RCEP, WP  no.7/October 1999






Poverty, Inequality and Social Protection










International Management Foundation










Cornelia Mihaela Tesliuc


Lucian Pop





0         Introduction and Executive SummaryAbstract..................................................................................................................................................................... 221

1       Poverty and Inequality in Romania.......................................................................................................... 443

1.1        A World View: How Poor Are The Romanians?............................................................................. 443

1.2        Poverty and Inequality: Levels and Trends During Transition...................................... 665

2       Who Are the Poor?............................................................................................................................................... 997

2.1        A Profile Of The Poor.................................................................................................................................... 997

2.2        Multivariate Analysis of the Correlates of Household Consumption.................. 131311

2.3        Poor Employees and Pensioners: The Largest Poverty Groups........................................ 161614

3       Two Kinds of Poor: Transient and Permanent Poverty.......................................................... 181816

4.      Effectiveness of the Social Safety Net in Tackling Poverty.............................................. 212119

4.1        The Romanian Social Safety Net....................................................................................................... 212119

4.2        The Social Safety Net’s Impact on Poverty................................................................................. 232321

4.3        Households and the State: Inequality and Redistribution............................................ 282826

5       Policy Conclusions And Recommendations.................................................................................... 323229

6   BibliographyREFERENCES........................................................................................................................................................................ 333330

7   Annexes............................................................................................................................................................................... 343432

7.1   Data Sources............................................................................................................................................................... 343432

7.2    Methodology for Poverty Measurement.................................................................................................. 353532

7.3  Statistical Annexes: Tables and Regressions Estimation............................................................. 383835

0  Introduction And Executive Summary  22

1  Poverty And Inequality In Romania  44

1.1  A World View: How Poor Are The Romanians?  44

1.2  Poverty And Inequality: Levels And Trends During Transition  54

2  Who Are The Poor?  74

2.1  A Profile Of The Poor  74

2.2  Multivariate Analysis Of The Correlates  Of Household Consumption  104

2.3  Poor Employees And Pensioners: The Largest Poverty Groups  124

3  Two Kinds Of Poor: Transient And Permanent Poverty  144

4  Effectiveness Of Social Safety Net In Tackling Poverty  164

4.1  The Romanian Social Safety Net  164

4.2  The Social Safety Net Impact on Poverty  174

4.3  Households and the State: Inequality and Redistribution  214

5  Policy Conclusions And Recommendations  244

6  Bibliography  254

7  Annexes  274

7.1  Data Sources  274

7.2  Methodology for Poverty Measurement  274

7.3  Statistical Annexes: Tables and Regressions’ Estimation  Error! Bookmark not defined.Error! Bookmark not defined.



Introduction aAnd Executive Summary

In everyeach society, there are rich andas there are poor. The focus of this paper is on the latter category, the poor, whicha category that grew in Romania in number and severity at unprecedented levels during transition. Even byunder a conservative definition of poverty, we found a six-fold times increase in the number of poor during the transition decade. This paper[1] investigates why poverty increased so much, and what can be done to alleviate it.

This increase in poverty is common to many transition economies, and it is typically considered a by-product of the reform processes. Most economic reforms have their victims, despite good intentions. Typically, reforms are implemented to change the current situation for a better one, judged by some “welfare compensation criteria”. Some reforms are defended by saying that for a better tomorrow, some should suffer a bit today. These reforms are usually labeled “macrostabilization” programs. Such programs are triggered by unsustainable excesses of aggregate spending over domestic production. The measures designed by economists and implemented by policy-makers to “restore macroeconomic stability” try to curb aggregate domestic demand, that means in most cases a reduction in the overall level of consumption – hence welfare --, in that country. A lot of people would see their welfare falling.

Proponents of another type ofOther reforms arguewould say that changes that negatively affect some particular groups are beneficial to the overall economy, because those who gain, gain more than those who loose[i]. These are called “structural adjustment” programs. These programs promote changes at the sector level, by restructuring some inefficient operations or by changing or discontinuing some inefficient practices, with the belief that the benefits accruing to the rest of the economy are larger than the pain of the “victim” sector. In this case, some groups whose earnings depends on the shrinking sectors would see their welfare falling.

Among transition economies, fast reformers succeeded into restoringe their economies on a path of sustainable growth path, while hesitant reformers did notdon’t. The link between positive growth and household welfare is trivial. At micro level, the growing transition economies cured much of the “transient poverty” and, thanks to sizablesizeable social budgets and lower demand for social assistance, alleviated much of the “permanent” poverty as well. The other transition economies, Romania included, are still far from this target. We illustrate, in the first chapter, the marginal impact of “absolute” poverty in countries like the Czech Republic, Slovakia, Poland or Hungary, contrasted with itsthe high magnitude and severity in Romania, Bulgaria and republics of the fFormer Soviet Union Republics, Romania and Bulgaria. This speaks toabout the importance of finalizing the reform agenda in Romania, and the desirability of an increase in the pace of reform marching speed.

MovingLanding from this regional perspective toin the situation in Romanian realities, the paper documents how much poverty grew in the last decade. If the analysis ended here, Left to this topic, our paper would add no value-added to the research on Romanian research, as 1999 has beenis a relatively abundant year forin poverty-oriented studies[ii], in contrast with the situation two or three years ago. Such relative abundance of poverty studies raises the question: why more? Apart from its “regional perspective” on poverty and inequality, we investigate three topics that are, to a large extent, new to the social and economic research:

·        First, we quantified the contribution of various factors to the household welfare, and investigated why the poor are poor. Leaving the unobserved or latent “macro” factors aside, we go down to microeconomic datasets to investigate what determines household welfare. We build a model to unveil the factors that are associated with poverty, and to identify the policies that can help poor get out the poverty pool. We found that those in poverty are those less endowed with human and physical assets, unable to cope with the income shocks that have accompanied the transition period. Some of them may exit from the poverty pool, mainly if their human capital improves. For others, especially for those lacking labor resources, poverty may be permanent if social safety nets fail to elevateshield them from poverty.

·        Second, thanks to new opportunities opened by improved data collection, we investigate the dynamics of poverty, by looking at the entry into and exit from the poverty pool. This exercise is of immediate interest for social policy, because we identify the groups that were able to cope with the hardship of transition, and contrast these with the ones that fared worse overin time. As expected, we found that poor households headed by employees or pensioners are more able to exit from poverty than the others, notably those whose heads are e unemployed-, farmers- or self-employed-headed households. Social policy-makers should pay more attention to the latter category.

·        Finally, we investigate the effectiveness of the Romanian social safety net in shielding the poor against poverty. Notably, we are asking to what extent social programs tacklehow much poverty is tackled thanks to social programs, if the social programs provide adequate coverage to the poor and if the funds are reasonably well targeted to those in need. We found that the overall social spending almost halves poverty in Romania, and that most social programs with a poverty alleviation focus are well targeted to the poor. However, the targeting technique of the main programs is to addresses narrow welfare risks such us unemployment, or temporary loss of income due to illness, childcare or handicap. The only program with a broader focus – the social aid program – suffers from lack of funds and excessive discretion in implementation. Consequently, the coverage of the anti-poverty programs in Romania is inadequate. To illustrate this, we showmention at this point that the percentage of poor households that do not receive any support fromthrough social programs is lowerhigher than the figure for those infor the not-poor category. The fact that the programs are well targeted means that new funds are required to lift the “post-social programs poor” out of poverty. Given the budget constraints, we formulate a proposal for the improvement of the social aid program, able to cure only the extreme poverty.

The paper is structured in five sections.  Section one compares poverty and inequality in Romania with other countries in the region, and the dynamics of poverty and inequality in Romania. The second section presents a profile of the poor, including the extreme poor, a “static” household- level -model of the determinants of consumption’s (welfare) determinants for the whole population, as well as tailored models for the largest groups of poor: employees and pensioners. Section three investigates the dynamics of poverty, estimating who isare those who are improvinggetting better or deterioratingworse overin time. We distinguish among transient and permanent poor, two categories whose escape from poverty would require different solutions. Transient poverty would shrink when growth resumes, while curing permanent poverty would require social programs. As manymuch of the other papers at this conference deal with the issue of growth, we narrow our focus in section four to the social sector interventions, especially in their efficacy in poverty alleviation.  Section five concludes the paper.

1.     Poverty and Inequality in Romania



Poverty aAnd Inequality iIn Romania

0.1                  A World View: How Poor Are The Romanians?

Before beginning an in-depth analysis of poverty and inequality in Romania, we will step back a little, and look at these two phenomena over the globe. In particular, we would ask how manymuch people live on less than one (international) and four US$ a day in Romania[iii], and in other regions in the world? And how unequal is the distribution ofn income, andor consumption in Romania and in other regions? Such comparison would reveals the heterogeneity of these phenomena across the globe.

For international comparisons, we used regional poverty indicators estimated by the World Bank against a “severe poverty line”, that counts as poor all persons living on less than one 1985 international dollar a day.  The poverty threshold used is quite low, signaling malnutrition, the absence of adequate purchasing power to satisfy the basic food requirements. Poverty estimates for various regions were produced by the World Bank (World Bank, 1996), for 1993[iv]. Using the same methodology, we computed a similar estimate for Romania, based on the 1997 Living Standard Measurement Survey data. For inequality comparisons, we compared a recent (1997) estimate of the Gini coefficient[v] for Romania with regional estimates produced by Deininger and Squire[vi] (1996).

Text Box: Table 1.1 Extreme Poverty and Inequality in Romania (1997) and in the World (1993)
	Headcount %	Income shortfall%	Gini coefficients
Romania	1.9	0.3	0.30
Eastern Europe	n.a.	n.a.	0.29
Eastern Europe and Central Asia	3.5	1.1	
East Asia and the Pacific	26.0	7.8	0.38
Latin America and the Caribbean	23.5	9.1	0.49
Middle East and North Africa	4.1	0.6	0.38
South Asia	43.1	12.6	0.32
Sub-Saharan Africa	39.1	15.3	0.47
High-income Countries	n.a.	n.a.	0.34
Total	29.4	9.2	0.38
Source: 	World Bank (1996) for poverty, Deininger and Squire (1996) for inequality and authors’ estimation for Romania (1997)
Note:	Poverty is computed as region average (population-weighted), against an absolute poverty line of 1USD a day (1985 international USD) per capita.
	Gini coefficients are region unweighted averages, based on per capita income or consumption distributions for 1993-95. For Romania, Gini is based on consumption.
Few Romanians weare living on less than one US$ a day in 1997, and those who dido, weare very close to this level (Table 1.1). This means that malnutrition, one of the cruel signs of poverty, is a very marginal phenomenon in Romania. In this respect Romania, as well as the other countries in Eastern Europe, Central Asia, the Middle East and North Africa, contrasts with the rest of the world: both the incidence and severity of poverty incidence and severity are low. The plague of malnutrition is still widespread in other regions of the world, such as South Asia, Sub-Saharan Africa, East Asia and the Pacific, Latin America and the Caribbean. Thus, the significant drop in economic activity and welfare registered in Romania during the “transition decade” did not pulled the standard of living of those “relatively” poor into the range of malnutrition. In part, the near-absence of severe poverty in Romania may be associated with a low level of inequality. Although rising by almost 50 percent from the “near-equality” levels registered underduring central -planning, Romania’s inequality is at the average level of its region (ECA), one of the lowest in the world.

From an economic policy perspective, a more interesting comparison would be among Romania and other transition economies, given the common legacy and reform processes experimented with in the last decade. To reflect the higher level of per capita income in transition economies, poverty comparisons should use higher thresholds. Milanovic (1998) suggested a regional poverty line that counts as poor all persons living on less than four 1990 international dollars a day, for which he estimated poverty statistics for Central and Eastern European countries, as well as for the former Soviet Republics, for various years in the period 1993 to 1995. Using the same methodology, we computed a similar estimate for Romania, based on the 1997 Living Standard Measurement Survey data. For inequality comparisons, we compared a recent (1997) estimate of the Gini coefficient[vii] for Romania with country estimates produced by Milanovic[viii] (1998).

