenefits). 129 Due to limitations inthe ABF data, the program cost includes only the cost o f 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 o f ROL spent per 1 ROL reduction in the poverty gap, and the lower the number (preferably close to one), the better the outcome. T o 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, 1995- 2002 I I Extreme Poverty I Total Poverty 19971 19981 1g991 20001 20011 20021 19951 19961 19971 19981 19991 2Oool 20011 200 social insurance Icontributorvbenefits I 561 681 491 481 4.71 451 5.31 551 12.91 1761 12.21 1211 1231 1211 1541 155 pension length-of-sewice 6 4 7 7 5 5 5 2 5 4 5 1 5 9 6 2 15 7 20 91 14.41 14 01 1491 14 31 18 51 19 0 pension disability -. 6 5 6 6 4 9 4 9 4.6 4 3 5 7 5 8 17 91 197 11.9 13 0 12 3 11 7 17.8 pension. survivor 5 1 6 4 4.5 4 3 4 3 4.0 4 5 4 5 114 17 6 11.5 11.1 109 10 6 pension. lamer 4 3 5 6 3.9 4 0 3 5 3 6 3 6 4.1 104 162 10.9 10.7 107 11 6 pension. war veterans 4 4 6 5 6 2 6 6 4 4 4 3 6.1 5 6 7 6 14 3 18.0 22 3 103 11.1 pension. social assistance 3 1 3 3 3.4 2 8 2 5 3 1 4 9 5 1 8 0 6 2 6 8 4 3 8 3 6 1 unemployment benefit 3 9 4 7 3 6 4 1 3 6 3 5 4.2 4.1 8 2 11.9 9.3 10.5 8 6 5.2 redundancy payments 9 6 6 7 socialsecurity benefits for maternity, childcare leave 9 4 13 4 7.1 8 4 6 9 8 7 9 6 15 9 24 7 26 7 55 6 social assktance Inoncontributorybenefits 4 6 5 2 4 1 3 8 3.5 3.2 3 8 3 3 11.0 9 4 child albwsnce $ 4 7 5 9 4 0 3 6 3.5 3.2 3 7 3.9 11.8 9 4 SChDlarshlp 12.1 217 121 162 12.9 11.8 6 9 7.8 27.5 69.9 38.3 133.0 42.7 167 supportfor people with disabilhlies 3.9 4 7 3 1 3 5 2 9 3 0 3 9 3 6 8 0 9 7 7 7 MIG In2002. MISS in 95-96 2 0 2 2 1 6 1.8 2 6 2 0 3.9 other social assistance benefits 3 2 2 5 6 5 5 3 3 7 3 1 3 1 4 9 20 6 16 1 6 8 10.2 other noncontributorybenefits 6 1 8 9 5 2 5 5 5 4 5 2 6.7 6.8 18.2 15.4 15.1 13.8 23.5 21 2 allowance forwar veterans 5 7 8 6 5 0 5 2 5 1 4 7 5.1 5.4 16.3 15.3 16 5 158 allowance lor victimsof political persecution 26.9 131 7.3 6 0 6.7 6 5 192 23.3 204.1 Total SP 5 3 6 4 4.6 4 5 4 4 4 3 4.9 4.9 Comparing the costs6 and bene@ of social protection programs, the most eficient 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 o f programs for which we can estimate these indicators with sufficient precision in ABF i s "support for people with disabilities" and the child allowances. E. Overall Effectiveness of Social Protection Programs in ReducingPoverty The indicators o f coverage, absolute target incidence, and relative target incidence (importanceiadequacy), all reveal important information about the effectiveness o f social protection programs. This section seeks to combine those multiple indicators for a more comprehensive review o f these programs, in particular with respect to their effectiveness inreducing poverty. Figure 1 plots in a single graph the three related concepts o f 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 11). The x-axis presents the coverage o f the poor. The share o f total benefits received by the poor i s plotted on the y-axis (absolute target incidence). Adequacy (relative incidence) is captured by the size o f 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 o f these graphs, with a large bubble (equal to the size o f 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 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 l o w coverage and benefit adequacy. Second, the child allowances are less well targeted but have the largest coverage o f the poor and extreme poor. Finally, the rest of the programs have high leakage, l o w coverage o f the poor. Some programs within this group maybe good candidates for rationalization, with the resulting savings to be addedto the MIG. 131 Figure 2 illustrate the improvement in coverage that occurs with the implementation of the M I G in 2002. 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 o f the poorest 20% o f the population (89% o f them). Figure 2. Change inthe Effectivenesso f the SocialAid / MIGPrograminReducingPoverty TotalPoverty Extreme Poverty Social Aid 19974001, MIG 2002 Social Aid 1997.2001, MIG 2002 . . I I 100- I 1007 , . . I I I I ! 5 0 5 10 15 m C h e r q e L_______ Source: WBStaflEstimations based on ABF 2002 The Government has a good tool to reduce extreme poverty inRomania: 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 o f 0.4% o f GDP or even more, will cover an increasingproportion o f the extreme poor. Some o f the extreme poor, however, are likely to be missed by the program, as illustrated in Table A1 and A2 annexed. In2002, the program was successful in covering larger households, those where the couple was living together or unmarried (pooling resources, while one o f the adults claimed the benefits only for himand all dependents), the Rroma (compared to an average coverage o f the poorest decile o f 17%, the coverage o f extremely poor Rromas was 3 6%), those headedby adults w ithno o r few formal s chooling, by farmers, unemployed and housewives. In general, MIG recipients are households familiar with the social assistance system; most o f them (87% o f total recipients) receive a number o f other social assistancebenefits, especially child allowances. The extreme poor households who failed to be covered by MIG benefits are especially urban households, households affected by industrial restructunng 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 o f poverty, is the main challenge faced by the current administration. The next section suggests a number o f changes inprogramimplementation to facilitate such an outcome. F. Improving SocialAssistanceAdministration The previous sections showed that, among social assistance programs, the MIG program has the best targeting performance, but a relatively low coverage o f the poor. This section will take a closer look at the institutional arrangements o f MIG, trying to highlight two o f the most important weaknesses o f the program: (i) arrangements;and(ii) funding implementation. 1. Funding arrangements Unclear distribution o f the financing burdenbetween the central and local administration According to the law, the financing o f 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 andor "equilibration grants"'. Theoretically, the combination o f central and local financing provides a good mix o f incentives for the local administrations. A decentralization o f 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 o f 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 o f program financing resulted poor targeting, with larger programs in relatively richer areas and smaller programs in poorer areas, determined by the financing capacity o f the local administrations alone. The optimal solution to this problem i s 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 i s a risk that local administrations will abuse the program (allowing substantial leakage o f funds) if the capacity o f 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 o f co-financing by local authorities provides the right incentives for them to use the funds ina "responsible" manner. While these arrangements may work, the implementation o f the program was hindered by the lack o f clarity with respect to the share o f co-financing o f the local versus central government. The MIG law leaves unspecified the share o f 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 o f the program as was the case with the Social Aid program in 1996 (after the first year of implementation, when financing was providedby the central budget). Given the "guaranteed" feature o f the MIG, the program may end up crowding out other local expenditures, considered more important by local constituencies. Unpredictable funding o fthe MIG Local authorities (Local Councils) cannot plan their budget without a clear provision specifying their level o f co-financing, and face similar difficulties in planning their cash-flow. In 2002, this impossibility to predict the volume o f resources provided by the central government for MIG pushed many Local Councils to limit the demand for the program. Some examples o f limiting demand practices are: imputing "by default" revenues from informal work during the summer and consequently paying only 40% -60% o f the nominal benefit amount, and/or setting the number o f community work days at 9 dayslmonth independently o f 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 i s played by the Judet Council, which is in charge to distribute the funds among localities. The lack o f 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 o f enforcement with regard to benefit payments 7 The "equilibration grants" are resourcestransferredto the poorestlocalities, on the basis of the principle of "social solidarity". 134 The main difference between the MIG and the former means-tested Social Aid program i s that MIG i s "guaranteed". However, at the moment there i s no authority responsible or able to enforce this provision. The Ministryof Labor and Social Solidarityhas 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 o f funds towards local budgets i s entirely under the control o f the autonomous Judet Councils, and the eligibility criteria can be relaxed or tightened by the local authorities (Local Councils). MIGresources from the centralbudget are not clearly earmarked The central budget earmarks inblock the resources for the MIG and heating allowances, and from 2003 the salaries o f 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 o f disabled, making thus impossible to control the local budget share allocated for MIG co-financing. This arrangement has a direct impact on spending priorities o f local public administration, which has incentives to set the payment o f 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 o f MIGbenefits and heating subsidies. Combined with the lack o f enforcement regarding the obligation to cover entirely the amounts due to MIGbeneficiaries, the actual arrangement leads to overdue or partial payments o f MIGbenefits. 2. Implementation Discretionary power o f local public administration with regard to the criteria for assets evaluation and eligibility. A large share ofthe income ofthe 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 imputedvalue for the productive assets owned by the households, based on norms established by the local administration. The procedures to determine the "imputed value" o f productive assets (such as livestock, land or agricultural equipment) vary from one Local Council to another. Intheory, the local administration i s 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 o f productive assets to be taken into account inthe means testing procedure to the Local Councils. This provision generated large discrepancies between the imputed values o f similar assets over relative homogeneous areas*. Evidence from earlier surveys confirms the existence o f price differences across localities within judets, but o f 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 o fthis 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 kmd o f documents that should be provided by the potential beneficiaries. Many Local Councils ask for documents legalized by notaries, although the law does not require such type o f evidence, thus raising the monetary and time cost for the applicants. The Government and the Ministryo f Labor and Social Solidarity may consider following with detailed implementation norms to clarify and limit the amount o f paperwork and documents requested from the applicants. The Government may consider strengthening the design, financing andimplementation o fthe MIGprogram: 1. A more equitable co-financing scheme, based on an algorithdformula that takes into account the financial capacity o f localities together with the expected number o f 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 between42 thou and 420 thou, etc 135 Government may use one o f the basic needs poverty maps currently used by FDRS. Altematively, reliable estimates o f poverty at locality level may be produced by combining the information from the 2002 census with the 200213 AI3F. If the capacity to monitor local compliance inproviding the required co-financing is weak, the second-best altemative is to continue with 100% financing o f the program from the central budget. 2. Homogenization o f assets evaluation criteria at judet level. The Government may consider the harmonization o f the good practice o f certain Judet Councils, who have established minimum and maximum imputed values for productive assets used to determine household income, nationwide. A possible altemative i s to abandon the assets evaluation altogether, and use the information from the global income tax. Such approach is not yet recommended, as the global income tax captures imperfectly agricultural income. 3. Adequate enforcement mechanism for the MIG. I f the legislative maintains the local government co- financing requirement, this should be complemented with an adequate enforcement mechanism. Without such mechanism, some practices observed in 2002 may continue at even larger scale, such as the accumulation o f arrears in benefit payment, or the partial payment o f the MIG benefit (in 2002, a number o f local administrations paid only 80% o f the MIG benefit -i.e. the estimated central budget share). The institutionalization o f an adequate enforcement mechanismwill also prevent the use o f value judgments or stereotypes regarding the "deserving poor" at local administration level. 4. Eligibility criteria. Currently the law disqualifies those eligible beneficiaries who have debts to the State. The Law or the secondary legislation should mention explicitly that in such cases, the benefits can be used first to cover these debts, and then to allow the delivery o f the benefits to those beneficiaries cleared by debts for the remaining months. 5. Workfare. The workfare requirement proved to be a good self-targeting mechanism, but the law should state explicitly that the number o f hours should be proportional with the level o f the benefit. Currently many beneficiaries are giving up the program because in some areas they are requested t o work the maximum number o f hours (72)regardless the size o f the benefit. 