INTRODUCTION

Poverty rates for the Roma in Southeast Europe are higher than for other vulnerable groups.¹ The Roma not only have less money to spend but also face limited access to education, employment and housing (UNDP, 2002). What makes the living standards of the Roma different from those of refugees, internally displaced persons (R&IDPs) and the majority population that live in a similar socio-economic environment? How do Roma households that fall below the poverty line differ from those of other vulnerable groups' that live on equivalent expenditures – less than 4.3$PPP a day?² Do determinants of welfare, such as education and employment, influence the three groups in a similar way, and to what extent does the Roma's ethnicity contribute to their poverty? This paper will address these questions and analyse the causes of poverty for Roma, R&IDPs and the majority population living in the same socio-economic environment. It will also examine how these groups cope with their poverty. The paper uses data from a new survey, completed in 2004 in Southeast Europe, of Roma, refugees, IDPs and the majority living in close proximity to the Roma.³

POVERTY AMONG THE ROMA

Figure 1 shows that the proportion of the Roma living on less than $4.3 (PPP) a day is considerably higher than the share of the majority population that falls below this poverty line. R&IDPs also face much higher poverty rates than does the majority population, but generally they are still better off than the Roma. (The only exception is Croatia, where the share of R&IDPs living in poverty exceeds that of the Roma.⁴)

Figure 1.

Share of population living on less than $4.30 (PPP) per day.

Full figure and legend (48K)

Lack of education is considered to be one of the main determinants of poverty. Enrolment rates in primary school clearly show that the Roma are worse off than the majority population and R&IDPs⁵ (Figure 2). While enrolment rates are close to 100 per cent for each age group of the majority population, and while they fluctuate between 85 and 96 per cent for R&IDPs, they are below 80 per cent for each Roma age group. For the higher age groups, Roma enrolment rates decline to almost 40 per cent. This drastic gap in enrolment rates for the Roma is one of the main determinants for their low educational achievement and poor job opportunities.

Figure 2.

Enrollment rate in primary education.

Full figure and legend (47K)

Over 60 per cent of the Roma working age poor have not completed primary education. Figure 3 shows the consequences of their low educational attainment on the labour market. Poorly educated Roma and R&IDPs get primarily unskilled jobs, while they hold increasingly skilled positions as their education level increases. The data also suggest that the high Roma unemployment rates are due, to a large extent, to the lack of demand for unskilled labour.

Figure 3.

Type of occupation by education.

Full figure and legend (80K)

The data also suggest that discrimination plays a role on the labour market. In comparison with the majority population, the share of the Roma employed in unskilled occupations is much higher across all educational levels. Even with secondary education, 41 per cent of Roma are employed in unskilled occupations, twice the level of the majority and R&IDPs. The explanation here seems straightforward: once the Roma are identified as Roma, their chances of participating in the labour market at a level suitable to their skill levels shrink dramatically. Their ethnicity offsets other attributes such as 'educated', 'qualified', etc.

Another relation between education and participation in the labour market can be seen when examining the duration of unemployment. Respondents with primary or lower education levels stated that they had not held a job since 1995 or earlier, a period of at least 9 years, more often than respondents with at least a secondary education (56 per cent compared to 46 per cent). The fact that the Roma are unemployed for long periods of time explains why many are ineligible for unemployment benefits and thus must rely on meagre social assistance.

Other measures of human development show that Roma lack housing and basic services in much greater proportions than do the majority population and R&IDPs (Figure 4). Almost 70 per cent of Roma households could not afford to purchase basic medicines over the preceding 12 months compared with 38 per cent of R&IDPs and 30 per cent of the majority population. Twenty-five per cent of Roma live in dilapidated houses or slums, while 11 per cent of R&IDPs and just 3 per cent of the majority population live in these conditions. More critically, over 70 per cent of Roma households lack a toilet or indoor bathroom compared with 22 and 27 per cent for the majority and R&IDPs, respectively. In addition, 22 per cent of Roma households do not have indoor plumbing while the share for the majority and R&IDPs is less than half this share.

Figure 4.

Share of households that have no access to essential drugs, secure housing, improved sanitation, improved water source.

Full figure and legend (42K)

THE COMPOSITION OF THE POOR

The previous section showed that the Roma suffer lower living standards than the other vulnerable groups as illustrated through a number of indicators. This section will analyse the different characteristics of those household members who live in poor households, and compare these to the non-poor population.⁶ The incidence of poverty is usually higher for those with low education, high unemployment and larger household size. Cross-tabulating this 'poor' and 'non-poor' household-member sub-sample with other characteristics such as education, employment and housing reveals a disconcerting picture of poverty among the three groups.

The share of children living in poor households is higher than the share of children living in non-poor households for all three groups. This difference is biggest for the majority households (17 per cent of the non-poor majority are of age 0–14 years – compared to 25 per cent of the majority that live in poor households). In R&IDPs households, the difference is 3 per cent – with an equal high share of children living both in poor and non-poor households. The shares for Roma households are highest while the difference between poor and non-poor households is small. Thirty-four per cent of Roma living in poor households are children compared to 31 per cent of non-poor Roma.