Text Box: Table 1.2. Poverty in Central and Eastern Europe at 4 PPP$, 1995*
	Poverty	Inequality (Gini)
	Headcount	Number of Poor	Income (cons) shortfall (IS)	IS as % of GDP	Consumption per capita	Income per capita
	%	millions	%	%	coefficient	coefficient
Romania	65	14.7	24	5.10	0.30	
Bulgaria	15	1.3	26	1.10		0.34
Poland	20	7.6	27	1.40	0.31	0.28
Czech Republic	<1	0.1	23	0.01		0.27
Hungary	4	0.4	25	0.20	0.27	0.23
Slovakia	<1	0.0	20	0.01		0.19
Slovenia	<1	0.0	31	0.02		0.25
Source: 	Authors’ estimation for Romania (1997) and Milanovic (1998) for the rest of the countries.
Note:	The poverty estimations use the 1990 international US$. Gini coefficients are region unweighted averages, based on per capita income or consumption distributions for 1993-95. For Romania, Gini is based on consumption
Among the countries from Central and Eastern Europe, Romania has the highest poverty rate but similar poverty deficit[ix]. Two thirds of the population in Romania are living on less than four PPP$ a day in 1997, being in average 25 percent below this level (Table 1.2). These two numbers imply high costs of curing poverty, costs estimated – under perfect targeting and no administrative costs –at more than 5 percent of GDP[x]. As for income inequality, Romania resembles Poland, Bulgaria, the Czech Republic or Hungary, with a higher level of inequality than Slovakia or Slovenia.


The unwanted leading seat Romania’s unwanted distinction for having the highest poverty level in the region holds in poverty level in the region results fromis due to three factors:  a) the pre-transition differences in income per capita and income distribution;  b) the growth performances; and  c) the increase in inequality during the decade of transition in each of the countries. In 1989, Romania had the lowest income per capita among this group of countries, fairlysomehow close to the Bulgaria’s (Table A1.1 and A1.2 in Annex 3). Although, during the transition decade, all these countries witnessed a dramatic decline in output during the transition decade, mirrored by a sharp reduction in the living standards up to 1993, only Romania and Bulgaria experienced again recessions again in 1997 and 1996, respectively. Finally, while all the countries abandonedleft central -planning with low inequality – Gini’s around 0.2 –, Romania was among the countries with the highest increase in inequality.

There is an trivial parallel betweenamong growth performance and success in poverty alleviation. Among transition economies, fast or committed reformers succeeded into restoringe their economies toon a path of sustainable growth path, while hesitant reformers did notdon’t. At the micro level, the growing transition economies cured much of the “transient poverty” and, thanks to sizablesizeable social budgets and lower demand for social assistance, alleviated much of the “permanent” poverty as well. The other transition economies, such as Romania, Bulgaria, and countries from the Former Soviet Union, are still far from this target. We illustrated, in this section, the marginal impact of “absolute” poverty in countries like the Czech Republic, Slovakia, Poland andor Hungary, contrasted with itsthe high magnitude and severity in Romania, Bulgaria and the republics of the fFormer Soviet Union Republics, Romania and Bulgaria. This speaks toabout the importance of finalizing the reform agenda in Romania, and the desirability of an increase in the pace of reform marching speed.


1.2.        Poverty aAnd Inequality: Levels aAnd Trends During Transition

From a practical point of view, anti-poverty policies are “national” policies, and both poverty measurement and the instruments for its alleviation are designed relative to the country’s standards. This is the case in Romania, and in most of the countries in the world. Although recently the authorities and the domestic research community began to use systematically a Romanian methodology to quantify poverty, we are captive to the methodologies used to produce earlier estimates to asses its evolution over a longer time span.

Text Box: Table 1.3 Poverty, Inequality and Growth during Transition
Year:	Headcount rate	Inequality (Gini)	GDP Growth	PC GDP Dynamics
	%	Consumption/Capita	Annual %	1989=100
1989	3.7	0.21		100.0
1990	NA	NA	(5.6)	94.4
1991	NA	NA	(12.9)	82.2
1992	NA	NA	(8.8)	75.0
1993	20.0	0.23	1.5	76.1
1994	19.2	0.30	3.9	79.1
1995	16.3	0.31	7.1	84.7
1996	11.6	0.30	3.9	88.0
1997	19.3	0.28	(6.6)	82.2
1998	22.0	0.30	(7.3)	76.2
Source: 	Own estimations for 1995-1998, and World Bank (1997) for 1989, 1993 and 1994
Note:	Poverty is estimated using a line of 46,045 ROL per capita, in Jan-1995 prices, roughly equal to 50% of the average household consumption per capita in that year. Gini coefficient is estimated based on the distribution of consumption per capita.
In Table 1.3 we present the evolution of poverty against a benchmark pioneered by the World Bank (1997) for the period from 1989 to 1994. Based on this line – equal to a consumption of 46045 ROL per capita in January 19-95 prices – we estimated the incidence of poverty for the recent years 1995 to 1998. In the first decade of the transition period, the incidence of poverty increased six times, from 3.7 percent in 1989 to 22 percent in 1998. In the same table, we present two other indicators that are likely to cause the estimated increase in poverty: inequality and per capita GDPGPD growth. During the first six years, the data suggest that the worsening income distribution did not affected poverty significantly. As expected, the main determinant of the poverty incidence was the evolution of GDP per capita, the correlation coefficient betweenamong the two series being –0.979.

Text Box: Table 1.4.  Poverty in Romania  1995-1998	
	1995	1996	1997	1998
Poverty headcount P0 (%)	25.27	19.85	30.81	33.82
Total number of the poor (‘000)	5,725	4,488	6,945	7,609
Consumption shortfall (ROL, Jan-95 prices)	18,069	16,264	18,234	19,166
Consumption  shortfall as % of poverty line	25.44	22.71	25.73	27.01
Total poverty deficit as % of GDP	1.55	1.01	1.95	2.42
Elasticity of P0 to GDP	..	-1.35	-1.57	-1.29
Poverty Gap Index FGT1 (%)	6.43	4.51	7.93	9.13
Poverty Severity Index FGT2 (%)	2.42	1.54	3.00	3.56
Extreme poverty headcount (%)	7.96	5.07	9.53	11.70
Source: Own computation based on LSMS 1995-98 
Although Romania does not have an official poverty line, recent studies used systematically similar methodologies for poverty measurement systematically (Wagner et al. 1998, Dinculescu and Chirca 1999, ChircaChirc_ and Tes_liuc 1999), building a national “expert consensus”. Recent studies commissioned by the “Presidential Commission Against Poverty” endorsed this methodology, tantamount to official acceptance of this “expert consensus”making this “expert consensus” close to an official acceptance or even support. The core poverty analysis undertaken in this study appliesis done against the same poverty benchmark and methodology as the one used in the above-mentioned studies. The poverty line used in this study is defined as 60 percent of the monthly average consumption per adult equivalent in 1995, equivalent to US$D 40. For 1996 and 1997, all the monetary variables were deflated to January 1995 prices, to account for seasonal variation in ROL purchasing power. All persons living in households with a lower level of consumption per adult equivalent than the poverty line were counted as poor. In Table 1.4 we present poverty statistics for 1995-1998 estimated according to this methodology for 1995-98. The reader interested in the methodological aspect of poverty measurement may consult Annex 2.

In Romania, both the incidence and the severity of poverty are high, and have increased over time (see Table 1.4). Poverty was a marginal phenomenon at the outset of the transition period, but became a major problem thereafter. Romania experimented with gradual reforms for almost a decade, a combination of stop-and-gogo-and-stop policies. These proved to be very costly, so that by 1998, GDPGPD was still at 76 percent its pre-transition level, with further declines expected in 1999. A decline in living standards mirrored tThe decline in economic activity was mirrored by a decline in living standards, notably in the level of current consumption per capita. Poverty was somehow aggravated by the increase in inequality, due mainly to the new occupational risks, like unemployment, and the new opportunities, such as the freedom of entrepreneurship. While in 1989 the number of poor in Romania was estimated to be at around 1 million persons, in 1998 almost eight million persons or 34 percent of total population were poor, and more than two millions of those can be classified as extremely poor.

Poverty in Romania is shallow, meaning that most of the poor are clustered not far below the poverty line. A low Gini index among the poor (0.1), and a relatively low consumption gap (the average consumption of the poor households is only 25 percent below the poverty line) illustrates this. PDue to this, poverty is therefore highly elastic to GDP movements. The 3.9 percent growth in GDP during 1996 broughttook 1.2 million peoplepersons out of poverty, reducing the poverty headcount from 25.27 percent in 1995 to 19.85 percent in 1996. The 6.6 percent decline in GDPGPD in 1997 pushed another 2.5 million peoplepersons in poverty, increasing the headcount to 30.81 percent in 1997. In 1998, the second year of economic decline when GDPGPD fell 7.3 percent, the number of poor increased bywith 650 thousand people. Hence, most of these are “transient poor”, people who have recently become poor due to declining macroeconomic performance. Restoring growth will take them out of poverty. Severity of poverty also increased with declines in GDP decline, doubling in 1998 to 3.56 percent from 1.54 percent in 1996. That means that the poorest fare worst in 1998, compared with earlier periods.

The overall consumption shortfall is relatively small as a percentage of GDP is relatively small, making poverty alleviation policies financially feasible, at least under a scenario  of perfect targeting and low administrative cost scenario. The poverty deficit as a proportion of GDP was 1.5 percent of GDP in 1995, 1 percent in 1996, 2 percent in 1997 and 2.4 percent in 1998. A small poverty deficit carries a favorable message for anti- poverty policy.  Poverty eradication or important alleviation is possible with a relatively low cost, assuming perfect targeting. If we take into consideration the phenomena of  “leakage” (payments to the non-poor) and “spillover” (payments to the poor in excess of what they need to reach the poverty line) and the administrative costs of identifying the poor, the cost of poverty alleviation might go up. Many writers (Milanovic, 1998; Subbarao, 1997) emphasized the many difficultiesy of perfect targeting,, for many reasons, from lack of administrative capacity, to the perverse incentives generated by the system[xi], and recommend anthe increase of the estimates by a factor ranging between 2 and 3 times. Takingen an average of these recommendations, we arrive at a cost between 2.5 and 5 percent of GDP, which is substantial.

The next question, answered in the following chapters, is: MustIs the cost of covering the poverty deficit supplementadditional to the cost of existingcurrent social programs, or can some less effective social programs be eliminated or reduced to free funds?some of them can be substituted to current programs that fare worse in term of poverty alleviation, and that can be reduced or eliminated. Our analysis suggests that most of this expenditure would require additional funds. Based on this finding, the final chapter endorses an intermediate solution, presented in Dhanji et al. (1999), in which only the extreme poverty is tackled, atwith an additional cost of 0.33 percent of GDP.

Text Box: Table 1.5. Evolution of Inequality by Are of Residence Gini Index based on Consumption
Year	Total	Urban	Rural
1989	0.210	0.192	0.226
1995	0.308	0.295	0.317
1996	0.298	0.290	0.301
1997	0.284	0.275	0.292
1998	0.301	0.289	0.312
Source: Computations based on LSMS 1995-98.
Inequality, as measured by Gini coefficient, is not high in Romania (Table 1.5).  During the eight years of transition, it increased from 0.21 in 1989 to 0.30 in 1998. However, with a Gini coefficient of 0.30, Romania compares now with the Western European economies with a heavy welfare state, well-known for their low level of inequality. Curiously, inequality decreased in 1997, the first year of  a rapid enterprise reform program, implemented in parallel with labor redeployment measures. This may be due to the hesitant path of enterprise reforms in 1997, and the possible over-supply of the social safety net package. By area of residence, the inequality was greater in rural areas throughout the transition period.