6. The Local Councils use o f the workfare requirement i s almost exclusively focused on small ecological projects such as street cleaning or garbage collection. Traditionally, these activities were dominated by workers with none or l o w education. Many o f the new poor, such as skilled blue collar workers who lost their j obs when their s tate-enterprise was c losed or restructured, c onsider such workfare requirement stigmatizing, but would be willing to work in activities that would make better use o f their skills. The Govemment and the Local Councils should consider altemative workfare options, which may benefit not only the new poor, but also the local community. IV. Summary of KeyIssues andPolicy Recommendations The Romanian social protection system is widespread - it redistributes about 10% o f GDP andreaches 83% o f the population -, and contributes substantially to poverty reduction. Inthe absence o f the social protection transfers - accounting for the behavioral response o f the households - the number o f poor would increase by almost 50%, while the number o f extreme poor would double. A large share o fthis impact is due to pensions, inpart due to their sheer size, inpart due to their targeting. In the absence o f an adequate pension, most pensioners will end up poor. The current process o f pension recorrelation, that places an emphasis on both equity aspects (alignment o f the pension benefits with past contributions) and protection against extreme poverty, contributedto this result. During the transition period, the social assistance system was reformed substantially. From a system that used categorical benefits (such as the child allowance programs), starting 2002 the system strengthened the 136 MIG, a means-tested program well targeted toward the poor. The Govemment should be commended for all these improvements inthe social assistancesystem. To address the problem o f extreme poverty, the Government should place a greater emphasis on the MIG. The targeting performance of the program is among the best inthe region. The coverage o f the program has increased dramatically in2002 versus 2001 (from 0.5% o fthe population in2001 to 5% in2002). Despite the increase in coverage, the program coverage o f the extreme poor i s still inadequate (only 30% o f the extreme poor are covered by the program). The uncovered poor are especially urban households, households affected by industrial restructuring which do not take up the program due to stigma costs, and households with able- bodied individuals. To ensure adequate coverage o f the extreme poor, the MIG needs to be expanded (from 0.3% of GDP in 2002 to more, possibly 0.6-0.8% o f GDP, to cover the poorest 10% o f the population with moderate leakage to well-off households). A moderate increase in the income threshold (over and above inflation) would extend the coverage o f the benefit and will reach more poor people. To make room for more resources to expand the MIG, the Govemment can cap the child allowance benefits to their current (nominal) levels, and use the resulting real savings for the MIG. Equally important, the Government may consider strengthening the design, financing and implementation o f the MIG, according to the recommendations made in the last section. However, sustained poverty reduction cannot be promoted by passive social protection policy alone. As illustrated inthe poverty profile (Tesliuc et all, 2003), a large share o f poverty i s associated with the process of narrowing o f the salaried labor market, whichpushedmany people into less onerous activities in farming, non-agricultural self-employment, or unemployment. Somehow surprising, Romanian self-employed seems to have chosen this eaming alternative not because it promises higher incomes compared to formal employment, but because they do not have access to the salaried wage market. The poverty rates among self- employed or farmers, economically active groups, are close to the rates o f unemployed, Such findings suggest the need of a thorough investigation o f the labor market, as well as o f the other factor markets. For a large number o f the able-bodies poor and extreme poor, policies to stimulate labor mobility geographically and occupationally, and to improve the business environment, are key for policy reduction. These aspects were covered in another background paper (Mete, 2003) which analyzed the relationship between labor market and poverty. 137 Bibliography Holzmann, R. (2002): "Risk and Vulnerabilitv: The forward looking role o f Social Protection ina globalizing world", in: E. Dowler and P. Mosely (eds.): Poverty and Social Exclusion i n North and South, London and New York (Routledge). Holzmann, R. and S. Jorgensen (2001): "Social Risk Management: A new conceptual framework for social protection, and beyond", International Tax and Public Finance 8/4, 529-556. Jalan, Jyotsna, and Martin Ravallion (2002) "Estimating the Benefit Incidence o f an Antipoverty Program by Propensity Score Matching", Development Research Group, World Bank, Washington D C Lanjouw, Peter and Ravallion, Martin (1999), o f the Program Capture", World Bank Economic Review, Vol 11, N o 2,257-73 Mete, Cem (2003) "Labor Force Participation, Unemployment and the Poor", Background paper for Romania Poverty Assessment, World Bank, mimeo Ravallion, Martin (2000) "The Mystery o f the Vanishing Benefits: Ms.Speedy Analyst's Introduction t o Evaluation", Development Research Group, World Bank, Washington D C Tesliuc, C., Pop L., and Tesliuc E. (2001) "Romania: Social Protection and the Poor", Polirom Printinghouse, Romania Tesliuc, C., Pop L.,and Panduru F. (2003) "Poverty in Romania: Profile and Trends during 1995-2002", Background paper for Romania Poverty Assessment, World Bank, mimeo van de Walle, Dominique, (2001) "Viet Nam's Safetv Net: Protection and Promotion from Povertv", mimeo, Development Research Group, World Bank, Washington DC van de Walle, Dominique, (2002) "The Static and Dynamic Incidence o f Vietnam's Public Safew Net", mimeo, Development Research Group, World Bank, Washington D C 138 Statistical Annex Table A 1. Participation inMIGprogram among the Extreme Poor, by Location Location Program Participation Rate Total 17% Region 1 25O/o 2 13% 3 12% 4 I6% 5 15% 6 15% 7 15% 8 17% Area of residence urban II o/o rural 20%o 139 Table A1 cont'd. ParticipationinMIGprogramamongthe ExtremePoor by HouseholdCharacteristics HH Characteristics Program Participation Rate Total 17% HH Size 1 7% 2 4% 3 11% 4 21% 5+ 25% Gender HH Head male 17% female 16% Marital status married 15% living together 38% divorcedlseparated 19% widowed 15% unmarried 29% Nationality romanian 15% hungarian 5% rroma 36% other 3% Education HH Head no formal schooling 35% primary, grades 1-4 17% middle, grades 5-8 15% vocationallapprentice 13% highschool,grades 9-12, incl. lower hig 15% post-secondaryor foremen's school 6% higher school, short and long term 2% Occupation HH Head employee 3yo self-employed non-agricultureincl. fami 14% self-employed agriculture incl.family h 34% unemployed 27% pensioner 7% housewife 27% other (dependent, militaryservice, etc. 52% MultipleSP Benefits? One benefit 4% Multiple benefits 33% Multiple SA Benefits? One benefit 6O h MultiDle benefits 79% 140 Table A 2. Participation inMIGprogram among the Extreme Poor, Logistic Regression Dependent Variable: MIG Recipient Odds Ratic Std. Err. z P=-Z hsize 1.228 0.067 3.76 0.000 Region 1 -Iregion-2 0.591 0.153 -2.03 0.043 -Iregion-3 0.776 0.177 -1.I 0.267 1 -lregion-4 1.342 0.283 1.40 0.162 -Iregion-5 1.036 0.285 0.13 0.899 -Iregion-6 0.621 0.172 -1.72 0.086 -Iregion-7 0.718 0.215 -1.I 0.269 1 -Iregion-8 0.962 0.440 -0.08 0.932 Rural 1.085 0.207 0.43 0.669 female HH head 1.412 0.246 1.98 0.048 Age HH head 0.976 0.006 -3.80 0.000 romanian hungarian 1.229 0.464 0.55 0.584 rroma 1.474 0.368 1.56 0.119 other 0.919 0.972 -0.08 0.936 no formal schooling primary, grades 1-4 0.430 0.097 -3.76 0.000 middle, grades 5-8 0.386 0.096 -3.83 0.000 vocationallapprentice 0.300 0.089 -4.06 0.000 highschool, grades 9-12, incl. lower hig 0.219 0.084 -3.94 0.000 post-secondary or foremen's school 0.288 0.226 -1.59 0.113 higher school, short and long term 0.167 0.181 -1.65 0.099 employee self-employed non-agriculture incl. fami I.865 1.621 0.72 0.474 self-employed agriculture incl. family h 9.647 7.171 3.05 0.002 unemployed 7.565 5.632 2.72 0.007 pensioner 1.247 0.937 0.29 0.769 housewife 15.739 12.751 3.40 0.001 other (dependent, militaryservice, etc. 21.654 17.101 3.89 0.000 141 Table A 3. Increase inInequality in Absence of Social Protection Programs Simulation Gini index. with and without Social Protection Programs 1995 1996 1997 1998 1999 2000 2001 2002 Consumotion PAE 0316 0308 0296 0 293 0 286 0 280 0 284 0288 social insurance icontributory benefits 0.352 0.341 0 327 0 325 0 316 0.308 0.315 0.320 pension length-of-sewice - 0.337 0.329 0.315 0 313 0.304 0.297 0.303 0.308 pension -disability 0.319 0.311 0.299 0 296 0.288 0.283 0.287 0.291 pension suwivor - 0.319 0.311 0.299 0 296 0.288 0.283 0.286 0.290 pension -farmer 0.319 0.311 0.299 0 296 0.288 0.283 0.287 0.291 pension. war veterans 0.316 0.308 0.296 0 293 0.286 0.280 0.284 0.288 pension social assistance - 0.317 0.309 0.296 0 293 0.286 0.280 0.284 0.288 unemployment benefivlabour market-entry benefuincome support 0.321 0.311 0.299 0 296 0.290 0.283 0 286 0.290 redundancy payments 0.316 0.308 0.296 0 293 0.286 0.280 0.284 0.288 social security benefits for temporary disability, maternfly and child care it 0 316 0.308 0.296 0 293 0.286 0.280 0.284 0.287 social assistance inoncontributory benefits 0.321 0.313 0.304 0 301 0.292 0.286 0.290 0 296 chiid allowance 0.320 0.311 0.302 0 300 0.291 0.285 0.289 0 292 scholarship 0.316 0.308 0 296 0 293 0.286 0.280 0.284 o 288 support for people with disabilities 0.317 0.309 0.297 0 294 0.286 0.281 0.284 0.288 soc assist provided by mayor's ofices (MIG in 2002) MISS in 95-96 0.316 0.308 0.296 0 293 0.286 0.280 0.284 0.290 other social assistance benefits (includes MiG in 95-96) 0.316 0.309 0.297 0 293 0.286 0 280 0.284 0.288 other noncontributory benefits 0317 0.308 0,296 0 293 0.286 0.281 0.284 0.288 allowance for war veterans 0.317 0.308 0,296 0 293 0.286 0.281 0.284 0.288 allowance for victims of political persecution 0316 0308 0296 0293 0286 0280 0284 0288 Total SP 0358 0346 0335 0333 0 323 0314 0321 0329 Table A 4a. Social Protectionbenefits incidence on the poverty status of beneficiaries - 2002 BEFORE AFTER extreme extreme poverty poverty Extreme Poverty headcount SE 95%CI headcount SE 95%CI social insurance/contributory benefits 0.2523 0 0088 0.2350 0.2696 0.0949 0.0064 0.0823 0,1075 pension - length-of-service 0.2344 0.0095 0.2158 0.2530 0.0603 0.0055 0.0494 0.0711 pension disability 0.1852 0.0183 0.1552 0.2152 0.1000 0.0137 0.0730 0.1269 pension survivor 0.2038 0.0154 0.1736 0.2340 0.1306 0.0136 0.1038 0,1574 pension fanner 0.1892 0.0188 0.1582 0.2202 0.1411 0.0138 0.1140 0.1682 pension war veterans 0.2297 0.0666 0.0989 0.3605 0.0940 0.0480 -0.0003 0.1882 pension social assistance ----- 0.1564 0.0612 0.0362 0.2766 0.0732 0.0518 -0.0286 0.1751 unemploymentbenefiti labour market- entry benefit 0.1878 0.0156 0.1572 0.2185 0.1099 0.0130 0.0844 0.1353 redundancypayments 0.2482 0.1117 0.0288 04677 0.1633 0.1080 -0.0490 0.3755 social security benefits for temporary disability, maternity and child care leave 0.0822 0.0276 0.0280 0.1365 0.0289 0.0174 -0.0053 0.0631 social assistancelnonconmbutory benefits 0.1686 0.0096 0.1499 0.1874 0.1392 0.0086 0.1223 0.1561 child allowance 0.1557 0.0094 0 1373 0.1741 0,1359 0.0086 0.1190 0.1528 scholarship 0.1515 0.0565 0.0405 0.2625 0.1190 0.0570 0.0071 0.2310 support for people with disabilities 0.2879 0.0292 0.2305 03453 0.2102 0.0284 0.1543 0.2661 MIG 0.5426 0.0400 0.4641 0.6211 0.4203 0.0400 0.3417 0.4989 other social assistancebenefits 0.2057 0.0981 0,0130 0.3984 0.1773 0.0977 -0.0148 0.3692 other nonconhbutory benefits 0,1446 0.0177 0,1099 0.1794 0,1086 0,0161 0.0769 0.1403 allowance for war veterans 0,1569 0.0186 0,1203 0.1936 0.1187 0.0172 0.0851 0.1524 allowance for victims of political persecution 00155 0.0092 -0.0025 0.0336 0.0037 0.0038 -0.0038 0.0112 Total SP 0,2322 0.0082 0.2160 02484 0.1153 0.0067 0.1021 0.1284 BEFORE AFTER total poverty total poverty Total Poverty headcount SE CI headcount SE CI social insurance /contributory benefits 0.5150 0.0103 0,4947 0.5353 0.2970 0.0100 0.2774 0,3167 pension length-of-service 0.4905 0.0118 0,4673 0.5137 0.2294 0,0102 0.2093 0.2494 pension disability 0.4390 0.0176 0 4044 0 4736 0.3027 0.0176 0.2682 0.3372 pension survivor --- 0.4960 0,0180 0.4606 0.5313 0.3810 0.0180 0.3457 0.4164 pension farmer 0.4774 0.0198 0.4384 0.5164 0.4085 0.0188 0.3716 0.4454 pension war veterans -- 0.4225 0.0738 0 2775 0.5675 0.3080 0.0711 0.1684 0.4477 pension- social assistance 0.4238 0.0821 0.2626 0.5851 0.3050 0.0779 0.1518 0.4581 unemployment benefit/ labour market-entry benefit 0.4670 0.0182 0.4312 0.5028 0.3647 00188 0.3278 0.4017 redundancy papents 0.4229 0.0949 0,2365 0.6093 0,1633 0.1080 -0.0490 0.3755 social security benefits for temporary disability, matemity and child care leave 0.1823 0.0400 0.1037 02610 0.0998 0.0301 0.0406 0.1590 social assistance /noncontributory benefits 0.3652 0.0122 0.3412 0.3892 0.3355 0.0119 0.3121 0.3589 child allowance 0.3539 0.0123 0.3298 0.3780 0.3311 0.0119 0.3076 0.3546 scholarship 0.3103 0.0622 0.1881 0.4325 0.2439 0.0620 0.1220 0.3658 support for people with disabilities 0.5892 0.0306 0.5291 0.6493 0.5119 0.0336 0.4459 0.5778 MIG 0.7760 0.0323 0.7125 0.8395 0.6989 0.0364 0.6273 0.7705 other social assistancebenefits 0.4598 0.0923 0.2785 0.6411 0.3139 0.0944 0.1285 0.4992 other noncontributory benefits 0.3961 0.0240 0.3488 0.4433 0.3389 0.0245 0.2906 0.3871 allowance for war veterans 0.4190 0.0249 0.3700 0.4679 0.3640 0.0245 0.3159 0.4121 allowance for victims ofpolitical persecution 0.1556 0.0421 0.0729 0.2382 0.0899 0.0392 0.0128 0.1670 Total SP 0.4530 0.0100 0.4334 0.4727 0.3044 0.0098 0.2850 0.3237 Note: poverty headcount was computed by netting out 50% of each SP transfer from household consumption Table A 5. Social Protection benefits incidence on the poverty gap of beneficiaries - 2002 BEFORE AFTER extreme extreme Extreme Poverty poverty gap SE 95% C I poverty gap SE 95% C I social insurance /contributory benefits 0.0680 0.0031 0.0619 0.0741 0.0185 0.0016 0.0153 0.0216 pension length-of-service 0.0617 0.0033 0.0552 0.0682 0.0111 0.0013 0.0084 0.0137 pension disability 0.0467 0.0049 0.0371 0.0563 0.0200 0.0034 0.0133 0.0268 pension 0.0535 0.0057 0.0423 0.0646 0.0265 0.0043 0.0180 0.0349 pension farmer ----survivor 0.0422 0.0044 0.0335 0.0510 0.0260 0.0030 0.0202 0.0318 pension - war veterans 0.0423 0.0111 0.0205 0.0642 0.0174 0.0089 0.0000 0.0348 pension - social assistance 0.0425 0.0212 0.0008 0.0841 0.01 17 0.0078 -0.0037 0.0271 unemployment benefit/ labour market- entry benefit 0.0441 0.0046 0.0350 0.0531 0.0208 0.0028 0.0153 0.0263 redundancy payments 0.0945 0.0488 -0.0014 0.1903 0.0337 0.0215 -0.0085 0.0760 social security benefits for temporary disability, maternity and child care leave 0.0142 0.0058 0.0028 0.0257 0.0026 0.0016 -0.0006 0.0057 social assistance /noncontributory benefits 0.0462 0.0036 0.0391 0.0534 0.0313 0.0027 0.0260 0.0366 child allowance 0.0381 0.0032 0.0318 0.0443 0.0302 0.0027 0.0249 0.0355 scholarship 0.0515 0.0225 0.0072 0.0957 0.0367 0.0171 0.0032 0.0703 support for people with disabilities 0.0792 0.0102 0.0591 0.0993 0.0432 0.0070 0.0294 0.0570 MIG 0.1819 0.0202 0.1423 0.2215 0.1050 0.0147 0.0761 0.1339 other social assistance benefits 0.0524 0.0266 0.0002 0.1047 0.0309 0.0207 -0.0097 0.0715 other noncontributory benefits 0.0322 0.0050 0.0224 0.0420 0.0210 0.0040 0.0131 0.0290 allowance for war veterans 0.0350 0.0053 0.0245 0.0454 0.0230 0.0044 0.0144 0.0315 allowance for victims o f political persecution 0.0034 0.0021 -0.0007 0.0075 0.0007 0.0008 -0.0008 0.0022 Total SP 0.0671 0.0033 0.0606 0.0735 0.0252 