Figure 5 shows that the Roma living in poor households are less educated (only 5 per cent completed secondary education) than the other two vulnerable groups (32 and 39 per cent of majority and R&IDPs, respectively, completed secondary education). Over 60 per cent of working-age poor Roma have not completed primary education. What is especially interesting is that non-poor Roma respondents' education levels are not too different (13 per cent with secondary, 43 per cent with primary and 43 per cent with elementary education), whereas members from non-poor households of the other vulnerable groups have considerably higher education levels than their poor counterparts (almost 50 per cent have secondary education). Lack of education is often considered the first reason for unemployment and a main cause of poverty. Interestingly, all Roma households – regardless of whether they are above or below the poverty line – suffer from very low education levels. This suggests that although low education levels are significantly correlated with the poor household category (=-0.361^**), a substantial proportion of the effect revealed by the bivariate correlation might be composite effects, which can be attributed to the ethnic identity or other socio-economic factors. This relationship will be explored in detail in the subsequent multivariate analysis. It is safe to say that the incidence of poverty for the Roma is likely to be higher than for the other two groups irrespective of educational achievement.⁷

Figure 5.

Composition of the poor by level of education. Note: School age groups are: elementary (6–10 years old), primary (11–15), secondary (16–19) and university (20 and above).

Full figure and legend (49K)

Low levels of education obviously determine literacy rates. Only 66 per cent of poor Roma respondents can read and write, compared with 92 per cent of poor majority and 87 per cent of R&IDPs respondents. Why are Roma education levels so low? Lower school attendance rates for Roma contribute to their low levels of education. Among 6–22 years old, 47 per cent of non-poor Roma respondents and 61 per cent of poor Roma respondents do not attend school anymore, whereas only 18 per cent of non-poor majority respondents and 29 per cent of poor majority respondents do not attend school. All households, regardless of whether they are poor or not, state that the main reason why their children stop attending school is that the costs of education (fees, transport, books, etc) was too high.

Both poor and non-poor households receive income from a variety of sources (Figure 6). All non-poor households receive a high share of their income from state benefits. While for the majority population and R&IDPs, this source of income constitutes the second largest share after wage income, it is the major source for non-poor Roma households. Looking at the poor, work performed in the informal sector constitutes the second largest share of income after benefits for Roma and R&IDPs. For members of the majority population, however, the second largest share of income comes from selling their own agricultural products. This is less dominant for Roma and completely absent for R&IDPs. In contrast, home consumption of agricultural products considerably reduces household expenditures for all groups. If the household were to buy the home production on the market, non-poor households in all groups would pay between 25 and 28 per cent of their monthly average expenditures. In contrast, the share of home production for poor households would be 63 per cent for the majority, 64 per cent for R&IDPs but 104 per cent for Roma. This shows that poor households, especially poor Roma households, depend on subsistence agriculture as a means of coping with poverty, however do not sell their products to increase household incomes.

Figure 6.

Composition of the poor by sources of income.

Full figure and legend (67K)

In terms of employment, the survey shows that differences between poor or non-poor households are not very substantial. Fifty per cent of poor Roma household members stated that they are not working compared to 43 per cent of non-poor. The length of unemployment is very high, leaving Roma households on average 6 years out of the labour market regardless of whether they are poor or not. More significantly, out of those respondents above 15 years that declared unemployment status or being housekeeper, 70 per cent of poor Roma respondents stated that they have never been employed compared to 67 per cent of non-poor Roma respondents. This percentage is much lower for the poor majority and R&IDPs households (57 and 54 per cent, respectively), but lack of experience in the labour market of members of non-poor households of these two groups also reaches 45 and 46 per cent.⁸ The data reveal that lack of experience in the labour market and long-term unemployment seem to be a general problem for the three groups, poor and non-poor households alike – although Roma are again hit harder by long-term unemployment.

Differences in employment between the three groups arise when looking at the kind of work they do. While non-poor Roma mainly occupy jobs that require unskilled workers (37 per cent), non-poor majority and R&IDPs occupy jobs that require skilled workers (29 and 35 per cent, respectively). The poor majority occupy jobs that require skilled workers (24 per cent), unskilled workers (21 per cent) or farming (20 per cent). In comparison, poor Roma and R&IDPs hold mainly unskilled jobs (34 and 27 per cent, respectively) or they are landless workers (16 and 19 per cent, respectively). These numbers show that Roma, regardless of whether they are poor or not, work mainly as unskilled workers, whereas for the majority or R&IDPs, their level of poverty seems to be explained by their lack of skills.