2.   Who Are tThe Poor?

In this chapter, we will investigate the socio-economic characteristics of the households associated with poverty status. The answer would isbe provided through simple bivariate statistics in the first section, where we present differences in headcount rate and consumption shortfall as a function of socio-demographic, and economic characteristics of the households. In section two we investigate more formally, using multivariate analysis, the correlation between household welfare – proxied by household consumption per adult equivalent – and the level of household resources. The last section of the chapter uses the same multivariate technique to identify the factors associated with poverty among the largest occupational groups: employees and pensioners.


2.1     A Profile oOf The Poor

This section builds a poverty profile, by socio-demographic, economic and regional characteristics.  Among the socio-demographic factors associatedion with poverty are a high dependency ratio, having a female heading the household, being young or belonging to the gypsy community. Households would face lower poverty risks if their endowment with resources is greater, from human capital, productive assets like land, or consumption assets like durables or houses. Finally, we signal the regional dimensions of poverty, as we findfound a higher incidence in the rural area, or in the North-East, South and South-West of the country. The reader isn warned to interpret these results with caution, some of them being only “fallacies of composition”, as the multivariate analysis in the next section would reveal.

The next subsection presents the main factors that are associated with poverty, the latter being measured by the headcount rate and the consumption shortfall. The interested reader may fiound additional details in Annex 3, Table A3.4.


Socio-Demographic Characteristics

Dependency ratio (family size and number of children) Households composition is one of the most significant correlates of poverty, since the numbers of income earners and dependentsdependants determine the consumption needs and the ability to satisfy those needs. Throughout the world it has been observed that the poverty headcount increases with household size or number of children. This correlation holds for both rural and urban areas, and its capacity to discriminate between low-incidence and high-incidence poverty groups is the highest in any year from 1995 to 1998 (Figure 2.1). Households with five members face more than 50 percent chance of being poor , and for those with six or more members the odds rise to two thirds. These two categories account for 50 percent of the total poor in 1998. In contrast, single- or two-person families face a very low risk of poverty.


Extreme poverty, a concept which try to identify those people who face an extremely severe poverty, is also very high among large families (Table A3.5, Annex 3).  In 1998, 19 percent of the households with five members and 35 percent of the households with more than five members lived in extreme poverty. That means that a year ago, 1.6 million peoplepersons in Romania were living with  less than 40 percent  of the average consumption per equivalent adult.

In most cases, large households include a large number of children, and thus, households with more than 3 children have a high probability of being poor. Families with no children face half the risk of poverty of families with one or two children.  With each additional child after, the first after two children, the poverty incidence increases by 50 percent, rising to 84 percent in households with four or more children (Figure 2.1).

Text Box: Fig. 2.1 Poverty Headcount and Consumption Shortfall by:
Household Size							    			Number of Children
Source: Own estimation based on LSMS archives, 1995-1998

In 1998, although the magnitude of poverty was larger, the relative incidence was similar thanas in 1995. This means that the position of the poorest relative to the average remained unchanged.  In 1998, the households with three or more children, accounting for 5 percent of the total population, have the highest incidence of poverty: 65 percent of the families with three children and 84 percent of the families with four children or more are in poverty. Together, these groups account for 17 percent of the total poor.

More than 30 percent of poor households  with one or two children and more than 50 percent of those with three orand more children live in extreme poverty. With each additional child  the chance of living in severe poverty almost doubles rising to 45 percent in the poor households with four or more children.  The strong correlation between poverty and number of children in the household makes family benefits (such as the child allowance schemes) important instruments of safety net policies.

Age.  The high incidence of poverty among households with children determines that one third of all Romania’s poor are children less than 15 years of age.  Very popular among politicians is the idea that  pensioners are the most vulnerable group.  The data in the LSMS provides quite contradictoryan opposite evidence.  The incidence of poverty is 18 percent in the  group aged 56-65 years and only 10 percent in the age group over 65 years.  These two groups together account for only 10 percent of thein total poor in Romania.

Gender. Our bivariate estimates show that, over the period 1995-1998, female-headed households face lower risks of poverty as compared with o their males counterparts but larger consumption shortfalls. However, this is simply a “fallacy of composition”.  As revealed by the multivariate analysis in the next section reveals, male-headed households would have a 5 percent higher consumption, ceteris paribus.  Most of the female-headed households belong to the one-person category, being either unmarried females or widows. So, on average, female-headed households have lower dependency ration than male-headed ones. Because the dependency ratio effect is much stronger, bivariate estimates would show – erroneously – lower risk of poverty among females.

Even if we consider that the poverty headcount may accurately measure the risk of consumption poverty that female-headed households are facing, there are other characteristics of this group that point to its vulnerability. Most of the female--headed ed householdss are elderly widows living alone, who are faceing not only the risk of inadequate consumption, but also health hazards.  Their needs for care and old-age treatment are not accurately measured by our indicator[xii]. Given the differential  life expectancy of women and men, poor women comprise a large share of elderly women.  We consider the subgroup of elderly, single-female households, as a group at significant poverty risk.  This group represents about 80 percent of the total female-headed households.

Ethnicity.  By ethnicity, the only group whose poverty incidence departs from the average are the gypsies. The incidence of poverty among gypsies is 3.5 times higher than the average poverty rate and their consumption  was 40 percent smaller than the average consumption per equivalent adult in 1997. Part of this poverty differential betweenamong the gypsy community and the rest has, however, other causes than those associated with ethnicity, however. Most often gypsy households are of large size and include a large number of children, have  low levels of education and a high propensity towards  informal activity, factors that are associated with high incidence of poverty.  The other ethnic groups (Hungarian, Germans, others) have the same incidence of poverty  as the Romanians.


Economic Characteristics of the Households

Text Box: Fig. 2.2 Poverty Headcount and Consumption Shortfall, by:
Education of HH head		         					Occupational status of HH head
Source: Own estimation based on LSMS archives, 1995-1998
Human capital. The bivariate analysis suggests that the education level of the household head is a significant indicator which discriminatinge between low and high poverty (Figure 2.2.). Similarly, the consumption shortfall decreases monotonically with each additional level of education.

However, the capacity of “education level” to discriminate between poor and not poor is much weaker than the multivariate analysis suggests. While the incidence of poverty is highest among households with no formal schooling is the highest (42 percent), it isbut not much higher than that of those headed by individuals with gymnasium or a vocational education. One reason for this similar poverty incidence is the fact that in Romania very few people are illiterate, and those who are belongpertain to the generation aged over 65. This category of elderly, howeverone the other hand, have had the time to accumulate relatively more  assets (savings), plus were the main beneficiaries of the process of land restitution (Chirca and Tesliuc, 1999).  The high incidence among households whose head has vocational education is a consequence of the fact that typically demand for workers with vocational education declines during the restructuring.

In terms of consumption welfare, the returns on education start to be significantly higher only for those households headed by individuals with post- secondary or higher education.  The bulk of poor, however, live in those households whose head has only primary or no formal education (24 percent), gymnasium (25 percent) or vocational education (29 percent).

Occupation. By occupation, the highest risk of poverty is to be found among households headed by the unemployed-, farmers- or theand self-employed-headed households (Fig. 2.2).  Employee and pensioner households are bellow the average risk of poverty; however, these two groups amount for two thirds of total number of poor in all the four years.  Similarly, the consumption shortfall is  significantly larger than average for the unemployed, farmer and self employed households, and smaller for the employees and pensioners. Some new occupations that emerged during transition (e.g. non- agricultural self-employment) are associated with high poverty risks, a situation that is atypical in the region (Grootaert, 1998). Unemployed-headed households face the highest incidence of poverty, although their relative position improved each year since 1996.


Text Box: Fig. 2.3 Poverty headcount and Consumption Shortfall by Land Owned
Poverty headcount and poverty gap by region
Source: Own estimation based on LSMS archives, 1995-1998

Ownership of assets
Households with more than two hectares of land have lower poverty rates than the average (Fig.2.3). At this stage it is difficult to determineappreciate if land ownership influences welfare, while accounting forwithout keeping under control other factors.  Most often the ownership of land is of small size and is associated with pensioner type households and of small size.  The ownership of land does not make much difference in the consumption shortfall.  This is because those who do not own land may own other important assets such as education or work experience.

2.1.3       Regional Aaspects of pPoverty

There are two main regional dimensions of poverty in Romania. First, poverty incidence differs markedly between rural and urban areas. Throughout the period, rural poverty was about 50 percent higher that the urban one. For instance, in 1998 about 41 percent of the rural population lived in poverty, while in the urban areas only 28 percent of the inhabitants were poor.

Second, there are differences in the poverty levels betweenamong regions, as emphasized synthetically in Figure 2.3. The figure is based onto a very dis-aggregated definition of regions (15), presented in the Green Paper on Regional Development (Government of Romania, 1995) higher than the “official” one (8).  From an institutional point of view, Romania was “divided” into 8 development regions which, for most development indicators, are not homogenous. Such classification would be of little use for our purpose, because it will hide the higher intra-regional differences. The LSMS allows us to compute poverty at the level of 15 statistically representative sub-regions, identified in figure 2.3 by the initials of the component judets. The poor regions are North East (Botosani, Vaslui, Iasi), South (Teleorman, Giurgiu, Ialomita, Calarasi) or South West (Valcea, Gorj).


2.2.              Multivariate Analysis oOf tThe Correlates  oOf Household Consumption

The basic model estimated in this section is a variant of those presented in Braithwaite and Grootaert (1998) and Chirca and Tesliuc (1999). ThisSuch welfare model is a reduced form equation of the various structural equations which express the income-earning and consumption behavior of the households (see e.g. Glewwe, 1991). The model starts from the observation that household consumption – a widely used proxy for household welfare – is determined, first, by the level and quality of the resources a household owns, and second, by the returns that household may derive from these resources. To compare the consumption of households with different sizes and demographic structures, consumption was expressed in an “adult equivalent” measure. In addition, some control variables such as family size and the gender of the household head were introduced as right-hand side variables.

The simplest form of model to be estimated is:

where the dependent variable C is household consumption expenditure (per adult equivalent), HC and PR are vectors of household resources (measuring human capital and, respectively,  physical and financial capital, respectively), R is a vector of location variables that may influence the rate of return on resources HC and PR, and a, b and d are coefficients to be estimated. e is the disturbance term.

The dependent variable in the model, C, is measured in natural logarithms. Current consumption includes consumption of food, non-food and services. This variable was regressed on three types of variables (HC, PR and R) which are explained as follows:

1.    Household human capital variables. The first block of variables measures the human capital available at the household level. Human capital is usually measured by the level of education, experience proxied by age, the sector of activity and occupation. For education, we used as predictor the average years of education of adult members. Experience was proxied by the age of the household head. For occupation and sector of activity, we used the number of wage-earners, farmers, pensioners, unemployed and employers in the household. One would expect higher consumption be associated with higher level of education and experience. As for the occupational status of the active members of the household, one would expect consumption to be positively correlated with earning capacity. The magnitude of occupational coefficients will be used to rank the impact of various occupation on the expected level of consumption per adult equivalent.

2.    Household physical resources variables. The second block of variables measures the physical and financial resources the household owns, and includes livestock, area of land owned and the stock of savings. For livestock, the predictor used is the level of the stock measured in large cattle units[xiii]. The amount of agricultural land owned was expressed in hectares, and the stock of savings at the end of the period in ROL. One would expect a positive correlation between the availability of land, livestock or savings and consumption.

3.    Regional variables. Household consumption may vary regionally, due to unobserved location-specific variables.  To account for these, we introduced a vector of dummy variables for each region of Romania (eight regions), plus one for the area of residence (urban-rural)

Other variables introduced in the model, contained in a residual block, includeare the size and demographic structure of the household, the latter being measured by the age and gender of the household head. Household size, expressed in equivalent adults, was introduced to control the economies of scale associated with larger households. The demographic structure of the household controls for the fact that consumption varies with age and gender.