0.0020 0.0213 0.0291 143 B E F O R E A F T E R total poverty total poverty T o t a l Poverty gap SE C I gap S E C I social insurance /contributory benefits 0.1672 0.0049 0.1576 0.1768 0.0702 0.0034 0.0634 0.0769 pension length-of-service - 0.1559 0.0053 0.1456 0.1663 0.0491 0.0031 0.0431 0.0552 pension - disability 0.1288 0.0079 0.1132 0,1443 0.0734 0.0067 0,0603 0.0865 pension survivor 0.1468 0.0082 0.1308 0,1629 0.0923 0.0072 0.0783 0,1064 pension farmer 0.1309 0.0082 0.1148 0,1469 0.0992 0.0068 0.0859 0,1125 pension war veterans 0.1231 0 0275 0.0690 0.1771 0.0785 0.0228 0.0337 0,1232 pension social assistance ---- 0.1156 0.0322 0.0523 0.1788 0.0610 0.0229 0.0161 0,1060 unemployment benefiti labour market-entry benefit 0.1314 0.0076 0,1165 0,1463 0.0853 0.0063 0.0730 0.0977 redundancy payments 0.1616 00609 00420 02812 0.0738 0.0482 -0.0209 0,1685 social security benefits for temporary disability, maternity and child care leave 0.0469 0.0126 0.0221 0.0717 0.0188 0.0069 0.0052 0.0324 social assistance Inoncontributory benefits 0.1132 0.0056 0.1022 0.1232 0.0932 0.0048 0.0838 0.1026 child allowance 0.1034 0.0052 0.0931 0.1137 0.0912 0.0048 0.0818 0.1006 scholarship 0.1051 0.0321 0.0421 0.1681 0.0831 0.0284 0.0273 0.1390 support for people with disabilities 0.1933 0.0145 0,1648 0.2218 0.1453 0.0124 0.1209 0.1698 MIG 0.3327 0.0235 0 2865 0.3789 0.2496 0.0206 0.2090 0.2902 other social assistance benefits 0.1263 0.0458 0.0364 0.2163 0.0961 0.0428 0.0121 0.1802 other noncontributory benefits 0.1024 0,0088 0.0851 0.1197 0.0783 0.0080 0.0627 0.0939 allowance for war veterans 0.1096 0.0092 0.0916 0.1276 0.0854 0.0082 0.0692 0.1015 allowance for victims ofpolitical persecution 0.0258 0.0062 0.0137 0.0379 0.0061 0.0029 0.0005 0.0117 Total SP 0.1533 0.0049 0,1436 0.1629 0.0800 0.0038 0.0726 0.0875 Note: thepoverty gap was computed by netting out 50% of each SP transfer from household consumption 144 Table A 7. Urban Area: Changes inthe Coverage with Social Protection Programs, 1995-2002 1995 1996 1997 1998 1999 2000 2001 2002 social insurance /contributory benefits 41.3% 39.7% 40.1% 44.4% 46.6% 47.4% 48.8% 46.9% pension length-of-service - pension disability 6.1% 6.1% 6.4% 7.3% 7.4% 7.7% 9.8% 10.1% pension survivor 3.8% 4.0% 3.7% 3.8% 3.9% 4.1% 4.1% 3.9% pension farmer --- 22.7% 23.4% 24.4% 25.4% 26.0% 27.9% 29.1% 29.5% 1.7% 1.7% 1.4% 1.2% 1.2% 1.5% 1.2% 1.5% pension - war veterans 0.2% 0.2% 0.2% 0.2% 0.1% 0.2% 0.1% 0.1% pension social assistance - 0.7% 0.8% 0.6% 0.2% 0.1% 0.2% 0.2% 0.1% unemployment benefifflabour market-entry benefiffincome support 10.2% 7.1% 7.2% 10.7% 12.5% 11.6% 9.8% 7.6% redundancy payments 0.4% 0.1% social security benefits for temporary disability, maternity and child care leave 2.1% 1.5% 1.4% 2.0% 1.9% 1.2% 1.5% 1.0% social assistance /noncontributory benefits 54.5% 52.8% 62.5% 57.3% 55.9% 55.8% 55.9% 56.8% child allowance 53.1% 51.3% 54.2% 55.8% 54.7% 54.4% 54.2% 54.6% scholarship 1.7% 1.3% 1.1% 1.O% 0.7% 0.8% 1.O% 0.7% support for people with disabilities 0.8% 1.2% 1.5% 1.4% 1.3% 1.6% 1.5% 1.5% soc assist provided by mayor's offices (MIG in 2002) MISS in 95-96 0.6% 0.4% 0.4% 0.6% 0.5% 2.0% other social assistance benefits (includes MIG in 95-96) 0.8% 1.0% 13.4% 0.9% 0.4% 0.6% 1.O% 0.9% other noncontributory benefits 1.8% 1.7% 1.5% 1.6% 1.4% 1.3% 1.2% 1.4% allowance for war veterans 1.6% 1.6% 1.4% 1.5% 1.2% 1.2% 1.1% 1.2% allowance for victims of political persecution 0.2% 0.2% 0.2% 0.2% 0.2% 0.1% 0.2% 0.2% 'Total SP 80.0% 77.7% 82.6% 83.0% 83.1% 83.1% 84.3% 83.4% Table A 8. Rural Area: Changes in the Coverage with Social Protection Programs, 1995-2002 1995 1996 1997 1998 1999 2000 2001 2002 social insurance/contributory benefits 58.0% 54.7% 53.3% 55.3% 57.7% 56.0% 57.2% 56.6% pension length-of-service - 25.2% 25.7% 25.6% 26.1% 27.1% 27.2% 28.9% 29.0% pension -disability 5.1% 5.2% 5.5% 6.1% 6.1% 6.4% 7.1% 7.7% pension- survivor 6.8% 7.1% 7.3% 7.7% 8.4% 8.2% 9.0% 9.5% pension -farmer 24.5% 23.7% 23.5% 24.4% 24.0% 24.1% 25.2% 24.3% pension-war veterans 0.8% 0.7% 0.8% 0.6% 0.5% 0.5% 0.5% 0.5% pension social assistance - 1.4% 1.5% 0.5% 0.3% 0.4% 0.2% 0.2% 0.2% unemployment benefit'labour market-entry benefit'income support 13.7% 8.1% 7.8% 9.2% 12.0% 10.1% 8.0% 6.3% redundancypayments 0.2% 0.1% social security benefitsfor temporary disability,maternityand child care leave 1.O% 0.6% 0.8% 0.7% 0.6% 0.6% 0.5% 0.5% social assistance /noncontributorybenefits 53.6% 52.4% 62.5% 55.6% 56.9% 57.2% 58.0% 59.3% child allowance 51.7% 50.4% 52.9% 53.7% 55.4% 55.6% 56.2% 56.6% scholarship 0.6% 0.6% 0.4% 0.4% 0.2% 0.7% 0.3% 0.3% support for peoplewith disabilities 2.5% 2.6% 3.0% 2.9% 2.7% 2.5% 3.1% 3.1% soc assist provided by mayor'soffices (MIG in 2002) MISSin 9596 0.9% 0.5% 0.4% 0.5% 0.5% 6.1% other social assistance benefits (indudes MIGin 9596) 1.1% 1.5% 14.6% 1.1% 0.8% 0.5% 0.6% 0.5% other noncontributorybenefits 5.7% 5.4% 4.8% 5.1% 4.6% 4.0% 4.2% 3.6% allowancefor war veterans 5.5% 5.3% 4.6% 5.0% 4.4% 3.9% 3.9% 3.4% allowancefor victims of political persecution 0.2% 0.2% 0.3% 0.2% 0.2% 0.2% 0.4% 0.3% Total SP 87.2% 86.0% 88.4% 88.4% 89.8% 89.3% 90.3% 91.0% 147 P 2 N Hvi -z3s I- l-00 F 0 3 N 0 * - L ,- >' 0 0 x N O ~ WN 0 N 0 N 0 0 0 N * B U n LI1 3W P h E 'C n: eL 0 8 d3 U 0 N I I- 0 8 8 - P * cm i' 0 0 0 0 3 3 3 m am 0 N !l m ?m d N m a I- 2 eI 2 3 N 0 - ?% s A c 2 z i'N m 3 m '0 N I-- W m m L 5 N a i : : N 0 N 0 0 N H I w 0 N 0 z 0 0S I i 8 0 0 fnL 5 0 0 i; N 0 6 N 0 0N 0N 0 N N 0 0 N 10 3 x i $ i x N 0 N 0 x 0 0 1 05 - c0 W 0 I-- W N R 22 c -A - c I- c00 c 3 N W 3 3 -* 00 N m00 I- 0 8 w w 3 0 8 0 6 0 N 0 0 N 0 N N N 8m -? m 0 8 2 l- 1 00 N8 I s xs 0 0 0 0 0 l- -r -, 3 4 z 2 4 z Q - 0 8 6 Q 0 0 N N :: N :: 0 N m v, 3 3 z i 4 z 4 z 4 z 4 z 3 4 4 2 2 4z s N 0 0 N s N N 0 -88 1 P D 4 b, 3 s N r"c I z m m 8 0 2 m 4 I m 8 m 8 E: 4 ? M e I I li r I i? N If: I 3 5 35 ? r ';a3 v) e ! $ Q 0 Q mS d 8 2m .-E - E Mapped in or Mapped out? The Romanian Poor inInter-Household and Communitv Networks June Idh.2003 Maria Amelina Dan Chiribuca Stephen Knack 1 Introduction . B.MainFindings ................................................................................................. A. Description ofthe Study.................................................................................. 170 170 171 ........................................ 173 2. SourcesAo.fPublic C. Policy Implications............................................. Income-an Overview and Private Formal Transfers ............................................................... 174 176 B.FormalPrivate SocialFlows............................................................................ 179 181 D.Informal Income.............................................................................................. C. Formal private organizations ........................................................................... 182 183 B.Maps ofInter-Household Transfers.................................................................. A. Informal Inter-Household Flows by Category.............................................. 190 193 H.Cash at aPremium................................................................................... G. Parents and Children................................................................................ 193 Interactions with households from other cities, villages and abroad: .............195 I.GeographyofTransactions-DistanceataPremium................................. 196 4. Who Participates inInformalTransactions? A. Household welfare........................................................................................... 196 197 B.Household Characteristics ............................................................................... 197 C. Social Capital .................................................................................................. 198 199 E. Inter-household Flows andMIGAssistance ..................................................... D.Inter-household Flows: Multivariate Tests....................................................... 200 5. Social Capital, Local Governance and Empowerment o f the Poor A. Social Capital and Access to Resources -The Case of Triple Exclusion?........201 200 B. Social Capital, LocalGovernance and PublicAssistance ................................. 204 D.Decentralization and Local Government Efficacy............................................ C. Social Capital and Quality o f Public Service Delivery ..................................... 208 212 6.Policy Implications 215 References 217 Appendices 219 168 Many colleagues contributed to the design, data collection, and analysis. We thank Emanuela Galasso, Philip Keefer, Ghazala Mansuri, Alexandre Marc, Dena Ringold and Claire Wallace for helpful and insightful comments. A h a Barsan and Diana Marginean assisted in the design o f the survey and in the data collection. N i l s Junge and Victor Giosan provided background information. Lu Wang and Min Ouyang assisted in data analyses. Leah Cohen and Diana Marginean edited and formatted the final version. The Institute o f Advanced Studies (IHS) and Metro Media Transylvania helped to design the study and carry out the field work. The World Bank Resident Mission in Romania, most prominently Richard Florescu helped with the logistics. Indicators on Local Democratic Governance in CEE designed by the Tocqueville Research Center and the Local Government and Public Service Reform Initiative (LGI) o f the Open Society Institute (Hungary) informed the design o f the public officials' questionnaire used in the survey. We gratefully acknowledge the grant from the Austrian Trust Fundthat made a study o f this magnitude possible. All errors and omissions are ours. 169 1. Introduction A. Description of the Study This chapter analyses patterns o f economic and social interactions that sustain the poor or, alternatively, isolate them yet further from other households, from the communities inwhich they live and, by extension, from social networks and economic opportunities. The study also assesses interactions o f the poor with local and central government in terms o f the level o f trust and satisfaction with public officials, the level o f involvement in public actions and public decision- making and the ability o f local governments to respond to the needs o f their poorer constituency, especially inproviding social assistance and other MinimumIncome Guarantee (MIG) benefits. This study examines the associations that have not been adequately considered when assessing the well-being o f the poor. These are: transfers betweenpoor households and formal private associations; transfers made to and fiom poor households within informal inter-household networks - (friends, relatives, close associates); association between the level o f income o f a household and a community and participation incollective action and local government decision making. Mapping out the universe o f these interactions from the position o f the poor i s the primary goal o fthis analysis. It has been observed in Central European countries in general and inRomania inparticular that during transition strong social ties connecting relatives, immediate friends and associates have become stronger, while the weak ties connecting individuals and households through professional and social associations have become weaker.' Inthis context, the poor are reported to be falling out o f bothtypes o f associations. Strong familial and friendly networks have become difficult to maintain because o f high maintenance costs, such as reciprocal gift giving and costs o f transportation and telecommunication. Weak associational networks have become less accessible to the poor as well, due to the high costs o f accessing newly reconfigured social networks on the one hand and the disassociation fiom old networks through loss o f employment and migration to rural areas on the other. Another reported reason for the low levels o f inclusion o f the poor in voluntary organizations in the post-communist states may be the defacto closed, hierarchical nature o f these organizations. These patterns o f voluntary mobilization contrast with the bridging role o f associations in western democracies and need t o be examined in greater detail.* ' See for example Manning, N.,0.Shkaratan et al. (2000). Work and Welfare in the New Russia. Aldershot: Ashgate; Toth, J. I. and E. Sik (2002). "Hidden Economy in Hungary 1992-1999% Neef, R. and M. Stanculescu, eds. The Social Impact of Informal Economies in Central and Eastern Europe. Aldershot: Ashgate; Stanculescu, M. (2002). "Romanian Households between State, ,Market, and Informal Economies" inR.Neef and M. Stanculescu, eds. The Social Impact ofhformal Economies in Central and Eastern Europe. Aldershot: Ashgate. Uslaner, Eric M. (2003). "Trust and Civic Engagement in the East and the West", in Gabriel Bodescu and Eric M. Uslaner, eds., Social Capital and the Democratic Transition, Routledge. 170 The effect o f this double exclusion on the poor i s increased vulnerability, inability to adjust to the realities o f the market environment, and, therefore, increased dependency on the state for immediate subsistence and for reintegration into broader social networks and the labor force. Both local and national govemments need to reach out to their poorer constituency with well targeted social assistance programs and other policies that increase economic opportunities among the poor. To study pattems o f formal and informal transfers and relate them to the characteristics o f households and o f communities, a nationally representative household survey was carried out. The survey covers household income and transactions for 2002 and includes expanded sections on: 1) informal inter-household transactions (gift giving, exchanges, and barter);3 2) channels through which resources flow to and from households (between relatives, friends, and other informal associates); 3) forms these flows take (cash, goods, or services). The survey also captures social capital aspects o f the socialization o f the poor interms o f generalized and specific trust, cooperation with other community members, and participation in formal and informal collective action. The survey assesses the sense o f control and optimism, the perception o f the respondents' ability to influence govemment decision-making, and the satisfaction with services provided by local and national govemment. A parallel survey o f local public officials carried out inthe same localities helpedrelate the views of local officials on the effects and effectiveness of govemment activities to community perceptions o f the quality o f public service provision and other public interest issues. Quantitative s tudy was complemented by qualitative analysis. Seventeen focus group sessions conducted with poor and average income inhabitants o f urban and rural communities provided examples o f interactions between community members and public and private service providers. A separate subset o f focus sessions was conducted with poor urban and rural Roma groups and with public andprivate service providers. B. Main Findings The study finds the poor to be at a disadvantage infamilial, social, and public networks. Formal private transfers flow away from poor households. Consistent with the findings for other CEE countries, weak associational ties within private networks de facto act as strong closed networks catering to their immediate membership and demonstrating little altruistic interest in the poorer members o f the community. This study shows that clubs and professional and special interest associations primarily channel resources to better-off households, who are more likely to be their members. At the same time, the poor, particularly the rural poor, contribute disproportionately to many organizations, particularly to church groups. This i s consistent with data from some developed countries, where church related transfers were found to be regre~sive.