Information on household welfare and privately-owned, household resources have been collected through the Integrated Household Survey (LSMS), the 1998 Archive. The estimated model is presented in the Tables A3.6 and A3.7, Annex 3. The model, estimated through least-squares and corrected for the heteroskedasticity of the error term[xiv], is highly significant (probability of F-statistic is lower than 1 percent) and has high explanatory power.  The independent variables are able to capture 43 percent of the variation in the consumption per adult equivalent. Such goodness of fit is considered good for a cross section.

The log-linear functional specification was chosen over the linear form on the basis of the Davidson and MacKinnon test (1981). This impliesy that effects of household characteristics on welfare are proportional rather than linear, and the elasticity of consumption on various predictors would be a linear function of the value of that predictor. More simply, the elasticity would be, in absolute terms, lower for the sample of poor than for the sample of not-poor. To explore this feature of the model, we estimated the elasticity on each predictor at mean value for poor and not-poor. For not-significant predictors (at 5 percent confidence interval) we dido not estimated elasticity. The results are presented in Table 2.1. For dummy variables, were computed marginal changes (effects), indicating the percentage change in (natural logarithm of) consumption relative to its mean when the indicator variable changes from zero (e.g. rural, for the area of residence variable called Urban) to one (urban).

These elasticity coefficients are “marginal returns on the available resources”, showing how much consumption per adult equivalent would change when the independent variable would changes bywith one percent.  All elasticity isare below one, meaning that a one percent change in the (mean) level of any predictor would change consumption by less than one percent.  The variable with the biggest impact on consumption is family size, confirming our findings from the previous section: one percent change in the number of adult equivalents would reduce consumption by 0.545 percent .

Second in importance is the “average years of adult schooling”, which shows that a one percent change in schooling would increase consumption by 0.32 percent. A different specification, using interaction dummies for area of residence, yielded 50 percent greater returns from schooling in urban versus rural areas.

is much lower for the rest of the variables. The next most important variable is “number of employees” (elasticity 0.11) followed by ownership of physical assets like land or livestock (0.04-0.03) and increase in the number of other income earners. In line with earlier findings, the presence of a single farmer in the family would not change significantly its consumption, while an increase in those who are unemployed would reduce it.

Text Box: Table 2.1 Elasticity and Marginal Effects of Household Resources on Consumption, Poor versus Not Poor, 1998
Variable 	Poor	Not Poor	Elasticity Ratio
	Mean	Elasticity	Mean	Elasticity	Poor / Not Poor
Continuous Variables:					
 Adults Average Education Stock 	7.841	0.321	8.551	0.350	92
 Age of the Household Head	47.471		47.471		
 Employers (Number)	0.003	0.001	0.009	0.003	38
 Employees (Number)	0.685	0.111	0.709	0.115	97
 Farmers (Number)	0.632		0.218		
 Unemployed (Number) 	0.361	(0.020)	0.098	(0.006)	368
 Social Security Pensioners (Number)	0.226	0.027	0.472	0.056	48
 Farmer Pensioners (Number)	0.138	0.013	0.237	0.022	58
 Livestock, in Large Cattle Units	1.199	0.031	1.275	0.033	94
 Land Owned, in Ha	0.714	0.030	1.007	0.042	71
 Savings Stock, in ROL 	187,417	0.004	590,185	0.013	32
Household Size, in Adult Equivalent 	2.738	(0.545)	2.738	(0.545)	100
Dummy Variables:					
 Male	0.825	0.051	0.731	0.051	100
 Urban 	0.405	0.047	0.506	0.047	100
 Region: Center 	0.121	(0.123)	0.125	(0.123)	100
 Region: North-East 	0.202	(0.112)	0.152	(0.112)	100
 Region: North-Vest 	0.125	(0.070)	0.127	(0.070)	100
 Region: South 	0.167	(0.083)	0.152	(0.083)	100
 Region: South-East 	0.131	(0.078)	0.121	(0.078)	100
 Region: South-Vest 	0.106	(0.048)	0.119	(0.048)	100
 Region: Vest	0.086	(0.115)	0.097	(0.115)	100
Source: Own Estimations based on LSMS Archive 1998


Second inas importance iscomes the “average years of adult schooling”, which shows that a one percent change in schooling would increase consumption by 0.32 percent. A different specification, using interaction dummies for area of residence, yielded 50fifty percent greater returns from schooling in urban versus rural areas.

Elasticity is much lower fFor the rest of the variables, the elasticity are much lower. The next most important variable is “number of employees” (elasticity 0.11) followed by ownership of physical assets like land or livestock (0.04-0.03) and increase in the number of other income earners. In line with earlier findings, the presence of a single farmer-only in the family would not change significantly its consumption, while anthe increase in those who are unemployed would reduce it.

An increase in the human or physical capital of a household would increase more than proportionally the welfare of a not-poor household, compared to the poor ones.  The highest differential is on the stock of savings (32 percent, due to the differences in average endowments), followed by changes in household occupational structure (change in the number of employees, 38 percent;, social security pensioners, 48 percent;, and farm pensioners, 58 percent). Significant differences in returns between poor and not-poor exists on the income derived from land (71 percent), although not for livestock (94 percent).  As expected, anthe increase in the number of unemployed persons in a family would affect more than three times a poor household more , than a not-poor one by more than three times (368 percent).

Gender of the household head and location impact on household consumption. Controlling for all other variables, male-headed households enjoy greater average consumption thant similar, but female-headed ones.  Such gender differential is estimated at 5 percent.  By area of residence, urban households have antheir average consumption per equivalent adult 4.7 percent higher that those located in rural areas.  By region, the highest consumption – ceteris paribus – is in the capital city, Bucharest, and the lowest in the North-East, North-WVest and WVest.



0.3                  PPoor Employees aAnd Pensioners: The Largest Poverty Groups

The two largest groups of poor are those living in households headed by employees (39 percent) and pensioners (26 percent), despite the relatively low poverty incidence of this type of households. We investigate, through regression analysis, the characteristics of households headed by employees or pensioners that tend to be associated with poverty based on the LSMS 1998 sample. The estimations are presented in Tables A3.8 throughto A3.11 in Annex A3. The characteristics that tend to have a significant influence on the poverty status of these groups are presented bellow.

Employee households. The most important factors associated with the poverty risk are those regarding the household composition: first of all, the morehigher vulnerable households tend to be those with a larger size (i.e. number of members adjusted into adult equivalent); second, an increase of the dependency ratio (defined as number of children divided by number of permanent income earners[xv]) as well as the presence of unemployed members in the household raise significantly the odds of being poor; finally, the gender of the household head is another factor that has a relatively strong influence with respect to poverty: female headed households are more likely to be poor than the male headed ones.

A second category of factors influencing the poverty status concernsis the one regarding the educational level and job-related characteristics of the household head. EThus, one can observe that each additional level of education achieved by the household head has a very strong diminution effect on the household’s probability of being poor. With respect to job-related characteristics, households headed by employees with higher professional status (manager, professional, technician etc.) are less vulnerable to poverty risk than households headed by workers, especially unskilled ones. An interesting finding is that although the average level of welfare (measured by consumption per equivalent adult) of private sector employees does not differ significantly by comparison with thate one of state employees, the former category is more likely to be poor (ceteris paribus). A possible explanation of this phenomenon could be the higher inequality of welfare among private sector employees as compared with the state ones, as an effect of wages inequality. As far as the sector of activity of the household head is concerned, one can observe that the households headed by employees in agriculture or education are more likely to be poor, while those headed by employees in extractive industry or energy, gas and water (characterized mostly by a monopoly position) are the less vulnerable to poverty risk. Also, it is important to say that an unstable position of the household head on the labor market (i.e., having a limited period contract) increases significantly the odds of being poor.

Finally, there are two more relevant factors relevant tofor the poverty status of the employee- headed households: the first is the ownership of land, which has a high positive impact on reducing the probability of being poor;, andwhile the second concerns the impact of regional labor markets, and showings that households located in judets with higher rates of unemployment are more exposed to poverty.

Pensioners households. Pensioner households have the lowest incidence of poverty, after the employeers ones.  The most important factors that influence their poverty status are, as in the case of employee households, those related to the household composition: larger households are more exposed to the risk of poverty, as well as households with unemployed members.  On the other side, having employees in the household or being male headed by a male lower the household vulnerability to poverty. The education of the household head has also has a strong impact on the poverty status: those without schooling or with lower levels of education are the most likely to be poor. The type of pension received by the household head is another important determinant of poverty for pensioner households - the odds of being poor are greater for those collecting in the case of agricultural, disability and survivor pensioners than for those collecting being higher as compared with old age state insurance pensioners.

From a geographical point of view, one can observe that the probability of being poor is unevenly distributed among pensioner households, the regions North-East and West being the ones with the highest risk of poverty, as compared with Bucharest. Finally, it is worth being mentioninged that, despite the large impact of land ownership on reducing the vulnerability to poverty, the rural located pensioner households are more likely to be poor (ceteris paribus) than the urban ones.

3.     Two Kinds of Poor:  Transient and Permanent Poverty

2.Two Kinds oOf Poor: Transient aAnd Permanent Poverty

Conceptually, there are two situations that makes a person’s welfare falling below the poverty threshold: an income shock, or a low level of assets with growth-inelastic returns. 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 of the assets (including labor) they own. In this category one would include the unemployed who, in periods of economic recession, loose their jobs. When the economy recovers and employment increases, such individuals would re-enter the labor force and may escape poverty.

Other households would not be able to escape poverty even when economy recovers, because the assets they own do not generate sufficient income to lift them over the poverty threshold. The meager volume of assets such households own have similar returns under boom periods as under recession. Such households are called permanent poor. Typically, in this category are included the disabled, andor poor elderly unable to work..

The distinction among transient and permanent poor is not straightforward. Some authors (Rashid, 1997) consider as permanent poor those who, for some period of time, do not escape from the poverty pool. Other authors (Sen, 1976) 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 of poverty, and is relatively simple to measure.  The second classification is more close to phenomenon concepts, andbut requires subjective assumptions forto allow quantification. We estimated both of them, as they look at poverty from slightly different angles, thus bringing thus additional insights toon the poverty processes .

Text Box: Table 3.1 Entry Into And Exit From Poverty Pool, 1995-97
Poor in ..?	Individuals Belonging to an Household Headed by … in 1997:
1995	1996	1997	Employee	Self-employed	Farmer	Unemployed	Pensioner	Total
Not Poor Between 1995-97
No	No	No	64.6	28.8	37.2	29.1	72.2	63.6
Permanent Poor
Yes	Yes	Yes	5.1	27.3	20.0	15.5	3.1	6.5
Transient Poor
No	No	Yes	10.9	16.7	10.8	20.0	6.4	9.3
Yes	No	No	7.1	3.0	10.0	4.5	7.2	7.1
Yes	No	Yes	6.0	12.1	8.4	14.5	4.0	5.7
Atypical Poor
No	Yes	No	1.4	3.0	4.0	1.8	3.0	2.4
No	Yes	Yes	2.5	6.1	5.2	13.6	1.6	2.9
Yes	Yes	No	2.4	3.0	4.4	0.9	2.4	2.5
			100.0	100.0	100.0	100.0	100.0	100.0
Source: Own estimation based on the LSMS Archives 1995-97
Poverty Dynamics During Transition. To investigate the dynamics of the poverty in recent years, we used the subsample of the LSMS from 1995 to 1997.  Out of roughly 32,000 households surveyed each year, we selectedfound – thanks to the rotating panel feature of the data – about 3,000 households fromsurveyed each year. We used the panel to quantify the entry into and exit from the poverty pool, and to test if past poverty is associated with current poverty. OfFrom all the individuals in the panel, two thirds (63.6 percent) were not poor during the period (Table 3.1). The rest of the households were poor in at least one year.

Text Box: Table 3.2 The Structure of the Sample of “Poor in at Least One Year”, by Dynamics, 1995-97
	Individuals Belonging to an Household Headed by … in 1997:
	Employee	Self-employed	Farmer	Unemployed	Pensioner	Total
Permanent Poor	       14.4 	       38.3 	       31.8 	       21.9 	       11.2 	       17.9 
Transient Poor	       67.8 	       44.7 	       46.5 	       55.0 	       63.3 	       60.7 
Atypical Poor	       17.8 	       17.0 	       21.7 	       23.0 	       25.2 	       21.4 
Total	100.0	100.0	100.0	100.0	100.0	100.0
Source: Own estimation based on the LSMS Archives 1995-97

We would divided the households that were poor forin at least one year into three groups.  First, those that were poor throughout the period, fitting our first definition of permanent poor. Second, those who exited from poverty in good years (1996), but  entered in bad years (say, 1997). This group is close to our definition of transient poverty. The remaining group contains the exemptions, households that either fell into poverty when the economy went well, or exited from poverty in periods of recession. We have called them atypical poor. We presented, in Table 3.2, the structure of the subsample of the “poor forat least in one year”, for the whole subsample and disdiss-aggregated by the occupation of the household head in 1997.

Judginged by the dynamics of those suffering poverty, most of it – 60.7 percent – is transient poverty.  Surprisingly, permanent poverty is only 17.9 percent, significantly lower than the atypical poor (21.4 percent). Theise aggregate dynamics concealhide significant behavioral differences betweenamong various types of households, grouped by the occupation of the household head.  Extremely low levels of permanent poverty are noted for the employee and pensioner-headed households.  Such households seems to be able to restore their consumption above the poverty level in one or two years after the income shock. In contrast, a large proportion of the  self-employed or farmer-headed households “at least once poor” bear their poverty stigma year after year. This exercise is of immediate interest for social policy, because we identified the groups that were able to cope with the hardship of transition, and contrast these with the ones that fare worse overin time. 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 attention to the latter category.

As a feature of the transition process, one would notices the relative stability of the proportion of “atypical poor” for all types of households. In our opinion, this is indicative of the transformations that occur in the real economy – whichthat impactaffects primary market incomes such as wages and entrepreneurial income –, but also in the cash benefit system.

Various hypotheseis may be formulated to explain the high turnover into and from the poverty pool. First, the shallow poverty in Romania, plus the arithmetic of poverty measurement may be responsible for such results. About 8 percent of Romanians were 5 percent above or below the poverty line in 1997. Relatively small income shocks may have a small impact on their welfare position,change their welfare position a little, but still be enough this little is sufficient to change their poverty status. Second, during the three years under analysis thereit was an abundance of income shocks, a situation that is common in transition periods. As mentioned already, the enterprise reforms impacted on the wages and the welfare of some employee-headed households, as well as on those affected by the subsequent redundancies. The containment of the aggregate demand reduced the profitability of many businesses, hurting for instance the self-employed. Changes in the parameters of social programs, such as the lack of timely readjustment of pensions, child allowances or social aid, produced another type of shocks. Although such hypotheseis may be plausible, we caution the reader that they were not demonstrated, and their validation or rejection requires further research.

Text Box: Table 3.3 Household Capabilities to Escape Poverty: Transient versus Permanent Poor
	Poverty Headcount (%)	Poverty Structure (%)
Year	1995	1996	1997	1998	1995	1996	1997	1998
Transient Poverty					71.4	72.0	75.5	76.0
Permanent Poverty					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
 - head or member can not work	42.0	42.4	54.2	56.5	2.2	3.1	2.3	1.9
 - having four children or more	71.0	66.7	79.7	83.6	12.0	12.3	8.6	7.7
Source: Own estimations based on the LSMS Archives 1995-1998

The Poor That Needs Our Help. The second means for classifyingication theof poor into transient or permanent” poor uses as its criterion for determining the permanent poor the lack of means to overcome poverty. Here, we includeclassify as permanent poor the disabled, some of the elderly and families with a large numbers of children (four or above). According to this criteriona, about 28 percent of the poor were in permanent poverty in 1995 and 1996 (Table 3.3). As expected, in 1997 and 1998, years of economic decline, this proportion dropped to 24 percent as an effect of the large entry of “transient poor”. This definition gives us higher estimates than the “dynamic” interpretation, but in a relatively smaller range. In the end, it carries the same message, that much of the poverty in Romania is transient in nature.

The fact The positive message this analysis carries is that Romanians face mostly much of poverty faced by Romanians today is transient poverty in natureis a positive message. Transient poverty shwould shrink when growth resumes. This speaks toabout the centralprime role that macrostabilization and restoration of sustainable growth should play in a poverty alleviation strategy in Romania. For the remaining poor, social programs that shield them from poverty should be implemented. As much of the other papers at this conference deal with the issue of growth, we narrow our focus in the next chapter to social sector interventions, assessing their efficacy in poverty alleviation.


4.     Effectiveness oOf the Social Safety Net iIn Tackling Poverty

In this chapter we investigate the effectiveness of the Romanian social safety net in shielding the poor against poverty. In particularNotably, we are asking how much poverty theis tackled thanks to social programs alleviate, if the social programs provide adequate coverage to the poor and if the funds are reasonably well targeted to those in need. Also, we will try to assess the role of the state in reducing the inequality through redistribution.

We found that the overall social spending almost halves the poverty deficit in Romania, and that most social programs with a poverty- alleviation focus are well targeted to the poor. However, the main programs address only narrow welfare risks such us unemployment, or temporary loss of income due to illness, childcare or handicap. The only program with a broader focus – the social aid program – suffers from lack of funds and excessive discretion in implementation. Consequently, the coverage of the anti-poverty programs in Romania is inadequate. IndeedTo illustrate this, we mention at this point that a lowerthe percentage of poor households in the poor category that do not receive any support through social programs than do households in is higher than for the not-poor category. The fact that the programs are well- targeted means that additional funds are required to lift the “post-transfer poor” out of poverty. Given the budget constraints, we formulate a proposal for the improvement of the social aid program, able to cure only the extreme poverty.

This chapter draws heavily on a recent analysis of the Romanian social protection system done by Dhanji et al. (1999).


4.1.                        The Romanian Social Safety Net

Text Box:  Figure 4.1 Distribution of main cash benefits, 1997
Source: Statistical Yearbook 1998, National Commission for Statistics
Romania has a large array of social protection programs, both  (cash and in kind).  The cash transfers programs (which number more than 30) are the most important in terms ofas funds employed. A matrix-format description of all cash transfers is presented in Annex 4. Synthetically, they can be grouped into three broad categories[xvi]:

Social insurance: benefits awarded on the basies of a social insurance contribution record, or at the occurrence of a specified contingency, like unemployment, sickness or old age.  Social insurance includes pensions, work injury insurance, sickness benefits, unemployment insurance, severance payments, maternity and child care benefits. They formcut the largest share of totalin the total spending on cash transfers. In 1997, they accounted for 81 percent of the total.

Entitlements: universal benefits awarded on the basis of categorical characteristics, not related to the income of the recipients or contributions to benefits schemes. Entitlements include child allowance, incremental child allowance and birth grants. In 1997, they accounted for 14 percent of total spending on cash transfers (Fig. 4.1).

Social assistance: benefits awarded for people in “need”, based on a means test or the occurrence of an emergency situation. Social assistance includes social aid for low income families (an MIG-type program), allowance for wives of conscripts, family allowance for child placement, emergency help, allowances for prosthesis procurement, allowance for thermal energy, compensation for the increase in bread price, special aid for the disabled. In 1997, these programs accounted for less than 5 percent of total spending on cash transfers.

Many of these programs are inherited from the centrally  planned regime, and have begun to be reformed only recently (e.g.,: pensions). Among the new programs, unemployment benefits werewas introduced in 1991, and the minimum income guarantee in late 1995.  The child allowance benefit, whichthat during central planning during only covered the employees’ childrencentral planning covered only the employees’ children, was transformed into a universal benefit in 1993.

The structure of the Romanian social safety net has a heavy component of social insurance. In this respect, Romania is similar with most countries in the world. Somewhathow peculiar is, after 1997, the large reliance on entitlements to support households after 1997. The social assistance programs, targeted toward those in need, play a minor role in the total spending envelope of the cash transfer system.

By type of program, Tthe most important program by type one is the pension programs. In the four-year period covered in the paper, pensions represented between 65 and 80 percent of total cash benefits, or around 6 percent of GDP. In 1997, they accounted for 65 percent of total spending (Figure 4.1). Besides pensions, programs with sizable budgets are the child allowances and the unemployment benefit, with 0.6 to 0.7 percent of GDP. In 1997, their share in total cash spending was 14 and 12 percent, respectively. The social aid (MIG), the only social assistance program conceived as an antipoverty program with payments being made on the basis of need, provides the smallest cash transfer (less than 1 percent of total cash benefits in 1997). The overall cost of the cash transfer programs wasranged between 9.2 percent of GDP in 1997, upraising from a low of 7.7 percent in 1995.

Text Box:  Table 2.1 Benefit incidence on households and individuals
	State pensions*	Farmers pensions	Unemployment benefit	Child allowance	Social assistance**	Any type of transfer***
Percent of households receiving cash transfers
1995	39.82	14.37	8.46	36.40	2.16	79.68
1996	41.12	14.22	5.44	35.47	2.47	78.69
1997	41.43	14.07	5.38	37.46	16.81	82.07
Percent of individuals receiving cash transfers
1995	16.28	5.97	3.30	21.03	****	45.97
1996	17.00	5.91	2.07	20.43		44.56
1997	17.35	5.91	2.02	21.45		45.79
Source: Own computations based on LSMS, 1995-1997
 (*) old age, disability, survivor, social pension
(**) social aid, special benefits for disabled, other social assistance benefits
(***) does not include scholarships
(****) most of the social assistance benefits are awarded to households not individuals
The majority of Romanian households receive at least one cash transfer from the state (Table 4.1). Between 1995 and 1997, the percentage of households serviced by the cash transfer programs ranged from 79 to 82 percent. In terms of beneficiaries, the cash transfer reached between 45 andto 46 percent of all individuals (not counting those that benefited from transfers designated to households, not individuals).

Programs like pensions and child allowance were, as expected, remarkably stable. Significant changes occurred only in the coverage of the unemployment benefit and social assistance. Our estimates mirror tThe reduction in unemployment in 1996 is mirrored by our estimates. In 1995, 3.3 percent of individuals cashed in unemployment benefits. After 1996, only 2 percent were still cashing in unemployment benefits, in line with the fall of the unemployment rate and the exit of some long-term beneficiaries from the list of assisted persons. Social assistance covered a much larger share of the households in 1997, compared with the previous two years, an increase of 6 to 8 times in the number of assisted households.  This was due to the introduction of some temporary assistance programs, such as the “compensation for the raise of the bread price”, a benefit introduced forin several months in 1997 (March to October), that was distributed to a much larger number of households than the existing programs.


4.2.    The Social Safety Net’s Impact on Poverty

In this section, we investigate how efficient is the Romanian cash transfer system is in alleviating poverty. Among the transfer programs, some of them have a poverty alleviation objective and others do not. Our investigation will be limited to programs that have poverty reduction as a primary or secondary objective, excluding, for instance, pensions.

Text Box: Figure 4.2  Lifting the Poor out of Poverty
Source: Estimations based on the 1995-97 LSMS Archive
How many people would be poor without the current system of cash transfers? The success of any social transfer system is measured by the extent to which it contributes to poverty alleviation. Figure 4.2 illustrates this using two poverty measures, the headcount rate and the poverty gap index (FGT1).

The cash transfer system reduces poverty substantially, even when pension impact is not taken into account. In the absence of all cash transfers (pension included), poverty rate would increase by 70 to 96 percent, from 25.3 to 45.3 percent in 1995, 19.9 to 39.3 percent in 1996 and 30.8 to 52.3 percent in 1997. Much of this increase in the number of poor, between 70 and 80 percent, would occur in the absence of pensions, or if pension replacement rates would deteriorate substantially. However, pensions do not have poverty alleviation as their explicit function poverty alleviation, thus we will not analyze them from a poverty alleviation point of view. In the absence of all cash transfers but pensions, the poverty rate would increase by 17 to 18 percent, from 25.3 to 30.4 percent in 1995, 19.9 to 23.8 percent in 1996 and 30.8 to 36.9 percent in 1997.