~ Informal inter-household flows are income neutral. These results run counter to the economic literature on inter-household transactions, which finds the net effect o f informal flows in developing countries to favor poorer and more vulnerable households and to act as means-tested public transfers. However, the finding is consistent with recent sociological studies o f transitional economies footnoted above, which show post-socialist dislocation to cause rifts in informal ties, thereby diminishing access for the poor to the informal networks to which they once belonged. Comparedto the majority o f budget householdlevel surveysthis study examines consistently both the income and the expenditure effects of informal transfers, looking at gift giving and exchanges symmetrically, both as a source of revenue for a householdand as an expenditure item. For contributions o f the poor to religious charities inthe US, see Verba, Sidney et al. (1995). "Voice and Equality: Civic Voluntarism" inAmerican Politics. Cambridge: HarvardUniversity Press. 171 A detailed block o f questions on types and channels o f informal transactions used in the study revealed much more active participation o f the population as a whole in informal transactions than i s captured by traditional budget surveys. 97 per cent o f respondents reported participating in informal gift giving and exchanges, with similar shares of poor and better-off households participating ininformal inter-household transactions. The reason for the low or negative net effect o f informal inter-household transfers is not the paucity o f transactions, but the prevalence o f outflows over inflows for most types o f informal transactions. Informal transfers are reported to constitute 8.5 per cent of household income for inflows and 12.3 per cent for outflows. Inthe case o f the poorest quintile, a slightly higher share o f outflows i s attributed to three factors: 1) a high level o f transfers from poor rural households in foodstuffs and financial assistance (from pensions and sale of agricultural output) to the households o f their adult urban children and other relatives; 2) the high price o f remaining in networks that include higher income households (the price o f symbolic capital); and 3) the patterns o f recollection, with respondents remembering even the smallest gifts, exchanges and payments flowing out o f the household (higher frequency, lower value o f transactions), while recalling only the larger gifts, payments and exchanges flowing in (lower frequency, higher value). Qualitatively, the rich and the poor participate in different types o f informal inter-household transactions. The poor are less likely to participate in altruistic g$-giving and more likely to engage in reciprocal transactions, such as exchanges o f goods and services and payments for minor services (minor repairs, child care, tutoring). A higher share o f poor rural households participated in exchanges and payment for minor goods and services. This pattern may be the result o f stronger personal ties within rural communities where associational transactions costs are lower and exchanges among neighbors are more frequent. About one half o f all households are involved in informal lending. For the poor, lending and borrowing comes in the form o f multiple transactions that are small in value, and help smooth consumption for a household anticipating monthly public assistance or public benefits transfers (pensions, minimum income assistance, child allowances). This pattern brings together private and public flows in a single risk management system, with public assistance leveraged against informal flows from neighbors, friends and sometimes relatives. The poor have lower levels o f trust inneighbors and other people in general. The poor are not as connected and have markedly fewer people they can rely on in solving pertinent life problems (health, legal, administrative, problems with the police, bank, assistance in getting a job). Lower levels o f trust and participation translate into lower "dividends" from social capital, such as assistance inneedand informal help with employment. Inthis context assistance from the government becomes particularly important. Indeed, poor and rural households are more likely to trust both local and national government and be satisfied with services provided. This may be the result o f lower expectations o f the poorer and less educated population strata or it may reflect the high level o f dependence on the government by those excluded from private networks. Hopefully, it is also the result o f the effort and performance o f local governments, particularly those in small rural areas. The MIG programs, despite a number of flaws discussed in more detail below, appear to be well-targeted to poorer households with lower incomes, a highnumber o f children, and fewer assets. 172 The poor are much less involved in collective action and are less active in demanding that their public needs are met. They exercise their voice in public service issues less effectively. Households from the richest quintile are more likely to be "civic activists"; attend public meetings, participate inprotests, and alert the media to local problems. They are twice as likely as households inthe poorest two quintiles to contact local officials about a public issue, and ten times as likely to contact national officials. Because o f their higher incomes, they are also more capable o f pursuing better public services through informal means. Richer households offer "gifts" to public officials more often to help resolve their problems. At the community level, controlling for other variables, households living in localities with higher levels o f civic activism (measured by an index o f attending public meetings, participating in protests, alerting media to 1ocal problems, notifying the police o r court of1ocal problems) receive higher MIG assistance which may point to the positive role o f accountability o f local governments in MIG allocations. Another important community level finding is that poor communities have been found to be at a disadvantage in terms o f access to fiscal transfers. Poor households, particularly rural poor households, are more likely to live in areas that receive less earmarked and non-earmarked transfers from the central and county governments. All public transfers except for the national Minimum Income Guarantee program (MIG) are found to be regressive. Combined with lower levels of private support, a lower ability to self-organize, lower levels of public activism among the poor, and self-admitted higher dependence on state support these reducedflows may perpetuate the cycle of poverty and exclusion. C.Policy Implications The findings presented here have important policy implications at two levels. First, this research deepens our understanding o f the concrete role major social assistance programs play in the livelihoods o f the poor, specifically, the national social assistance program, MIG. MIG payments constitute a fairly small share o f total income for the first quintile once all other income sources are taken into account (3.6 per cent), though this share increases quite substantially when the recipient households are factored out from the first decile (21.3 per cent o f income for the 28 per cent o f recipient households in the sample), rather than the first quintile. However, if we take into account the anti-poor nature o f other public transfers and the exclusion o f the poor from informal public and private networks, the relevance o f MIG payments inthe livelihoods o f poor households increases dramatically. The qualitative part o f the study indicates that MIG-related programs are valued very highly by poor recipients, as it provides cash in hand at regular intervals and helps to leverage other transfers - informal lending, exchanges, and even ad hoc employment. At the same time, the study uncovered important institutional and organizational constraints to the improved targeting and cost-effectiveness o f MIG flows. These include: 1) arbitrary allocation o f MIG management resources and assessment o f need by local government officials, 2) the need to finance a share o f MIG-related expenditures from local budgets, which are particularly burdensome for poor and rural localities, and 3) the high costs o f filing, information-gathering, and transportation faced by potential recipients in some areas. The finding that MIG assistance is higher (controlling for income and other factors) in large localities, in localities with greater locally-raised revenues, and in ethnically-homogeneous localitiessuggeststhat adjustments are neededto improve equal accessto social assistance. At the level of policy design, these findings can contribute to improved conceptualization and development o f policymaking instruments and approaches that work. The analysis shows that the poor live ina very distinct social environment. This environment affects their economic opportunities and coping strategies, inmany ways definingparticular channels through which the poor access social assistance and social protection programs. The challenge for the government i s 173 to incorporate this expanded understanding o f how the poor live and interact with other households, private organizations (including private service providers; trade, labor and landforest owners associations; the church), and local govemment into the design o f more readily accessible and effective social programs. Inlight of the findingthat other public transfers are regressive.atthe locality level, and that the poor are less organized in voicing their discontent and working collectively for their rights, the govemment needs to broaden its pro-poor approach to public transfers. To include the poorer constituency into both decision-making and service-providing networks, local govemments need to re-examine the way they interact with the poorer population. The menu o f possible procedural changes includes: 1) site visits to these areas followed by concrete community level actions, 2) specialized outreach programs, and 3) other efforts that include communities in the budgetary process and the evaluation of service provision. Inother words, local governments must choose from a menu o f practical empowerment mechanisms that have borne fruit in other parts o f the world and adapt them to localrealities and circumstances (Anderson, 2002). The paper i s structured as follows. The next part presents an overview o f all income sources, concentrating on public and private transfers and their role in the income o f poor households. The third part analyses the role o f various informal inflows and outflows in livelihoods o f households from different income groups. The progressivehegressive nature o f these transfers for the poor quintile is assessed. The second section o f this part looks at maps o f inter-household transfers interms of: 1) the nature o f relationships among transacting households, 2) the form o f transactions (cash vs. in-kind), and 3) the geography o f exchanges (same locality, other Romanian locality, other country). The fourth part discusses demographic, economic and social characteristics o f households in relation to their net position as either donor or recipient o f informal flows versus being uninvolved in informal transactions. Multivariate analysis i s used to provide a more rigorous examination o f the determinants o f inter-household flows, such as vulnerability, access to public social flows, and participation incollective action. The fifth part i s a chapter on social capital that examines the role o f trust and collective action in the access o f the poor to private and public flows. It assesses the relationship between the social capital endowment o f a locality and both the quality o f public services and the nature o f local government, noting also the tendency o f local governments to provide better services to better organized, networked communities, and ethnically more homogenous communities. The sixth section concludes with policy implications. 2. Sources of Income an Overview - The main difference in the s tructure of formal income sources between the poorest and other quintiles i s that household income for the poorest quintile is dominated by transfers, while for all other quintiles officially recorded salary constitutes the most important source o f income (Table 1). As expected, private business i s far more important for household incomes inthe highest quintile than for incomes of all other quintiles. Since transfers inthis analysis include both public and private flows, we will examine them inturn by category. 