The cash transfer system has an even a greater impact in reducing the overall poverty gap. The headcount rate tends to underestimate the poverty alleviation impact, as it does not take into account the fact that transfers may also alleviate the severity of poverty forof those still poor[xvii]. Another measure, poverty gap (FGT1), is sensitive both to the depth and incidence of poverty. The second graph on Figure 4.2 illustrates the dynamics of the poverty gap with and without transfers. In the absence of all cash transfers, the poverty gap would increase by 2.8 to 3.8 times, much more than the number of poor. As before, much of this increase in the poverty deficit would occur in the absence of pensions, or if pension replacement rates would deteriorate substantially. In the absence of all cash transfers but pensions, poverty gap would increase by 45 to 47 percent.

Text Box: Figure 4.3 Contribution to Poverty Reduction by Type of Program, 1997
Source: Estimations based on the 1995-97 LSMS Archive
What cash programs, excluding but pensions, had the greatest contribution to poverty alleviation? Compared with athe situation of no cash transfers except pensionspre-transfers but pension situation, the two programs with the largest impact on poverty alleviation are the child allowance and the unemployment benefit. These two programs tackled most of the potential poverty. Between 1995 and 19-96, unemployment benefits reduced the pre-transfer (but pension) headcount by 5-6 percent, and child allowances by 6-7 percent. In 1997, the upward adjustment in the child allowance resulted in an increased contribution of this instrument in poverty alleviation, at the expense of the unemployment benefit and other, non-contributory programs. As illustrated in Figure 4.3, the head count reduction determined by child allowance was even larger for 1997, atbeing about 10 percent. On the other side, social assistance, the only set of programs with a clear poverty alleviation mandate, contributed little to its goal. From 1995 to 1997, the rise in (current) poverty headcount associated with an absence of all social assistance spending would be less than half a percentage point.

Cash transfer programs in Romania do not seek high coverage of poor, but are rather “specialized” in some poverty risks. The concernissue with such a system is that many poor may fall between these programs, and notwould be not assisted. As an illustration one could note that while in 1997 the share of the (post-transfer) non-poor  households that do not receive any cash transfer is about 16 percent, the corresponding figure for the poor household is higher, being around 19 percent. Tables A3.11 and A3.12 in Annex 3 documents the impact of some cash transfer programs on the reduction of the number of poor and the poverty gap for 1997. For instance, unemployment benefits reduces substantially the poverty deficit among households headed by unemployed, while child allowances do the same for young households, with many children[xviii]. As mentioned above, social assistance has only a marginal impact, significant for households with illiterate adults, or living in overcrowded houses[xix].

When trying to evaluate the impact of a program on the poor, one must take into account three essential dimensions: a) coverage - the share of pre-transfer poor households that are covered by the program, b) targeting - the share of funds that are transferred to the poor, and c) effectiveness - the share of the benefit in the average consumption of the poor recipients. A simple Text Box: Figure 4.4 Coverage, Targeting and Effectiveness of Cash Transfers on the Poor, 1997
Source: Estimations based on the 1997 LSMS Archive
and intuitive way to present all this information is illustrated in Figure 4.4 for 1997 data. In the graph, program coverage of poor is measured on the X axis, program funds’ targeting of theon poor is measured on the Y axis, and program effectiveness – the share of the programs’ benefits in the consumption of poor beneficiaries – is proportionate with the size of the “bubbles”[xx]. Take as example the bubble that stands alone in the middle of the graph, on child allowances. This instrument covers 60 percent of the poor households (read X axis), transfers 51 percent of the funds to (pre-transfer) poor families (read Y axis), and representscontributes oin average with 11 percent ofin the total consumption of the poor recipients (read the “size” of the bubble).

As Figure 4.4 illustratesd in Figure 4.4 that, social programs falls into three categories, if judged by their coverage and targeting.  First, there are three programs that have little coverage and targeting, that can not be justified from a poverty alleviation point of view: the indemnity for politically persecuted, for war veterans and their widows, and scholarships. Fortunately, these programs are minor expenditure items and, except scholarships, have a different social goal, to reward citizens with special merits and their dependents. The efficiency of the scholarship program is questionable. MIn the design phase, most of the scholarships are designed to be needs-basedfulfill social function, with only few of them being awarded for “merit”, a variable that might be unrelated towith the income of the recipient. In practiceEx-post, however, most of the scholarships (73 percent of funds in 1997) went to non-poor, while only 27 percent reached the poor, a figure that is much lower than the share of poor in the total population (30.8 percent). Scholarships seem to have a regressive distributive impact, in favoring of children from middle to rich class. It seems that a thorough review of this instrument is desirable, from eligibility through screening up to benefit delivery, in order to improve its pro-poor impact for the poor.

Second, there are six programs that have good targeting, but relatively low coverage of the poor: the minimum income guarantee scheme, the unemployment benefit, the social pension, the sickness benefit, the maternity and child care benefit and the special aid for the disabled. These instruments each targeted poor households between 54 to 79 percent of the timeto poor households, and covereds between 1 and 13 percent of the poor in 1997. Three of them cover rare social risks (1 percent of the poor), such as maternity, child care, lack of earning ability due to old age, or sickness, and cover a sizablesizeable portion of thepoor’ consumption of the poor. Similarly, the aid for the disabled cover 4 percent of the poor, target 62 percent of funds to this group and contribute with 17 percent to the consumption of beneficiaries. Next, the unemployment benefit performs very well, covering 13 of the total poor population, targeting 68 percent of the program expenditures to this group and contributing with 21 percent to the average consumption of poor beneficiaries. Finally, social assistance recorded a good targeting performance in 1997, except the segment of non-MIG programs in 1997. The minimum income guarantee, a means- tested scheme to provide shield against deep poverty, covers 8 percent of our relatively “generous” category of poor and targets 79 percent of the funds to those in need. The MIG scheme contributes to 22 percent to the consumption of the poor beneficiaries. The coverage of the social assistance is rather limited, although it seemeds to improve in 1997. In this year, one can distinguish between a well-targeted MIG and the other programs, less targeted but with somewhathow better coverage. With respect to (overall) social assistance program efficacy, it reduced from 22 (30) percent in 1995 (1996), to 6 percent in 1997, increasing briefly to 30 percent in 1996.

Third, very isolated, is the child allowance scheme. Designed as a universal benefit, the child allowance scheme covers almost all families with children. It reaches 60 percent of the poor families, the rest being the pensioner households or young families without children. Its targeting efficiency varies, as expected, with the extent of poverty within Romania. In 1997, when poverty was the highest in the period 1995-97, the program targeting was the highest, at 51 percent, compared with 34 percent in 1996 (the lowest headcount rate) and 42 percent in 1995 (the middle headcount rate). The change in the level and formula of the benefit in 1997, although implemented only in March, has as effect an increased in the program efficacy in helping the poor[xxi]. The program increased its impact on the consumption of poor recipients share of program benefit in the consumption of poor recipients raised from 6 or 7 percent of their total consumption in 1995-96, to 11 percent in 1997. One would expect this indicator to perform better in 1998, a year of full program implementation.

From this overall picture, we canlet’s summarize the cash benefit performance on one additional dimension - area of residence: recent studies emphasize tThe high incidence of poverty in rural areas was emphasized in recent studies (World Bank 1997, Wagner et al. 1998, Dinculescu and Chirca 1999, Chirca and  Tesliuc 1999). The rural poor outnumber the urban ones by a factor of two, and the severity is higher in rural areas. Unlike inIn contrast with the urban areas, the rural inhabitants one effectedturned to a dramatic increase in informality, with a widespread increase in unregistered self-entrepreneurship and a sharp reduction in the number of wage earners. Also, rural areas suffers from severe aging, comprisingconcentrating 70 percent of the country’s elderly andwhich explainings thea larger share in pensions ((observed)). However, the other categories of post-transfer poor are poorly served. The unemployment benefit program denies unemployment benefits to those that were laid-off with more than 2 (4) hectareshas of land in plains (hills or mountains).

The above description of the Romanian social safety net leads to the conclusion that almost all cash transfers have a pro-poor targeting (except scholarships, and some merit based benefits), but the coverage of the poor is, except for child allowances, modest. The system addresses specific poverty risks, and many poor households may fall between these risks, and will not be covered by the social safety net. This raises the normative issue of the system’s fairness. Much of the problem is caused because various programs use different poverty lines, or none at all, in assessing eligibility.

Text Box: Fig. 4.5 Social Aid Income Threshold as Share of the Extreme Poverty Line
Source:  Report on social assistance in 1998, 1997, MLSP
The only program that is pro-poor, the social aid program, uses a very low definition for poverty, addressing, by design, only the severe poverty situations. At the beginning (August 1995), the program threshold was close to the "extreme poverty line" used in this paper[xxii] (around 85 percent of it). ThisSuch line covers the poorest decile in 1995, 1997 and 1998. The program threshold was eroded by inflation until the first quarter of 1997, when its real level stabilized at about 50 percent of the "extreme poverty line" (Fig.4.5).

The consequence of this “erosion” of  the program eligibility criteria had dramatic effects on the population that might benefit from this support. The program’s theoretical “theoretic” capacity to cover the extreme poor dropped from the poorest decile to only 0.5 percent of the poorest households. However, the authorities did not assessed the consequence of the “non-indexation policy” on the poor, rushing instead to praisecount on the plus side the lower spending on the program, and the reduction ion the number of beneficiaries. The effect of the measure on the poor was the exclusion of 95% of the initial beneficiaries – those in “extreme poverty” – from its payroll.

By 1997, the program outreach was 0.4 percent of the households, close to its theoretical“theoretic” coverage (0.5 percent). Our expectation that the program covers the extreme poor was invalidated by the data.. Dhanji et al. (1999) estimate that in 1998 only 6.8 percent of those whose consumption is bellow the program threshold benefited from social aid. Although 38 percent of recipients are in extreme poverty, belonging to the poorest decile, only 15.6 percent have a lower consumption than the "program official line". So, the program has a high inclusion error.

In addition to the low coverage and weak capacity in identifying the ultra-poor, the program (which is implemented and funded by the local authorities) seems to have - as suggested by some field studies - an uneven territorial application. We believe that the current funding arrangements for the program are a main cause offor this result. From a normative point of view, it would make more sense to centralizetransfer funding responsibilities to central level. If program funding were left to local resources, then poor communities would not be able to generate as much resources as the richer ones, despite the fact that their needs are higher. Poor communities would face difficult choices in the composition of program expenditures. In practice, the (predictable) opportunistic behavior of local administrators in poor communities aggravates the situation. The program targets a marginal stratum of their constituencies, the ultra-poor. Faced with the choice of servicing this group, or funding benefits that most of the voters would capture (such as health and education, roads, garbage collection and sewage), we were notdid not surprised to findus that the second option was chosen[xxiii]. The program design failed to take into account “political economy” considerations.

We suggest two design changes tTo enhance program effectiveness in achieving its goal, we suggest two changes in program design: a) transfer at least the funding responsibility from local to central level; and b) provide at least the necessary funding necessary to cover at least the “extremely poor”, and therebythus transforming the social aid into a veritable and secure component of the social safety net, able to "catch" those (ultra)poor serviced inadequatelynot assisted or insufficient helped or not at all by other programs. These recommendations are natural extensions of the previous remarks. Inadequate funding in the past has generated either a tightening of the program’s eligibility criteria, or arrears in the disbursement of the benefits. BecauseAs these funds goes to the most needy strata of the population, those who are unable to smooth their consumption over time through saving and dis-saving, thise lack of support in some periods will thrown them into malnutrition, with damaging consequences for their future. Although the payment’s arrears is a way local governments levy “seniorage tax” from their constituencies, it is a pity that such practices hurts the ones in most need.