174 A. Public and Private Formal Transfers Almost half o f all income in the poorest quintile comes from formal transfers (Table 1). The most important part o f public transfers are social benefits, which include old age and other pensions, scholarships, and child benefits. These benefits constitute 38 per cent o f total net income for the poorest quintile, with similar shares o f recipients from urban and rural households, as compared to 13.3 per cent o f income for high-income households. Interestingly, pensioners report leveraging their pension in negotiations with potential employers, as this secure source o f income makes them more attractive as workers that may require lower payment and lower benefits. Examples o f exploitationwere brought up infocus group sessions: "Private employers want to hirepensioners... because, they say, `I'llgive them IM lei as wages, then they have their 2Mfrom theirpension, and they will be happy. " (GT, average income resident, Breaza, urban) Broader access o f the elderly to social benefits often tums them into providers for their younger relatives: "Ihaveadaughter andason-in-law, andtwonieces,andtheir incomeis only 1.6Mlei. Thereare four of them in theirfamily ...If it wasn'tfor myself and my wife, andfor our pensions, how would they manage? So we, the elderly, choose to help them. Because, thanks be to God, I don't need anythingfancy, I have everything I need. But theyoung people, they are truly in a dire situation." (Low income respondent,Alunis, rural) "Now my only source of support is myfather's pension. There areJive of us in ourfamily, and we all live off myfather j.pension. (low income respondent,Alunis, rural) 'I Social assistance, which includes social allowances, emergency relief, and different heating compensations, constitutes 3.6 per cent o f income for the poorest quintile, i s available to 25.7 per cent o f households from the poorest quintile and i s progressive (Table 1). More urbanpoor households (35.1 per cent) receive social assistance than their rural counterparts (21.8 per cent) (Tables 1A and 1B in Appendix 1). For the poorest decile the importance o f social assistance i s more dramatic - 28 per cent o f households in this decile receive social assistance. Among recipient households this assistance constitutes 21.3 per cent o f total income, as compared to 23.3 per cent o f households from the second decile, which receive only 8.4 per cent o ftotal formal income from social assistance. Recipients o f social assistance (MIG) consider it to be a vital part o f their income, as assistance i s received at regular intervals and represents "cash in hand" for cash-strapped households. Focus group respondents indicated that the guaranteed income enabled them to borrow small sums o f money or products from neighbours, relatives, and food stores, thereby smoothing consumption in times o f hardship: "We have neighbors, relatives who can lend some things to us. There are also the elderly who receive pensions, and we sometimes go to them, and they lend us money. When we receive the (children's) allocation or social aid, we return the money we borrowed. (...) Because ifyou don 't repay your loan, no one will lend you any money in thefuture." (Low income respondents, Nereju, rural) 176 "R1: Too little help! Wedon't really help each other! Only when we come to the store and they give us things. But even hereyou know you must bring money topayfor these things. R2: Wecertainly know we must bring the money! R2: And even at the store, when doyou know that you can come and buy things? It is only whenyou know your (social) aid is coming, that you will get some money." (Roma, low income respondents, Alunis, rural) At the same time, social assistancerecipients and other low income focus group respondents complained about difficulties in accessing public assistance due to four main factors: 1) poor access to information describing the documentation needed to receive aid and various eligibility criteria, 2) the dismissive attitudes o f local officials, 3) the high costs for filing documentation, and 4) the inability o f poorer localities to cope with their social assistance mandates. Poor access to information, dismissive attitude: "There is a new social aid law, andfor newborns up to the age of six months you get 750,000 lei. The newspapers say so. When I went to ask about this, they said they had not heard about it, that I should go talk to Adrian Niistase." (Roma, low income respondents, Focgani, urban) Lack o f awareness: "Ihavenoticed thatplenty of thesepeople (thepoor, main&) havealmost never madeany attempt, or they simply did not know how to applyfor social aidfrom County hall. Iknow afew examples of individuals who were supposed to receive such benefits. They even told me that they went there, but were not given any help in this respect. So this is theproblem. Suchpeople occupy the last position on the social scale, if I may say so. If they were to benefit from social aid, they would implicitly qualibfor free medical assistance.A large majority of them did not know anything about this, others had not received any guidancefrom City Hall, that is they were misinformed." (Physician, Breaza, urban) Arbitrariness o f local officials: "Ifone owned a horse,for example, or a wagon, they would not give you social aid! Ifyou owned a Persian carpet (ie., of good quality) in your house they wouldn't give you social aid either. Also, if you own a television set, you don 't get any aid! But, one would think that perhaps, a 50-year old person, after 20 or so years of being married, may have been able to accumulate afew thingdor, to accomplish something.'' (Roma, low income respondent,Alunis, urban) Costs of accessing services: "Last year, they were very disrespectful towards me and my wife. They asked her to go to the notary, and certib a personal income statement, which cost 85,000 lei. The costs for obtaining documentation are very high.'' (Low income respondents, Galati, urban) Inability o fpoorer localities to cope with social assistance mandates: "There are decisions taken by the Government, but they are applied at the local level. They give social help to thosepeople with income below a certain level,for example. The money should come from the local budget, but the local budget doesn't have enough money. They provided social assistantsfor people with disabilities. These assistants are employees. First they were paid by the Ministry of Labor. After that, they asked the local governments to pay 25%. In the next year they asked them to pay 50%. Now the local governments must pay 100% of the social assistants' salaries. This costs the local government about 3 milliards of lei. And now we have reached the 177 point where the local government d oesn't even have enough money to repair a h ole i n a road. " (Local councilor, Breaza) "I'm speaking about the local budgets of the villages, which are extremely small, I know villages that can hardly cover their operating costs. Among the material services provided, the Minimum Income Guarantee(MIG) is the most affected, because there are not enoughfunds. Ihave no specijk data with me, but asfar as Iknow, there were more than twenty thousand demands. These were only partially covered. And even those that were approved, many of them were not paid infull." (Public serviceprovider, Targu Mures). Another category o f public transfers are payments made by local government (city hall, local council, prefecture, specialized national agencies), and by public services (hospitals, schools, kindergartens). These transfers include public assistance for high cost medical emergencies, public medical insurance, and ad hoc assistance for education; railway ticket subsidies for pensioners, veterans, and students; agricultural subsidies; and subsidized loans for the purchase of primary residence. These are important flows that reach 47.6 per cent o f the sample (1,234 households) (Table 1). Public transfers constitute 6 per cent o f total income for the poorest quintile; 4.4 per cent for the second quintile; and 9.2 per cent o f total income for the highest income quintile, demonstrating greater importance o f these transfers for the incomes o f richer households (Table 1). Separating these transfers into assistance from public services providers (assistance with health emergencies, childcare and education), and other public transfers from local or national govemment (subsidized investments, agricultural subsidies, assistance for transportation) helps discem the following divergent patterns (Table 2 and 2a). Assistance from public service providers i s more progressive, and reaches a comparable number o f households indifferent income quintiles, while other public transfers are highly regressive, with more highincome households receiving these transfers both inabsolute terms and as a share o f income. Table 2: Transfers fromPublic Service ProvidingInstitutions and from PublicAssistanceprograms, by Quintile Numberof Share of All Number of Share of All' HouseholdsReceiving Households HouseholdsReceiving Households Transfersfrom Public Receiving Transfers from Other Receiving Service Providers Public Programs 16.1% 18.6% 19.3% 208 20.5% 104 258 25.5% Source: PublidPrivate Transfers and Social Capital Survey, 2003 178 Table 2a: Transfers from Public ServiceProvidingInstitutionsand from Public Assistance Programs,by Quintile Average Transfers As a Share As a Share of Average As a Share As a Share of Variable from Public of the Income before T~~~~~~ Tz;;;er Income before Service Transfer the Transfer Programs the Transfer Providers Quintile 1 474,786 9.3% 2.00% 1,038,14 4.6% 4.38% Quintile 2 559,550 10.9% 1.17% 7.2% 3.37% Quintile 3 771,830 15.1% 1.19% 13.0% 4.49% Quintile 4 1,108,184 215% 1.34% 15.6% 4.23% Quintile 5 2,332,030 45.5% 1.52% 13,267,97 59.1% 8.67% bIIhouseholds I 1,025,157) 1OO%l 1.38% 4,492,3611 100%1 6.04% Source: PubWPrivate Transfers and Social CapitalSurvey, 2003 B. Formal Private SocialFlows Formal private networks include associations such as trade unions, church groups, parent committees, neighborhood associations, agricultural associations, professional associations and NGOs. Overall, active membership inthese organizations i s fairly high, as 1,302 households, or 49.3 per cent o f the sample, report beingmembers o f at least one o f these groups. Poor households report lower associational membership than more well-to-do households. 33 per cent o f households from the poorest quintile report membership in an association compared to 60.5 per cent for the highest income quintile. The highest overall membership i s reported for Trade and Labor Unions (29.7 per cent o f all associated members), Owners' Associations (24.0 per cent o f associated members), and Agricultural Societies (20.5 per cent) (Table 3). Of these three, Trade and Labor Unions have the lowest membership among the poorer quintile (around 4 per cent o f households) and Agricultural Societies the highest (11.43 per cent o f these households). The latter i s comparable to membership from the second through fourth quintiles and i s 30 per cent higher than membership from the highest income quintile (Table 3). Table 3: Membershipof HouseholdMembersin Organizations Size of Member Organization Households (as % of the sample ) Trade Unionor -abor Union 29.7% Owners' association 24.0% Agriculturalsociety with legal personality 20.5% Parents' committee 18.0% Church committeeor otherforms of collectivechurch coordination 16.5% Politicalparty 15.2% ProfessionalAssociation 12.4% Family-TypeAgriculturalAssociation 10.8% OtherAssociations 10.3% Artists' /sports association 9.6% Traders or BusinessAssociation 8.7% Moneyrotatingsystem 8.6% NGO or civic group 7.1% On average, formal private associations do not seem oriented towards assisting the poor. Both cash and in-kindpayments and services are provided to average and higher income households, who are more likely to be their members, and are regressive as a share o f household income. The highest income quintile receives 2 per cent of total income from private associations, while the poorest 179 quintile gets only 0.3 per cent o f income from these transfers (Table 1). Socially vulnerable groups, such as female-headed households receive a significantly lower number o f transfers than non-female headed households. Pensioners do not get significantly more assistance than other household groups. The poor report beingisolated from those groups that were previously open to them: "Before, people were not layered as such, in these categories. Now the Forest Association also associates itselfonly with the rich, and everyoneelse is excluded. This is how our world is nowadays.''(11,average income respondents, Nereju, rural) "It is vey hard! I asked someoneIknew ifIcould borrow his tractor, which Iused to bring home some woodfor heating. Before, the communa used to give us wood, that was actually the reason why the communa owned the woods. And they used to write down the names of all thepoor people in a table, and sell it to themfor a reducedprice; they used to chop the wood in smaller pieces, measure it in cubic meters, then give it to the poor, so that it would be enough for everyone)...'' (Low income respondent,Alunis, rural) At the same time, the poor are active contributors to private associations - 75 per cent o f households from the poorest quintile contribute to different public groups, as opposed to 77 percent from the highestincome quintile(Table 4). Table 4: Contributions to Private Organizations, by Quintile Total i Urban ~ ~ Rural Variable Number llean (Lei) l ~Number IMean(Lel)~ ~ Number IMean(Lel) of income ~ a share Quintile 1 390 706,861 2.8% 93 768,752 2.7% 297 687,480 2.8% Quintiie 2 388 783,823 1.5% 176 714,326 1.2% 212 841,518 1.9% Quintile 3 384 824,964 1.2% 219 656,971 0.9% 165 1,047,937 1.6% Quintile 4 390 942,682 1.1% 281 972,354 1.0% 109 866,189 1.1% Quintile 5 397 1,105,023 0.6% 296 1,177,137 0.6% 101 893,677 0.6% All households 1949 873,743 1.0% 1065 903,996 0.9% 884 837,295 1.5% The church, through church groups, and the school, through parent committees, are the largest recipients o f contributions (Table 4 A in Appendix 1). While richer households are more often represented in church committees, the poor (particularly the rural poor) are the most active contributors to church causes (Table 4B and 4C). The vulnerable groups are equally likely to contribute to church causes (Table 4D). There i s no indication that churches distributemore support to vulnerable households (female-headed households and pensioners) than to average households. Table 4B. Table 4C. Contributionsto ChurchActivities Membershipin Church Committees Quintile 1 1 6.8%1 Quintile2 7.6% Quintile 3 335 63,7% 183 54,5% 152 80,0% Quintile4 Quintile 4 345 65,3% 225 5 a , 8 ~ ~ 120 82,8% Quintile 5 323 61,1% 239 55,7% 84 84,0% All HHs 1,714 65.0% a62 55,0% a52 79,5% 180 Table 4D. Contributions to Church Committees and Activities, By UrbadRural and Source. Public/Private Transfers and Social Capital, 2003 As is evident from table 4A inAppendix 1, church committees or other forms of collective church activities are receiving by far the largest share o f small contributions: about 20 per cent o f the sample contributes up to $3, while almost 50 per cent o f the sample contribute $15 or less. Parents' committees receive the next highest share of small contributions (98 contributions o f $3 or less, or 200 contributions o f $15 or less). Furthermore, the poor, specifically those in Roma communities, report high payments for religious rites and little support from local clergy. For the Roma these costs are also combined with discrimination. "The Church! That's another problem. Well, all the Roma here in the communa are Orthodox. For example, at Easter thepriest goes around to everyone's home to sprinkle their house with holy water. I owe the priest some money, and was not able to pay him back. The holidays came and went... and the priest did not stop by at all! He went to the neighboring houses, and he did not even bother to stop by to ask me how I was doing, or ifI wanted him to bless my house.(... )Ifyou have money,you pay thepriest, $you don't, he doesn't stop by, he doesn 't care about you." (Roma, low income respondents, Alunis, urban) "We are discriminated againstfrom all sides. All the local leaders discriminateagainst us. My little girl was sick and Iasked thepriest toplease baptize her. He said, until you make yourpayment, Iwon't baptize her!" (Roma, low income respondents,Alunis, urban) C.Formalprivate organizations Bothprivate service providers andprivate institutional economic actors provide some assistance to a closed group o f employees, clients and associates. These transfers are also highly regressive, with the highest income quintile receivingmore assistance from formal private institutional actors, who are most likely their employers or private service providers (Table 5). 1 1 Table 5: Flows from Private Businesses and Private Service Providers 13,291 ~:~~~~ 584,563 I;::::\ :f 72,908 0.11% Q4 521,567 0.60% 12,000 0.01% Q5 33 3,390,530 2.00% 69,035 0.04% otal 144 956,829 1.20% 28 27,922 '0.03% Source: PubWPrivate Transfers and Social Capital Survey, 2003 181 To conclude, official public flows form a very important source o f income for poorer households, the most important being social benefits and targeted social assistance. Disturbingly, testimonials single out the high cost o f accessing social assistance in some localities, the arbitrary nature o f resource allocation by local officials, the inability o f poorer localities to fund the local share o f expenses, and discrimination against the Roma by leaders and elite groups of the community (for a more detailed multivariate analysis o f social and fiscal determinants o fMIG flows see section 5). At the same time, it is important to note that other public flows are not pro-poor by nature. Formal private flows are highly regressive pointing to the closed non-altruistic nature o f the newly formed interest groups and associations. At present, the poor do not benefit from non-government-sponsored formal transfers, which makes them even more dependent on the state for assistance. D.Informal Income Informal incomes are vital for both urban and rural poor households. These flows include: 1) earnings from informal wages, small scale agricultural production and leasing o f land; 2) inter- household gift giving; and 3) exchange^.^ All households inthe sample participate insome type o f informal transactions. 12.2per cent of total income inthe poorest quintilecomes from informal sources.. The lowest share o f informal income i s 9.8 per cent for quintile 3, and the highest share i s 13.2 per cent for the highest income quintile (Table 1). Patterns o f formal and informal flows differ for rural and urban populations. The rural population receives more than one third o f their total income from informal wages, small scale agricultural production, and land leasing, while the share o f these items i s 20 per cent less for the urban poor, or 16.2 per cent (Table 1A and lB, Appendix 1). Small scale private plot production allows the rural poor to feed their families from own agricultural production and to sell this produce at local markets. Agricultural seasonal work i s also an important source o f income for rural landless households. Inurban areas, service day jobs are very important for the survival o f poor households, though the poor -the poor Roma inparticular- report that it i s increasingly difficult to find odd jobs (cleaning, carpentry, etc.) in towns, partly because o f the decrease insolidarity within ethnic groups and networks o f relatives: Low wages for odd service iobs "We work afull dayfor only 20,000 lei. I t is very difficult today to raise three children. They (the relatives) all have their own companies, but they will not hire us.'' (Female, low income respondent,Breaza, urban) "Seeds are the major source of income. We sell seeds. We wash, whitewash; I have sometimes earned 35,000 a day, and could not do anything with this money ... (Seasonal work... And this ifwefind offers for suchjobs. When Easter comes, we work (...) We work wherever we are offered jobs. I know old acquaintances who offer mejobs typically for one day only. We also find out about other jobs from newspapers." (Roma, low income respondent,Foqani, urban) *Since there is no prior tradition o f netting out payments inincome calculations, informal payments inTable 1 are included as payment inflows and added to "other sources o f income". For a more detailed analysis of informal payments see the section on inter-household transfers below. 182 Both the household level survey and focus group testimonials demonstrate that it i s easier to get ad hoc informal jobs in rural areas. There i s evidence o f self-organization o f the poor inteams of day laborers: "Those who are poor, who lack any assets, they have a hard time making ends meet. They make a living by working as day laborers, mainly in agriculture. They may also raise animals. Sometimes they form teams and go throughout the country seeking employment (in agriculture, on the fields.) They don't earn a lot of money. They typically come back with agricultural produce such as corn or money. Enough to cover some basic life necessities, but nothing more." (President of Nereju Community Organization, Nereju, rural) Participants o f poor Roma focus groups singled out summer agricultural employment as a particularly vital source o f income: "Everyone, everyone goes to work for food, as day laborers. Wefind employment, day by day, but not now, only in the summertime... I n the wintertime you die of starvation. I n the summertime, you can m anage. The b est timep eriodf or the R oma h ere is i n the summer, when you can breathe a little, when we go to Gheorgheni and pick blueberries and raspberries ... "(Low income respondents, Alunis, rural) Since the study i s more focused on formal and informal social transfers within open and closed networks, we will not concentrate on eaming in the informal economy (see chapter ...for more detail), and insteadwill study informal inter-household transfers. 3. Informal Inter-Household Flows The second informal source o f income is net informal inter-household transfers. These include: 1) altruistic gifts presented to and received from other households, and 2) exchanges o f goods and services between households. In total household income net gifts (inflows minus outflows) are negative for all categories o f households but the richest quintile. Net exchanges are also negative for poor rural households and for the second quintile o f urban households (Table 1A and 1B in Appendix 1). We now analyze informal flows in more detail, in order to: 1) assess the role o f informal incomes inthe livelihoods o f poor households; 2) determine the likelihood o f receiving and transfemng resources to other households; and 3) construct a map that relates the nature and the geography of informal transfers to household income, characteristics, and social environment. A. I nformal Inter-Household Flows by Category Informal flows are found to be substantial and widespread. Households in all income categories participate extensively ininformal inter-household transfers (almost 97 per cent o f all sampled households). Flows among households are significant when measured as a fraction o f household income before inter-household transfers. Gross informal outflows equal 12.3 per cent and gross informal inflows equal 8.5 per cent o f net income before inter-household transfers. Not surprisingly, these shares are higher for inflows and outflows for the poorest quintile, 17.8 and 17.6 per cent respectively. Ifwe compare these income shares to MIG-related transfers, we will see that informal gift-giving flows inabsoluteterms are 5 times greater than MIG-relatedtransfers! This comparison bringsout again the importance o finformal flows inthe livelihoods o fRomanianhouseholds in general and poor households inparticular. 183 Table 6A. Total Inter HouseholdTransactionsOutflows Quintile Inter-Household Gift Payment Exchange Income before Transaction Inter- Household Transaction Q2 7,067,152 4,738,569 1,677,287 651,296 52,400,00( 7,673,685 5,180,584 1,912,508 580,593 71,OOO,OO( 10,351,711 6,829,662 2,737,056 784,993 88,300,00( 13 5% 9.0% 3.2% 1.2% 10.8% 7.3% 2 7% 0.8% 11.7% 7 7% 3.1% 0 9% 184 Table 6B. Total Inter-HouseholdTransactions Inflows Quintile Inter-Household Gift Payment Exchange Incomebefore Transaction Inter- Household Transaction 4,860,970 2,382,542 1,426,012 1,052,416 27,600,00( 4,879,986 3,032,865 1,108,693 738,428 52,400,00( 5,443,322 3,189,966 1,428,305 825,051 71,OOO,OO( Q4 7,276,531 5,194,846 1,119,909 961,776 88,300,00( 12,315,270 9,108,721 1,754,592 1,451,957 170,000,00( 100% 49.0% 29.3% 21.7% 62.1% 22.7% 15.1% 58.6% 26.2% 15.2% 71.4% 15.4% 13.2% 14.0% 13.2% 16.2% 14.7% 12.8% 15.7% 13.9% 20.9% 16.4% 17.3% Q4 20.9% 22.7% 16.4% 19.1% 21.6% 1:: Q2 I 9.3% 5.8% 2.1% 1.4% 7.7% 4.5% 2.0% 1.2% 8.2% 5.9% 1.3% 1.1% 7.2% 5.4% 1.O% 0.9% 8.5% 5.6% 1.7% 1.2% Source: PubMPrivate Transfers and Social Capital, 2003 185 Gross transfers sent are about one third higher than the transfers received. This trend has already been noted among poor households in Romania in a study o f extreme poverty.6 The reason may be, as is borne by the data, that respondents recall the smallest gifts and loans extended to other households (the average value of outflows i s smaller than the average value o f inflows), and only the few and more significant contributions they have received from other households (the frequency o f inflows is lower than the frequency o f outflows). Similar perceptions were captured infocus group testimonials: "The neighbors won't help me anymore, everybody minds his own business. IfI ask for a cabbage he says that he doesn't have any. If anybody were to ask me for anything, Iwould give it to themfor free. "(v,low incomeperson, Breaza, urban) Since this study contains a much more detailed survey o f informal inter-household transactions than most o f the previous inter-household studies which were based primarily on national household budget surveys, it may be the case that this trend has become more noticeable with more careful differentiation o f various types o f inter-household transactions. Another outstanding feature o f this study is that the net effect o f inter-household transfers i s income neutral (Table 7). The results are similar for urban and rural households, and are not reportedhere. Table 7. Inter-householdTransfers and Inequality I i Decile2 35,300,OO 4.3% 34,900,OO 4.4% Decile3 49,200,OO 6.0% 47,300,OO 5.9% Decile4 55,600,OO 6.8% 54,100,oo 6.8% Decile5 70,600,OO 8.6% 68,400,OO 8.6% Decile6 71,300,OO 8.7% 69,500,OO 8.7% Decile7 85,000,OO 10.4% 82,600,OO 10.3% Decile8 91,500,OO 11.2% 90,600,OO 11.3% Decile9 114,000,OO 13.9% 113,000,000 14.1% 226,000,OO 219,000,000 Variable Income before Share Incomeafter Share Inter-Household lnter- Transfer Household Transfer Quintile I 27,550,OO 6.8% Quintile2 52,400,OO 12.7% Quintile 3 70,950,OO 17.3% Quintile4 88,250.00 21.6% Source:Publidprivate Transfers and Social Capital, 2003 "ItIS a proven fact that the poor underestimate the help they receive and overestimate the help they offer." (Source: Stanculescu, M.& Berevoescu, I.(coord.) (forthcoming). "Saracie Extrema, Tranzitia Traita inGroapa de Gunoi. Romania, 2001." (Extreme Poverty, Transition Lived in a Waste Dump. Romania; 2001.) p. 176. Bucharest, Romania: Research Institute for Quality o f Life.) 186 The finding that inter-household transfers are income neutral runs counter to the results reported inrelated literature where inter-household transfers behave like means-tested public transfers and flow from the rich to the poor (Cox, Jimenez, Oraska, 1996; Cox, 2002; Jimenez, Galasso, Cox, 2001). This outcome can be attributed to the more detailed nature o f inter-household transactions mentioned above as well as to the particular configuration o f interactions among households in post-socialist countries in general and in Romania in particular. It is also consistent with findings of sociologists' on patterns of informal transactions inpost-Socialist countries. Social ties are found to be built over time, to depend upon trust, and to require maintenance. The poor lack these resources and are often excluded (Pahl 1988, Morris and Irwin, 1992). The situation seems to be exacerbated by post-Socialist dislocation, in which old personal ties are strained by newly-developed income inequality on the one hand and functional informal ties based on the ability to obtain things in a shortage economy are rendered useless by the advent o f the market economy on the other (Ledeneva 1998, Wedel 1986). Focus group testimonials help identify some particular c haracteristics o f informal outflows from poorer households. These are: 1) the disproportionate transfers to children made by poor rural parents in the form o f agricultural produce or income from selling agricultural output produced on small plots, and 2) the high price paid by the poor for remaining in reciprocal networks: "My daughter is in college. Iraise a pigfor them, every month Igive them 40-50 eggs, I give them 500,000 - 1M lei because Ifeel pity for them." (Average income respondent, Alunis, rural) High cost o fremaining innetworks: Ifrou cannot rise up to a certain level, you arepushed aside. What does it mean, "to be pushed aside? " You cannot access their circles. They have many cars, and they have a lot of money. 'I (Average income respondents, Alunis, urban) On the positive side, measuredagainst other income before informal transactions, most transfers appear progressive (Table S).' 'Forthe definition o f progressive /regressive transactions used here see Social Protection chapter in this volume. 