The implementation of the second recommendation requires two actions: a revision of the program eligibility criteria to cover the “extreme poor”, and an increase in the program’s budget. CTo compliancey with theour definition of “extreme poverty” used in the study would entail, this implies an increase of the benefit by 2.18 times. For instance, in December 1998, the line that would cover the extreme poor would be 328,500 ROL for the first person, instead of 151,000 as provided by the law. As for the funding, this would be equivalent to the extreme poverty deficit. In 1998, this amounted to 1,312 bBillion ROL, compared with the 112 bBillion ROL actually spent. This increase in the program-spending envelope by 11.7 times seems large. However, compared with the overall social budget (34,444 billionBln ROL), this is a minor increase, amounting toof less than four percent. Savings in some poorly- targeted schemes can accommodate this necessary adjustment.


4.3.                        Households and the State: Inequality and Redistribution

The effectiveness of a “welfare state” refers toconcerns its ability to redistribute. In this section, we are asking if the system of cash transfers and taxes existing in Romania redistribute or not. In other words, is the Romanian state taking from the rich and transferring to the poor? To aThe answer to such a question requiresneeds an exhaustive measurement of the transfers and public services received by the households (or, dis-aggregated into specific group of households), as well as the amount of gross income that is taxed (through direct and indirect taxes, and social contributions).

The main purpose of thise section is to analyze the impact of the fiscal and budgetary system on households, by estimating the net flows (effect) between the household and governmental sectors, by consumption decile, poverty status and area of residence. We will build a “household sector account” with Government, able to measure the pro-poor orientation of the whole fiscal and budgetary system. A breakdown into (pre-transfer but pension) consumption deciles (before all transfers except pensions) will provide robust measures of benefit and tax incidence on various income strata.

Text Box: Table 4.2 Summary Household Account with Government, by Decile, 1997
Consumption Decile	1	2	3	4	5	6	7	8	9	10	Col%
I. Pro Memoria, %
Share in total population	10	10	10	10	10	10	10	10	10	10	
Share in total households	6	7	8	8	9	10	11	12	13	16	
Share in total consumption	3	5	6	7	8	9	11	12	14	23	100.0
II. Flows from Government to Households, %
Total Benefits	10	8	8	8	9	9	10	11	12	15	46.2
Cash Transfers	9	7	7	7	8	9	10	11	14	17	28.1
Social Insurance	8	6	6	6	8	9	10	12	15	19	22.6
Entitlements	17	14	13	12	11	9	8	7	6	4	3.6
Social Assistance	24	13	10	9	8	8	8	6	7	6	1.2
Others Cash (non-contributory)	4	6	3	8	11	12	6	12	15	24	0.8
In-Kind Benefits	10	9	10	10	10	9	10	10	10	12	18.1
Health	7	5	7	8	9	9	10	12	14	18	8.0
Primary Care 	5	5	7	7	9	9	11	13	15	19	5.6
Secondary Care 	11	5	8	10	10	8	9	10	11	16	2.5
Education	12	12	11	11	11	10	10	8	7	7	10.1
Primary & Gymnasium	17	14	13	12	11	9	8	7	6	4	4.3
High-School and Vocational	13	13	12	12	11	11	9	7	7	5	3.2
University or Post-High-School	4	6	8	10	12	11	12	13	11	13	2.6
III. Flows from Households to Government, %
Taxes and Contributions	3	5	6	8	9	10	12	13	15	19	51.4
Taxes	3	5	6	8	9	10	11	12	15	22	32.5
Contributions	3	5	7	9	10	11	12	13	14	16	18.9
IV. Net  Flows to Government (taxes minus transfers, ROL Bln. )
Net Balance 	2,481	892	421	-260	-456	-849	-1183	-1170	-1602	-2534	-5.2
Source: Own Computation based on 1997 LSMS
Table 4.2 presents the 1997 household account with Government, by consumption deciles. As apparent from the top section of the table, households were ordered into deciles having an equal number of persons, based on their “pre-transfer but pension” consumption, per adult equivalent. Thus, the first decile will include the poorest 10 percent of the sample and the last decile the richest 10 percent. As expected, larger families are clustered in the poorest deciles, given the negative covariance between welfare level and family size observed in almost all poverty or inequality studies. For instance, the poorest 10 percent of Romanians are living in larger families, accounting for 6 percent of total number of families in the country. In contrast, the richest 10 percent live, in average, in smaller families. Their share in the total number of families in Romania is 16 percent. The final row of the top section of the table gives us the decile-distribution of our welfare indicator. In distributional analyses, this information was used to discriminate between weakly pro-poor to not-for-poor impact of a transfer. In 1997, the poorest decile consumed 3 percent of total household consumption (before transfers but pensions), contrasting with the 23 percent share of the richest decile, of 23 percent.

The second section of the table presents, in a tabular format, the benefits that flow from Government to households: cash benefits like social insurance, entitlements and social assistance, and in-kind benefits like health care and education services. The data in each column represent the share of benefits received by that population group as a share of thein total budget of the program, were all benefits were first deflated in constant prices. We will not do not here an analyzeanalysis of cash benefits here, as this was the topic of  a previous section. We highlight, however, a good anti-poverty performance byof the overall entitlements and social assistance systems, as opposed to “other cash” items. Social insurance has the expected shape. We will comment, briefly, on the distributional impact of the in-kind public services.

The central and local Government is providing a sizablesizeable array of in-kind goods and services to households. However, in terms ofas magnitude, almost all these outlays are spent for health care and education. WFrom survey data, we measured the incidence of these services by household type, using survey information on each service’s take-up. For instance, the use of primary care services by the poorest decile is measured by the share of the first decile’s patients that used public dispensaries or polyclinics in the total number of users in 1997. This share is considered equal with the share of public spending on primary care that the poorest decile benefited from. For secondary health care, we approximated the share of the poorest decile in total spending was approximated by using the percentage of the total number of hospital days occupied by patients from the number of days in hospitals of first decile’ patients in total number of hospital’ days. To estimate the overall use of public health care services, the respective shares were weighted by the public health care costs, by level of health care (primary vs. secondary).

The incidence of education services by household deciles is similar with the one observed in other countries: lower- level education goes primarily to low-income households, while university education does not. In 1997, 17 (13) percent of the pupils in primary and gymnasium (high school) education were located in the poorest decile, compared with only 4 (5) percent in the richest one. In the same year 13 percent of pupils in high school were from the lowest decile, compared with 5 percent from the richest.  In contrast, the share of university students from the poorest decile is only 4 percent, and of those from the richest one is 13 percent. Things are probably better, as our estimate for higher education’s incidence is probably biased against the poor, due to poor information collection[xxiv].

In contrast, the incidence of public health care services is different than in other countries in the region. Some studies (provide source) found that the poor tend to benefit from primary health care, while secondary care is used disproportionately by the rich. The estimates in Table 4.2 contradict thissuch finding. In Romania, use of public secondary health care facilities is rather uniformly distributed across deciles, while the rich use (public) primary health care disproportionately. Sector studies (Tesliuc C, 1999) have drawn a similar conclusion, and have offered as an explanation the fact that a large share of the burden of disease is fromin cardiovascular diseases, whichthat are more commonfrequent among rich than the poor. Also, the poor and the rich do not share disease prevention patterns is not similar for poor versus rich people, the latter one having morea pro-active, preventative behavior. Finally, as we measure the “perceived” health status, we may face data quality problems.

The third section of the table summarizesd the monetary flows from households to Government, in taxes and contributions.

The fourth section of Table 4.2 shows the net flows between households and Government. Apart from inter-decile redistribution, the Government levy throughin taxes and contributions 8,053 billion ROL from the richest deciles (fourth to tenth), out of which it redistributes 3,794 billion ROL to the poorest three deciles. Government net taxation amounts to the difference between these two quantities, or 4,260 bBillion ROL. We will call this quantity 100 percent net tax, because it is net of inter-decile redistribution. From households, the Government collects in taxes 189 percent of the net tax, redistributing 89 percentage points of it from the fourfourth to the richest deciles to the first three poorest deciles. Most of this redistribution occurs to the first (58 percentage points out of 89) and second deciles (21 percentage points out of 89). The answer to our initial question: “Does the Romanian state redistribute” is yes.

The last column in Table 4.2 gives the magnitude of all these transfers, as a percentage of total consumption. In aggregate, the Government is levieslevying from households the equivalent of 52 percent of their consumption, and transfers back 46 percent of the same consumption figure. The net taxation represents slightly more than six percent of total household consumption.

The provision of public goods seems to be well targeted, except for higher education and secondary health care. This result is quite common in all countries in the world. TPartly, this is due in part to the fact that observed because use of these services is correlated with the income position of the beneficiaries (endogeneity problem). One typical argument for the public provision of these services is their associated positive externalities. However, these tend to decrease with the level of the provision (they are greaterare more for primary and secondary education, than for the tertiary; and greateror for primary care thancompared to secondary care). Under tight fiscal constraints, a change in the provision mechanisms of these benefits more inclined for partial cost-recovery should be investigated, targeting (through means tests) the benefits to those unable to pay for them.

On the liability side, households contribute to the state budget quite progressively. The most progressive instruments are the property tax, the entrepreneurship income tax, and the VAT. Wage- related taxes and contributions are weakly pro-poor, as they should be. As property is more unequally distributed than income, an increased collection based on this factorsource will be “welfarist”. The authors reccommend sSuch a development is recommended by the authors, given the low share of thesesuch taxes in the overall fiscal burden of Romanianfaced by households in Romania.

Text Box: Figure 4.5 Net Flows between Government and Households, by deciles
(transfers minus taxes and contributions, 1997)
Source: Own Estimations from the 1997 LSMS Archive
As a conclusion, one can say that the Romanian welfare system is highly redistributive. This result is achieved by the combined distributional effect of cash benefits, provision of public services, and also taxes and contributions. The system improves the welfare of the first three deciles, redistributing a part of the net taxation on the fourth to richest deciles (Figure 4.5). Given the widespread poverty, such a feature is highly desirable. A change in the system parameters from more to less redistribution would not be opportune until growth will absorbs part of the transient poverty.




5.     Policy Conclusions and Recommendations

4.Policy Conclusions And Recommendations

This paper covers three new themes inof the domestic debate on poverty, inequality and effectiveness of social policy in curing poverty. First, it investigates the factors that may take poor out of poverty.  Second, it quantifies poverty by duration and, accentssignaling the large share of transient poor, a social burden that will ease if the economy can be placed again on a growth path.  Finally, the paper assesses the effectiveness of the social safety net in poverty alleviation.

The findings in each of the themes have direct implications for a more effective social policy-making. First and foremost, the paper confirms the centralprime role education plays as a correlate of household welfare, especially infor the urban areas. Second, the paper diagnoses much of the existing poverty as transient, and underscoressignals the essential role of stabilization and growth. Third, the paper concludes that the social safety net functions well in reducing inequality (meeting its redistribution objective), but fails to cover the poor (its poverty alleviation objective). To advance on the poverty alleviation front, the authors endorses a proposal developed together with Dhanji et al. (1999) to improve the operation of the social aid program, estimated to cost 0.33 percent of GDP.

More generally, the analysis supports a poverty-alleviation strategy that have as prime objectives:

·        TtThe completion of the reform agenda on structural adjustment and macrostabilization, two factors that are essential prerequisites for the restoration of the Romanian economy on a sustainable development path;

·        TtThe development of broad-based economic growth, by ensuring that the poor are able to maximize the benefits from growth, by a) providing the necessary framework for broad-based growth;  b) continuing the policies through which the poor gain access to essential assets, including education;  c) increasing the productivity of the poor;  and d) making sure that markets do not discriminate against the poor;.

·        TtThe development of the human capital of the poor, as one of the keys for reducing poverty. Special emphasis should be placed on early childhood development, primary and gymnasium education. The differential in access to education among areas of residence, of other characteristics, should be reduced;.

·        TtThe continuos improvement of the capacity of the social safety net toin alleviateing the poverty, especially among the permanent poor.