187 Householdtype NetInter- Informal Gift Payment Exchanges Total Income before Inter- Household Household Transactions Twnsactlons lei -39,493 -452,260 303,977 108,790 27,600,000 ~1 #ofHHs 504 482 338 325 518 lei -2,243,663 -1,705,705 -568,594 87,133 52,400,000 Q2 #of HHs 506 492 288 272 518 lei -2,391,444 -1,990,617 -484,204 244,458 71,000,000 Q3 #of HHS 496 481 273 247 518 lei -3,084,466 -1,634,816 -1,617,147 176,783 88,300,000 Q4 #ofHHs 498 484 244 242 518 lei -8,070,967 -4,183,539 -3,857,896 -4,513 170,000,000 Q5 #Of HHS 498 484 266 234 518 lei -3,164,701 -1,993,387 -1,244,773 122,530 81,900,000 le1 0.2% 4.5% -4.9% 17.8% 6.7% QI #ofHHs 20.1% 1%9% 24.0% 25.5% 20.0% le1 14.2% 17.1% 9.1% 14.2% 12.8% QZ *of HHs 20.2% 20.3% 20.4% 20.1% 20.0% lei 15.1% 20.0% 7.8% 39.9% 17.3% Q3 #Of HHs 19 8% 19.9% 19.4% 19.8% 20.0% lei 19.5% 16.4% 26.0% 28.9% 21.6% Q4 #ofHHS 19.9% 20.0% 17.3% 17.6% 20.0% lei 51.O% 42.0% 62.04/~ -0.7% 41.5% Q5 CofHHs 19.9% 20.0% 18 9% 17.3% 20 0% lei 100% 100% 100% 100% 100% Q5 #ofHHs 96.1% 93.4% 51.4% 45.2% 100% lei -3.9% -2.4% -1.5% 0.1% 100% -AllHHs Of HHs 96.6% 93.6% 54.4% 51.O% 100% The most popular form o f transaction i s gift giving with 93.6 per cent o f households exchanging gifts. Gift giving transactions are the highest in value among all inter-household transfers and are mildly progressive. Rural poor households perceive themselves as net givers, while urban poor households see themselves as net receivers o f gifts. This pattern i s consistent with the observation about gifts o f agricultural produce and pensions received from rural family members (Table 1A and 2B inAppendix 1). Cash Davments for minor services rendered to other households are the second most important transactions in terms o f mean value per household and the number o f transacting households (54 per cent o f households participating). Rural households are more active in these transactions (on average 73 versus 46 per cent respectively, Table SA and 8B, Appendix 1). The rural poor perceive themselves as net winners, while all urban and richer rural 188 households see themselves giving up more resources than they receive. Payments as a share o f income before inter-household transfers are mildly progressive as well. Exchanges of goods and services are more widespread in rural areas, where cash transactions seem to come at a premium- 67 percent o f rural versus 40 percent o f urban sub- samples (Table 8A and 8B, Appendix 1). These exchanges are highly progressive, with both the rural and the urbanpoor seeing themselves as net winners from these transactions. There are more rural poor households participating in inter-household payments and exchanges than urban poor households. This pattern may reflect closer relationships among neighbours in rural areas, as well as exchanges and payments for small agricultural services and produce (Table 8A and 8B, Appendix 1). Surprisingly, informallending i s found to be regressive. More households from the highest income quintile inrural areas borrow informally than from the lowest quintile (the opposite is true o f urban areas). This finding i s backed by statements made during focus group sessions indicating that the poor fear indebtedness and prefer to do without rather than borrow: "Ihavenobodytohelpme,my husbandworksaswell,Ihavetohavemilkeveryday. You can borrow moneyfrom a neighbour, but it's quite a shame to ask a second time. Icanbuyfrom thekioskoncredit,theownerknowsme,theydon'thelpeverybody,she understands that Idon't have money. (Low income respondent, Galati, urban) I' "Icanbarely makeendsmeetfinancially.Ihardly evenhaveenoughmoneytobuy medicine. Ihave nobody to help me. I don't dare go and buy on credit, Itry toface up to the situation according to my possibilities. " (Low income respondent, Galati, urban) "Generally, you do not borrow. I t is better to give up things." (Low income respondent, Alunis, rural) Table 8C - NetInformalLending Househokf Type MeanValue of As aShare of Inter- As a share of Total Income Transactions HouseholdTransactionsby Before Inter-Household Quintib Transactions I r;9% .n 189 Table 8D. Net Informal Lending (Outflows and Inflows) Q5 I 6,764,7991 52.2%1 4.0%1 6,034,073 50.9%1 3.5% All HHs I 2,591,4021 100.0%~ 32%1 2,369,268 100.0%/ 2.9% Source: PublidPrivate Transfers and Social Capital Survey, 2003 T h i s may be attributed, o n the one hand, to the lack o f tradition o f professional informal lending in post-Socialist countries, and on the other, to traditional reliance on the state for assistance. The poor, particularly the rural poor, (Table 9 A and 9B in Appendixl) lendmore than borrow, which may partly be attributed to the desire to stay in informal networks even if such membership comes at a cost. At the same time, participation ininformal lending is fairly high. Halfo f all poor households participate in informal lending. Taken separately, informal outflows and inflows equal approximately 4 per cent o f total income for poor urban households and 3 and 3.1 per cent respectively for rural households (Table 8E and 8F, Appendix 1). These values are comparable to the value o f MIG-related transfers. As a share o f income, negative balance for the year is fairly small (0.6 per cent of total income) and may represent part o f a dynamic borrowing and repayment pattern (Table 8F inAppendix 1). These appear to be consumption- smoothing, short-term loans taken against in-coming MIG or other public transfers that come with predictable periodicity. Therefore, lending appears an important coping strategy for poor households and part o f a consumption-smoothing, risk-sharing exercise. B. Maps of Inter-Household Transfers To better understandthe current paradigm o f exchanges among households, we examine its components: 0 Channels of exchanges - relationship among transacting households (relative, friend, other associate) 0 Modality of transactions -altruistic vs. reciprocal exchanges 0 Form of transactions -cash versus in-kind 0 Geography of exchanges - distance o f transactions, the same location versus other urbadrural localities inRomania or abroad. Frequency and variety o f informal transfers is much greater for richer households than for poorer households and geography o f inter-household connections i s much more extended. Richer households operating in the cash e conomy carry out more informal transactions in cash. Poorer households conduct more exchanges ingoods and services. Channels and Forms of Informal Transactions -Prohibitive Cost of Friendship Channels and forms o f informal transfers differ dramatically between high income and low income households. Poor households transact primarily with their neighbours, people likely to be in a similar situation, or with providers o f odd jobs. These exchanges are predominantly reciprocal, rather than altruistic. The richer households can "afford friendship" with more 190 informal transactions carried out between friends, and more transactions defined as gifts. (Table 13, Graph 1A and 1B) Table 13: Average Number of Transactions per Household, by Relationship IItype oftransfers elativesFriends Neighbors omebodyelse otal Graph 1A Graph 1B 1 Average number of transaction cases for 1 Average number of transaction cases for I all interhousehold transfers by type of i all interhousehold transfers by type of relationship (from HH) relationship (to HH) Very poor (0-4 USD) 0 RtCh (15 ic) I i mVerypoor(04USD) ORlch (E++) 3 5 - 3 0 0 5 lo 0 relatives friends neighbors somebodyelse i relatives friends neighbors somebody else Source: Public/Privaie Transfers and Social CapitalSurvey, 2003 Households from the poorest income group (from $0-$4 in ppp terms per equivalent adult) transact much less with relatives, both interms of frequency of exchangesand total volume of transactions, than richer households (Table 14). The average o f 2 giving transactions with relatives for the poorer households as opposed to the average of 3.5 giving transactions for high income households; the average o f 1.3 receiving transactions for the poor versus 2.3 receiving transactions for high income. Exchanges with friends are twice as numerous for high-income householdsthan for the poor. The most likely partner in informal exchanges for the poor are their neighbors. The poor record 40 per cent more giving transactions with neighbors than the high income households and twice as many receiving transactions. Another likely partner for the poor is an associate, which can be explained by the fact that the dominant transaction for the poor is through exchange, rather than altruistic gift giving, as is the case for richer households (Table 14). "To go and askfrom the rich... no, we don't do that. Wego to thepoor. Whenyou are poor, you only go to those who are poor, whenyou are rich, you go to those who are rich. For ifyou were to go and askfor helpfrom the rich, they will only laugh atyou. " (Low income respondents,hlereju, rural) The Roma poor report isolationwithin their neighborhoods: "...There are afew (Roma) families round here, on the street that goes up... We are like isolated mountains, I swear. We have no water, no roads, we have young 191 children... The neighbors won't let us get water from their wells." (Roma focus group, Alunij) ...andhigh levels ofmutual assistanceamong poorRoma: "The people in the ghetto help each other. We all rush to help each other. The kids do the same. We don't go to everyone. We know somepeople. In the end, it is the poor that help each other. " (Roma focus group, FocSani) The poor are much less involved in altruistic exchanges than high income households, but are more involved in gift giving with others (neighbours, other associates) - people in their immediate environment. Table 14: The Structure of Inter-household Transactions, by Type of Transactions (YOof total transactions) ittor Donations Relatives Friends Neighbors Somebody Else Total ousehold economicstatus From HHTo HH From HH To HH From HH To HH From HHTo HH From HH To HH Source. Public/Private Transfers and Social Capital Survey, 2003 The poor testify to isolation from old friends and to inability to sustain old contacts: "Relationships have, yes, cooled off among people, because of the differences between... how should Iphrase this, differences in financial means between people. We had somefriends who started their own business... now they are doing a lot better than w e are, and the relationships between us have cooled ofl because you feel at some point that you cannot tfloat' at the same level with them, and then you try to contact them less often. They see life differently, it is easier for them to make ends meet... that's why we it cooled off"(TM, average income respondent, Breaza, urban) "Today'spoor are in such terrible shape. (...) Because thosewho arepoor today were ourfriends from yesterday, they are the ones who lived next door to us. They are also their (our former neighbors 7 children. And these are all relatively educated people. (...) They arefaced with terribly embarrassing and humiliating circumstances. They cannotpay their bills. (...) Somefind understanding from people, institutions), others do not. (average income respondents, Focsani, urban) " "Ifyou had money you would be everybody's friend, because then you could help eveiybody. But $Idon't have money I cannot even step outside my house. I had friends and I helped them. But now, when they see that I can't work and I can't earn money they don't want to know about it, they don't want to hear anything or see 192 anything....I cannot go and borrow money from a former colleague, whom I used to help before ...he won't help me, he will say that he is unable to help me, that he can 't give me anything. (Average income residents, Alunis, rural) The rural poor t estify to the breakdown o f traditionalties o f cooperation and traditions of assistance to the poor within communities: "During Communism people helped each other more. When something happened to Mr. Ion, everyone wouldjump to his help. Now if something were to happen to him, everyone says, 0) Method Tobit Probit OLS Intercept -16.306 -- -25.097 (-1.41) (-1.60) Number o f adults -3.022 -.0297 -2.318 inhousehold (-5.11) (-2.87) (-3.68) Number o f 1.015 .0151 0.556 children (1.62) (1.35) (0.91) Age o f subject -0.897 -.0087 -0.720 (-4.07) (-2.50) (-4.59) Age squared 0.006 ,0001 0.005 (3.07) (1.78) (3.96) Log of income 2.988 -.0189 3.268 (exc1. transfers) (4.66) (-1.47) (3.1 1) Asset index 0.471 -.0003 0.624 (3.13) (-0.09) (3.44) Owns land 5.333 .0478 4.126 (4.18) (2.21) (2.47) Owns home -2.688 -.0154 -2.498 (-1.75) (-0.70) (-1.74) MIGcash income 0.186 .0031 0.150 (1.60) (1.16) (1.78) Other social 0.362 -.0009 1.086 assistance (6.38) (-1.24) (1.88) Pensioner in -4.452 -.0579 -3.047 household (-2.28) (-2.30) (-2.39) Rural -0.821 ,0516 -2.144 (-0.53) (1.54) (-1.33) Rural pensioner 5.033 .0339 3.543 (2.19) (0.93) (2.42) Public activism 2.074 .0156 2.062 index (2.45) (1.24) (1.95) Regression stat. Chi2=249.7 chi2=8 1.2 R2=.15 N 2590 2590 2143 All dependent variables are inmillions oflei. T or Z statistics are inparentheses. Probit coefficients are marginal effects evaluated at mean o f all other regressors. 234 Table 2: Inter-household Transactions Gross Outflows and Net Inflows Regressions Equation 1 2 3 4 Transfers Gross Gross Gross Net Variable outflows outflows > 0 outflows Inflows (if>0) Method Tobit Probit OLS OLS Intercept -78.485 -- -74.501 52.389 (-5.78) (-3.20) (3.39) Number o f adults -3.236 -.0014 -3.237 0.967 inhousehold (-4.74) (-0.43) (-3.67) (1.07) Number o f 1.335 ,0029 1.227 -0.567 children (1.83) (0.75) (1.22) (-0.62) Age of subject 0.128 .0004 0.140 -0.848 (0.50) (0.29) (0.70) (-3.93) Age squared -0.002 -.0001 -0.002 0.007 (-.62) (-0.57) (-0.85) (3.69) Log of income 4.720 .0026 4.576 -1.470 (excl. transfers) (6.29) (0.80) (3.01) (-1.39) Asset index 1.387 .0029 1.338 -0.863 (7.96) (2.86) (3.49) (-2.38) Owns land 3.085 .0146 2.480 1.304 (2.09) (1.43) (1.93) (0.87) Owns home -3.662 .0055 -3.984 1.533 (-2.04) (0.51) (-2.29) (0.96) MIG cash income 0.104 .ooo1 0.100 0.049 (0.77) (0.13) (1.45) (0.56) Other social 0.158 -.0005 0.567 0.187 assistance (2.3 1) (-2.85) (3.48) (0.80) Pensioner in -4.584 -.0245 -3.416 0.991 household (-2.04) (-2.22) (-2.53) (0.93) Rural 3.620 .0276 3.039 -4.701 (2.00) (2.25) (1.14) (-1.68) Rural pensioner 4.187 .0147 2.939 0.478 (1.58) (1.06) (1.35) (0.19) Public activism 2.320 .0026 2.366 -0.630 index (2.35) (0.39) (1.87) (-0.62) Chi2/F statistic chi2=257.5 chi2=l 56.2 Rz=.10 R2=.02 All dependent variables are inmillions o flei. T or Z statistics are inparentheses. A ** and N 2590 2590 2482 2590 * indicate significance at .01and .05 respectively for 2-tailed tests. Probit coefficients are marginal effects evaluated at mean o f all other regressors. 235 Table 3 shows determinants o f MIG assistance (cash benefits, means-tested heating subsidies and emergency relief), in millions o f lei. Tobit regression is used because the dependent variable is truncated at 0. 0 nly 4 24 o f the 2 521 households inthe sample received MIG assistance. Social assistance, inter-household transfers and local revenues per capita are all in millions o f lei. Equation 1 analyzes the determinants o f MIG assistance for all households, urban and rural. Equations 2 and 3 respectively divide the sample into urban and rural sub- samples. Equations 4 and 5 include all households, but separate MIG assistance into its two main components, cash benefits (equation 4) and heating subsidies (equation 5). In all equations, MIG assistance is unrelated to the volume of inter-household transfers: neither inflows nor outflows are significantly associated with MIG. Net inflows (inflows minus outflows) are also unrelated to MIG, when substituted (in results not shown) for inflows and outflows. Potentially, inflows converted to assets could still reduce MIG assistance through the significant assets variable in the regression. However, the inter- household transactions variables all remain insignificant when the assets variables are omitted from the model. Further tests not reported in the table divided urban from rural households, and found no evidence in either sub-sample that inter-household transactions affected MIG assistance. These results are consistent with the possibility that private transfers do not "crowd out" public assistance, neither by reducing the likelihood a household applies for MIG nor by affecting its eligibility. However, these coefficients could be capturing both a (negative) "crowd out" effect and the (positive) effect o f otherwise unobserved characteristics associated with need. Only if we could control fully for a household's need for public assistance, with variables other than inter-household transfers, could we confidently attribute the inter- household transfer coefficients to the effects o f crowd out or absence o f "crowd out". Age (of respondent) i s also unrelated to MIG assistance in all equations in Table 3 (and remains insignificant when a linear function o f age i s substituted for the quadratic). Households with more children receive significantly more MIG benefits, controlling for other variables. This result holds for the full sample (equation l), for urbanhouseholds (equation 2) for rural households (equation 3), and separately for MIG cash benefits (equation 4) and heating subsidies (equation 5). Benefits are supposed to be based in part on number o f children, so these finding are unsurprising, although still reassuring. Higher income (per adult equivalent, measured exclusive o f inter-household transactions and social assistance) i s associated with significantly lower MIG receipts, for the full sample (equation l),for urban households (equation 2), and rural households (equation 3). Surprisingly, MIG cash benefits are unrelated to income (equation 4), although heating subsidies are higher for lower-income households (equation 5). Households with more assets also receive lower MIG benefits. Asset types were divided into two categories: 1) goods that are potentially income-producing, or luxury goods, and are generally taken into account when MIG administrators evaluate applicants; and 2) other assets that generally are not considered.I6 As expected, the first type o f asset i s more strongly associated than the second type with reduced MIG benefits. Land ownership also reduces l6The first group includes cars, vans, motorcycles, agricultural equipment, animals, CD players, cell phones, ' refrigerators, freezers, microwave ovens, computers and internet service, and color television. The second group includes radios, tape players and stereo systems, vacuum cleaners, telephones, cookers and stoves, black and white television, cable TV, video players, book collections, washing machines, sewing machines, and bicycles. 236 MIG benefit^.'^ The coefficients on income and assets (including land) are higher for the rural sub-sample than for the urban sub-sample, suggesting better targeting in rural areas. Officials in rural areas, despite more limited administrative capacity for investigating applicants, might have more accurate information on their standard o f living. Although there are sizeable earmarked transfers from central government for social assistance, the revenue measure most strongly associated with higher MIG benefits is locally- raised revenues per capita. This variable is significant infour o f the five tests (equation 3 for rural households i s the exception). Localities are expected to cover about 20% o f the cost o f MIG from local revenues. There is wide variation in locally raised revenues, however, and many localities - particularly smaller and less wealthy localities which are able to raise fewer revenues locally - do not even come close to covering their 20% share o f costs. Public activism at the local level could be associated with improved accountability o f local government, which must consider the possibility citizens will protest in various ways if services are inadequate. To the extent local officials have control over MIG funds and implementation, households residing incommunities with greater activism may receive higher MIG assistance. A household's own activism matters, however, only to the degree it contributes to the "public good" o f monitoring local government. Therefore, a household's own level o f activism should be unrelated to MIG assistance, controlling for the mean community level. The household's own level o f civic activism turned out to be unrelated to MIG assistance, as expected, and this variable was omitted from the models reported in the tables, for space reasons. The community level mean o f the index, however, has a positive and significant (inequations 1,2 and 5) coefficient. Number o f years the mayor has been in office i s positively and significantly related to MIG assistance (equation 1). More experienced mayors may be more effective in lobbying higher levels o f government for transfers to local government budgets. However, this variable remains significant when those transfers are controlled for. More experienced mayors may administer social assistance programs more effectively, or, alternatively, those who manage to run more successful programs may increase their chances of re-election. Mayoral tenure is positive and significant in the urban (equation 2) and rural (equation 3) sub-samples, and for heating subsidies (equation 5), but not for cash benefits (equation 4). Households residing in localities with PSD mayors receive significantly higher MIG assistance (equation 1). This results holds for households in rural localities (equation 3), where 1imited locally-raised revenues makes town halls more dependent o n transfers from central and county government, which i s controlled by the PSD. The result does not hold for urban areas, which are less dependent on such transfers. However, the PSD effect remains when transfers from central and county government are controlled for (inadditional tests not reported in the table), and those transfers are not significant predictors o f MIG benefits received by households. It i s therefore somewhat o f a mystery why MIG benefits are higher inPSD-governed localities. Residents o f peripheral villages are often thought to be disadvantaged politically, relative to residents o f communal centers, where the town hall is located. In tests not reported in the table, no difference was found, however, between MIG assistance levels in central and peripheral villages. ~ "Inadditionaltests,homeownershipwasfoundtohavenoimpact. 237 Where poor people have more influence over local government policy, it i s natural to expect public assistance to be more generous. A survey ofpublic officials included a question asked about the level o f influence poor people (and various other groups) exercised over decisions by the town hall and local council. Where poor people were judged to have more influence, MIGbenefits are higher (equation 1). This variable i s not significant, however, when the sample i s broken down into urban and rural sub-samples (equations 2 and 3). It is a significant predictor o f heating subsidies (equation 5), but not o f cash benefits (equation 4). Several s tudies, mostly based on U.S. data, have found that s upport for redistribution and provision o f public goods is lower in more ethnically heterogeneous communities. In the 51 localities represented inthis study, the Roma population in the last census varied from 0% to 11.7%, averaging 1.9%. The effect o f Roma on MIG assistance could be positive (if they tend to be poorer, and their poverty i s not fully captured by the other regressors) or negative (if islesssupport forassistancewheremoreofitislikelyto gotoadifferentethnic there group). Percent Roma i s insignificant in most tests in Table 3, but it is positively and significantly related to MIG cash benefits in equation 4. In tests not reported in the table, percent Hungarian population (which varies from 0% to 99.3%, averaging 10% in the 51 localities) showed no consistent relationship with MIG assistance. I t i s possible that discrimination could occur within communities. For example, independently o f the ethnic composition o f the locality in which they live, Roma households could receive lower MIG benefits, controlling for income, assets, and so on. Unfortunately, there i s no way to identify Roma households in the data. Hungarian households can be identified, however. Hungarian ethnicity does not significantly lower a household's MIG benefits, even inoverwhelmingly non-Hungarian localities. Controlling for the effects o f other variables, cash benefits are somewhat higher for rural households (equation 4), but heating subsidies are larger for urban households (equation 5). The bottom row o f the table shows the means o f the dependent variables. The average MIG assistance for all households i s 819 thousand lei per month (equation l), most households but receive 0 ;among the 424 households receiving MIG benefits, the average is 4.87 million. The average is higher for rural than f o r urban households: 1.001million compared to 6 99 thousand. For the full sample, cash benefits averaged 451 thousand (equation 4), and heating subsidiesaveraged 159 thousand (equation 5). To summarize, there is very little evidence in these tests that the benefits to the poor from inter-household transactions are offset by accompanying reductions in their public assistance benefits. These tests are not conclusive, however, as they cannot fully control for factors that may affect either eligibility for MIG o r gifts and informal loans received. There i s s ome reassuring evidence that MIG works as planned: households with more children, lower incomes, fewer luxury assets or assets with income-producing potential, receive more assistance. Benefits also tend to be higher where there is more civic activism, and where the poor have more influence over local political decision making. More unsettling results, however, are the lower level o f benefits in localities that raise fewer revenues locally, and higher benefits in localities with mayors from the PSD or who have been in office longer, suggesting that political factors influence allocations. 238 Table 3: MIG TransfersRegressions Equation 1 2 3 4 5 Samde All Urban Rural All ~~- All MIGcategory All All All Cash benefits Heating subsidies Intercept 2.931 -2.116 7.496 -14.139 -1.396 (0.47) (-0.40) (0.47) (-1.86) (-0.60) Gross inter-hh 0.010 0.006 0.043 0.008 0.003 inflows (1.18) (1.18) (0.61) (0.91) (0.90) Gross inter-hh -0.017 -0.009 -0.131 -0.005 -0.010 outflows (-1.11) (-1.00) (-1.03) (-0.35) (-1.21) Age of subject 0.026 0.125 -0.437 -0.012 0.039 (0.17) (1.11) (-0.99) (-0.07) (0.68) Age squared -0.001 -0.001 0.003 -0.001 -0.001 (-0.22) (-0.96) (0.72) (-0.24) (-0.44) Numberof 2.196 1.545 2.745 1.920 0.589 children (4.99) (4.75) (2.23) (4.06) (3.60) Log of income -1.022 -0.670 -1.097 -0.227 -0.261 (excl. transfers) (-3.92) (-2.76) (-1.85) (-0.69) (-2.74) MIG-relevant -1.324 -0.807 -1.778 -1.185 -0.375 assets (-4.09) (-3.5 5) (-1.72) (-3.28) (-3.07) Non-MIG- -0.191 -0.088 -0.961 -0.214 -0.155 relevant assets (-0.93) (-0.60) (-1.50) (-0.93) (-2.00) Owns land -2.592 -1.200 -5.062 -1.941 -0.963 (-2.51) (-1.60) (-1.73) (-1.69) (-2.47) Locally-raised 4.880 3.173 -0.352 3.073 1.697 revenues p.c. (3.64) (3.03) (-0.06) (2.11) (3.34) Public activism 9.227 8.080 9.871 4.524 3.018 (locality mean) (3.07) (2.66) (1.35) (1.39) (2.42) Tenure inoffice of 0.689 0.320 1.ooo 0.089 0.280 mayor (5.68) (3.15) (2.73) (0.65) (5.87) PSD mayor 2.284 0.260 6.120 0.351 0.874 (2.44) (0.35) (1.85) (0.34) (2.46) Poor influence 1.301 -0.210 0.514 0.478 0.493 local decisions (1.93) (0.29) (0.32) (0.65) (1.75) Percent Roma -3.149 -2.949 7.905 41.585 -8.544 (-0.20) (-0.26) (0.13) (2.50) (-1-44) Rural -0.539 -- -- 2.357 -2.754 (-0.3 9) (1.54) (-4.92) pseudo R2 .04 .05 .04 .03 .10 Observations 2521 1529 992 2521 2521 Observations > 0 424 299 125 210 250 Mean, dep. var. 0.819 0.699 1.001 0.451 0.159 Method i s tobit regression. Dependent variable i s MIG assistance received, inmillions of lei. Inter-household transfers and local government revenues are also measured inmillions o f lei. T statistics are inparentheses. 239 Dependent Client satisfaction Corruption Trust inthe mayor variable 1 2 3 4 5 6 Intercept 67.6 119.2 -162.2 -133.6 3.02 4.12 (0.930 (1.92) (-2.72) (-2.83) (5.55) (9.98) Expenditures 1.59 1.35 0.17 0.40 0.58 -0.67 per capita (3.86) (3.69) (0.25) (0.60) (0.23) (-0.24) Logof locality - -1.98 -2.02 0.09 0.12 -0.12 -0.14 population (-2.18) (-2.40) (0.11) (0.16) (-2.55) (-2.68) Logof mean 0.62 -1.46 9.98 8.34 0.27 0.19 locality income (0.140 (-0.38) (2.89) (2.95) (1.23) (0.77) Neighbor 52.2 8.25 3.50 relations index (2.59) (0.48) (3.14) Ethnic -15.5 13.93 -.60 fractionalization (-2.09). (2.29) (-1.24) R' .39 .35 .31 .38 .31 .19 Mean. den var. 76.3% 20.7% 2.88 Sample size i s 50. T-ratios are inparentheses. 240 Appendix 3 Sample and SampleMethodologv The study used qualitative and quantitative methods to collect community and household-level data. During the preparatory stage, three focus groups and five open-ended interviews were conducted with rural and urban poor, as well as with public and private service providers. Duringthe second stage, the following surveys were carried out: a household-level survey and a local officials survey. Seventeenfocus groups and individual semi-structured interviews were conductedwith poor and non-poor rural and urban inhabitants, local officials, and public and private service providers, as outlined below. A locality card was filled for each locality enrolled inthe sample. The cardcontainedsocial, demographic, and economic data as well as specific informationabout budgetsand social services provision. The Household Survey: Sample size: 2641 households. Sample type: stratified, probabilistic, three-stage sample. Stratification criteria: 18 geographic areas based on historical regions, residence (urban- rural), urban locality size (4 types), degree o f development o f rural localities (3 categories). Sampling: probabilistic selection o f localities (5 1 total, 27 rural), sample units (streets, 264) and households. Households were selected by random route method. Forty households in rural areas and between SO and 80 households in urban areas, depending on locality size, were enrolled in order to make the connection between household and locality levels possible. Representativity: the sample is representative for Romanian household population, with a maximum sampling error o f Data was not weighted. In-depth interviews were carried out in "face to face" sessions. The Local Officials Survey: Sample sue: 200 Respondents: mayors, deputy mayors, local councilors, other local officials Sampletype: theoretical Focus grouw and semi-structured interviews: Seventeen focus groups and 33 individual in-depthinterviews were carried out insix localities. The localities were selected from three counties in different provinces. Localities differed by urbadrural, level of economic development (poor versus rich communities), and ethnicity. Participating localities are as follows: Breaza, Nereju, Alunis, Focsani, Galati, Tirgu Mures. Structure o f focus groups: 9 Focus groups with the poor: 6, one ineach locality. Focus groups with Roma: 2, inAlunis and Focsani. Focus groups with average people: 6, one ineach locality. Focus groups with public and private service providers, and local officials: 3, inFocsani, Galati, TirguMures. 241 1 Five in-depth individual interviews were conducted with local officials (mayors, deputy mayors, local councilors) in: Breaza, Nereju, Alunis, Focsani. 1 Twenty in-depth individual interviews were conducted with public and private service providers (social workers, teachers, priests, physicians, managers of . NGOdcivil organizations) distributed inall surveyed localities. Eight in-depth individual interviews were conducted with poor people: distributed inall surveyed localities. 242