Not least, the authors would welcome a more open dialogue on social and poverty issues, based on systematic analysis and monitoring of these phenomenaon, using all the “brainpower” existing in various places in Romania today. We believe that a more transparent policy of information dissemination – including here the dissemination of the micro datasets that are the bread and butter of such analysis – would be benefitcial for the ones that matters: the poor. The more access the research community can hasve to such information, the better the chances that good poverty -alleviation solutions willwould be found. The more open such data and analysis are tocan be used by allany of the stakeholders in Romania, the are better the chances that the deprived willwould get the attention they deserve.





Braithwaite J., Grootaert C., Milanovic B., (1998) “Determinants of Poverty and Targeting of Social Assistance, Eastern Europe and the Former Soviet Union”, World Bank mimeo.

Chirca, D. and Tesliuc, E.D. (eds), (1999), “From Poverty to Rural Development”, National CommissionComission for Statistics

Dhanji, F., L. Pop, C. Tesliuc, and E. D. Tesliuc, “Romania: Social Protection and The Poor”, forthcoming

Davidson, R. and J. Mackinnon (1981) “Several tests for model specification in the presence of multitude alternatives”. Econometrica. Vol.49 pp.781-793

Deininger, Klaus and Squire, Lyn (1996) “A New Data Set Measuring Income Inequality” The World Bank Economic Review Volume 10, Number 3

Dinculescu V. and Chirca, C. (coord.) (1999) “Saracia in Romania: dimensiuni si factori”, in “Saracia in Romania 1995-1998”, PNUD, Proiectul de prevenire si combatere a saraciei.

Glewwe P. (1991) “Investigating the Determinants of Household Welfare in Cote d’Ivoire”, Journal of  Development Economics, Vol. 35.

Green Paper on Regional Development (1996), Government of Romania

Milanovic, B. (1998), “Income, Inequality, and Poverty during the Transition from Planned to Market Economy”, World Bank, Regional and Sectoral Studies

Ministerul Muncii si Protectiei Sociale (1998), "Raport privind activitatea de asistenta sociala in anul 1997"

Ministerul Muncii si Protectiei Sociale (1999), "Raport privind activitatea de asistenta sociala in anul 1998"

Sen, A. (1976) “Poverty: An ordinal approach to measurement”, Econometrica, vol. 44, no. 2

Subbarao, K. et al. (1997) “Social assistance and poverty targeted programs”, World Bank Development Series

Tesliuc, C., (1999), “Economia Serviciilor de Sanatate si Alternative de Finantare”, PhD Thesis, Academia de Studii Economice, Bucuresti

Wagner, P., C. Chirca , C. Zamfir, C. Molnar, and S. Parciog (eds.) (1998) “Metode si tehnici de evaluare a saraciei”, Ed. Expert

Walle, D. van de, (1995), “Incidence and Targeting: An Overview of Implications for Research and Policy”, in D. van de Walle and K. Nead, “Public Spending and the Poor”, Johns Hopkins University Press

World Bank (1995) “Understanding poverty in Poland”

World Bank (1996) “Poverty Reduction and the World Bank.  Progress and Challenges in the 1990’”

World Bank (1997) “Romania. Poverty and Social Policy”

World Bank (1999) “Poverty and social developments in Peru 1994-1997”
Annex 1

Data Sources


4.3.Data Sources

All estimations presented in this paper are based on the Living Standard Measurement Survey implemented by the National Commission for Statistics (NCS), the archives from 1995 to 1998. The sampling frame of LSMS is the Master Sample of Territory Areas - referred to by its Romanian acronym EMZOT – built with data supplied by the January 1992 population and housing census (RPL’92) and comprising housing but not social residential institutions (hospitals, nursing homes, single-person hostels, garrisons, prisons). The master sample allows house sampling according to a two-stage selection scheme. In the first stage, the primary sampling units are survey areas. The second stage sample is composed of 12 monthly sub-samples (waves) and use a total number of 36,072 households. It allows estimations with errors less than 3 percent for all variables with a variation coefficient bellow 262.6 percent. A half of each monthly sub-sample is surveyed again in the same month of next year. This rotating panel feature of LSMS allows us to create, by attrition, a panel of 2,936 dwellings that were investigated in each year from 1995 to 19-97. Unfortunately, NCS renounced to this feature in 1998.


Annex 2


Methodology for Poverty Measurement

Text Box: Box 1. Poverty Indicators
Headcount index, or poverty rate (P0): percentage of people below the poverty threshold.
Consumption shortfall: the percentage by which the average consumption of the poor falls below the poverty line.
Poverty deficit: the sum of all consumption shortfalls, expressed as percent of GDP.
Poverty gap index, or the Foster, Greer & Thorbecke Index (P1): the sum of all consumption shortfalls divided by total population, expressed as percent of poverty line. The poverty gap index reflects the ratio of the cost of perfect targeting (the cost of supplementing the each poor person’s consumption by an amount sufficient to reach the poverty line) relative to he cost of eliminating poverty via untargeted support (by giving each person a transfer equivalent to the poverty line).
Poverty severity index (FGT2 or P2 Index): is a measure to correct the shortcoming of the poverty gap index which does not respect Sen’s “transfer axiom” by giving equal weight to each consumption shortfall regardless the deepness of the poverty gap, by weighting the consumption shortfall of the poor by itself.
Elasticity of headcount with respect to GDP: the percentage point change in poverty headcount divided by percentage change in GDP.
Contribution to total poverty: a measure of the percentage contribution of a subgroup to total poverty.  Complete elimination of poverty within a subgroup would lower total poverty precisely by this percentage

In this paper, we used five different methodologies of poverty measurement.  Three of them were used to allow international comparisons:, such as the absolute poverty lines of one US$ a day;, or four international PPP$ a day, as used in section one of chapter onefirst chapter, section one;, andor the line pioneered by the World Bank for Romania 1989-1994. The other two lines, based on a modified version of the relative method of poverty measurement,  were used in the domestic context, as they reflect better Romanian realities and social policy goals.

Romania does not have an official poverty line. However, recent studies used systematically similar methodologies for poverty measurement systematically (Wagner et al. 1998, Dinculescu and Chirca 1999, ChircaChirc_ and TesliucTe_liuc 1999), building a national “expert consensus”. As the Presidential Commission against Poverty commissioned some of the studies, this “expert consensus” approaches an official acceptance or even support. The core poverty analysis undertaken in this study is done against the same poverty benchmark and methodology as the one used in the above-mentioned studies.

Text Box: Table A2. 1 The “Nutritionist” Equivalence Scale
	Caloric intake	Equivalence coefficients
Boys aged 16  - 20 	3600	1.00
Men aged 21 - 65 	3500	0.97
Boys aged  13 – 15	3100	0.86
Women aged 21 – 56	2900	0.81
Girls aged   13 – 20 	2800	0.78
Children aged  10 – 12	2500	0.69
Children aged   7 -   9 	2100	0.58
Men aged 66+	2100	0.58
Women aged 57+	2100	0.58
Children aged  4 -  6	1700	0.47
Infants aged  2 -  3	1300	0.36
Infants aged   0 -  1 	1000	0.28
Source: Wagner and Chirca (eds) 1998, Methods and Instruments for Poverty Measurement, UNDP
We measured household welfare by the level of (current) consumption of the members of a household[xxv]. The consumption indicator includes all food, non-food and services consumed by the household members during a month, both from purchases and from own production (self-consumption). This indicator does not include imputed consumption of durables or owner-occupied houses, due to the lack of reliable rental prices for these assets, a consequence of their shallow rental markets. To compare the welfare of households of different sizes, consumption was divided by the number of “adult equivalents” existing in the household, using a scale derived by Romanian nutritionists (Table A2.1). The resulting indicator, “consumption per adult equivalent”, was used to rank the households according to their welfare.

All persons living in households with a lower level of consumption per adult equivalent than the poverty line were counted as poor.  The poverty line used in this study is defined as 60 percent of the monthly average consumption per adult equivalent in 1995, equivalent to US$USD 40.

In addition to this core poverty benchmark, we used three other “lines”, one to signal “extreme” poverty in the Romanian context, and two for international comparisons:

Ø      To signal “extreme poverty”, we used a line defined as 40 percent of the average consumption per adult equivalent in 1995, equivalent to US$USD 27 per month.

Ø      To signal the scope of mal-nutrition, we used “the PPP$1 a day” per capita poverty line, for which comparable country poverty estimates are found in World Development Indicators 1999. The reference PPP$ used for international comparisons was the 1985 international PPP$.

Ø      For regional comparisons with other Europe and Central Asia (ECA) countries, we used a 4 PPP$4 per day per capita poverty line, computed for the 1990 international PPP$.

We based our poverty measures on household consumption, not on income. We have considered household consumption expenditure a more reliable measure than income, for two reasons.  First, it reflects better permanent income better, particularly with the LSMS data we are working with. Romanian LSMS useshas the month as its reporting period, and structural features of the Romanian economy makes food self-/consumption an important consumption source. Food self-consumption is derived from agricultural income, usually obtained during summer and fall months. With the reporting period set at one month, the survey data do not capture agricultural income, but do measure instead self-consumption quite accurately. Second, reporting problems may arise with income when asking people to share information about it.  Evidence from nine countries in Eastern Europe and Former Soviet Union confirms that HBS expenditures tend to be higher than HBS income (Milanovic, 1998).

The survey data do not allow us to distinguish between intra-household allocation. Subsequently, our poverty analysis assumes that this allocation is fair, and – for food consumption – is proportional with the nutritional needs of each member, known to vary with age, sex and occupation. The adult equivalent scale used to “standardize” these needs is based on the “normative” caloric intakes recommended by Romanian nutritionists.

To measure the scope, or incidence of poverty, we have used the following indicators: the number of poor and the headcount index. To measure the severity or depth of poverty, we used the following indicators: the consumption shortfall, the poverty deficit, the poverty gap index, poverty severity index and the Gini index. To signal the stability of poverty in time, we computed its elasticity with respect to GDP.

This report defines the poor as the individuals with consumption expenditure per adult equivalent below 70,972 ROL (in January 1995 prices) per month.  This threshold represents 60 percent of the average consumption expenditure per adult equivalent.  The same poverty line has been used for all 4 years we are analyzing.

To compare the rural-urban poverty rate and its year-on-year variations we inflated the 1995 relative poverty line for 40 percent and 60 percent of the average consumption expenditure per adult equivalent using the NCS equivalence scale. All the members of the households under this line were regarded as poor.

Since earlier papers (World Bank, 1997) reported consistently lower food prices in rural areas, that would requiringe deflator corrections by area, we tested this assumption. We determined the average household food consumption in (unit value) prices by area. For 1996, the monthly food basket was three to six per cent lower in rural than in urban area in terms of money – a difference which we did not regard as significant. First, because we used unit values not prices to estimate the purchasinge power differential by area. Given the direct proportionality between price and quality, the lower unit value in rural areas may be linked to a somewhat poorer quality. Second, there are goods that are seldom bought in rural areas where self-own consumption is quite commonfrequent. Unfortunately, accurately determined unit values for such goods are not available.




Annex 3

Statistical AnnexesData: Tables and Regressions Estimation












Text Box: Figure A3.1 GDP per Capita, PPP (Current International US$)

Figure A3.2 GDP Growth (Annual %)
Source: World Bank World Development Indicators 1999
















Table A3.1 GDP per capita, PPP (Ccurrent Iinternational $)

Country Name








































Czech Republic




















Slovak Republic


























Table A3.2 GDP Ggrowth (Aannual %)                                                               

Country Name








































Czech Republic




















Slovak Republic

























Table A3.3 Poverty Profile, by Household Characteristics, 1995 - 1997

Poverty Line: 60% of Mean Household Consumption per Adult Equivalent in 1995

Households' characteristics

Poverty headcount (%)

Distribution of poor (%)

Shortfall as % of poverty line

Poverty gap FGT1

Poverty severity FGT2

Inequality (Gini)







































residence area




































































































































































































































household size



















1 person



















2 persons



















3 persons



















4 persons



















5 persons



















6 persons and more



















no of children



















no children