Looking at the data on self-employment, the share of households that have started a business is very low (13 per cent) across all three vulnerable groups. However, looking again at all three groups, twice as many households above the poverty line have started a business in comparison with those below. Most businesses started by members of the non-poor majority population have grown and developed, while businesses established by non-poor Roma and R&IDPs have deteriorated or gone bankrupt. Poor households across all three groups have faced difficulties in keeping their businesses going. Securing financing is usually more difficult for the poor than for the non-poor. Poor households across all three groups face difficulties taking advantage of savings accounts and other financial services. Over 90 per cent of poor households across all three groups do not use savings accounts or money transfers. Seventy-three per cent of majority households above the poverty line take out loans from commercial banks, while a high share of non-poor Roma households borrow from friends or relatives (41 per cent). For the Roma, the loan is often not business related but rather intended for family matters (32 and 43 per cent for the majority population and the Roma, respectively) and for home improvements (37 per cent of R&IDPs).

Poor households across all three vulnerable groups also have significant debts to local utility companies (Figures 7a–c), particularly to the electricity provider. Poor Roma households are, however, in a more critical situation than the other two vulnerable groups in terms of their outstanding debts for water, electricity and housing. Poor Roma households owe the electricity provider 800 per cent of their total monthly expenditures, compared with 132 per cent for non-poor Roma households. Poor Roma households also owe the water utility 374 per cent of their total monthly expenditures in unpaid bills. On average, poor Roma households have not paid their water bills for 20 months and electricity bills for 19 months. The majority and R&IDPs households that are poor also face much higher shares of outstanding payments for water and electricity than do households above the poverty line. For many, particularly the Roma, there is probably no way to break out of this circle of outstanding payments.

Figure 7.

(a) Share of outstanding water payments out of total expenditures. (b) Share of outstanding electrictiy payments out of total expenditure. (c) Share of outstanding housing payments out of total expenditures.

Full figure and legend (76K)

Access to housing is another parameter that illustrates the difficulties suffered by the poor across all three vulnerable groups (Figure 8). The percentage of households living in dilapidated housing or slums is higher for poor households than for households above the poverty line. One-third of poor Roma households live in dilapidated houses or slums, while almost 50 per cent of poor R&IDPs households live in accommodation for refugees (camps or adapted buildings). Two-thirds of poor Roma households and almost half of poor R&IDPs households do not have a bed for each household member compared with 44 and 18 per cent of non-poor households. Over 70 per cent poor Roma households do not have a toilet in their home, compared with about 50 per cent of non-poor Roma households. Over 40 per cent of poor majority and R&IDPs households lack a toilet, compared with less than 20 per cent of non-poor households. With regard to housing, poor households are definitely worse off than non-poor households, but the living conditions of non-poor Roma households is often close to that of the poor.

Figure 8.

Composition of the poor by housing type.

Full figure and legend (66K)

Poor access to healthcare seems to be a problem especially afflicting the Roma and R&IDPs (Figure 9). Although poor households report having a family doctor less often than non-poor households, differences between poor and non-poor households are not very large. Almost 80 per cent of poor R&IDPs households and 48 per cent of Roma households do not have a family doctor. The data also show that only 73 per cent of Roma children and 61 per cent of R&IDPs children living in poor households have received basic vaccinations compared with 80 and 78 per cent of children living in non-poor households. Incidence of vaccination for the children of the majority population is higher than 90 per cent for poor and non-poor households.

Figure 9.

Share of households that cannot afford medicine.

Full figure and legend (31K)

The major reasons put forward why some children have not been vaccinated are: lack of information (33 per cent of poor Roma) or lack of medical ID (73 per cent of R&IDPs). Another reason for the lack of healthcare is the cost of medicine. Over 70 per cent of poor Roma and 60 per cent of poor R&IDPs households cannot afford to buy basic medicine. Again, the difference between poor and non-poor Roma households is much smaller than for the other two groups.

DETERMINANTS OF WELFARE

The previous sections have analysed the different features and determinants of poverty among Roma, IDPs/refugees and the majority population living in close proximity to the Roma. They have demonstrated that the Roma have lower educational attainment rates and fewer employment opportunities and remain dependent on state benefits, often regardless of their level of impoverishment. These findings raise the question whether group identity – the fact that some of those surveyed belong to the Roma ethnic group – determines expenditure-based welfare levels more than the socio-economic factors like education, source of income or regional differences. By using multivariate regression analysis we can factor out the ethnic group component and estimate the partial correlation coefficient between household expenditures and different determinants of welfare while holding all other factors constant.⁹

The model uses the logarithm of equivalent household expenditures as the dependent variable.¹⁰ Explanatory variables are three education dummy variables covering highest achieved educational level of the head of the household, dummy variables for wage, benefits and informal activities as a major source of income of the head of the household, household size, geographic location, vulnerable group dummy variables and country dummy variables to cover country-specific effects.

The three education dummy variables are expected to produce positive returns compared to the lowest level of education (elementary), which will serve as a baseline variable. It is further expected that the benefits and informal-activity dummy variables will be negative compared to wage income, which serves as the base variable. Another characteristic that is influencing the incidence of poverty is the household size, which implicitly includes number of children and should be related negatively to log expenditure. The rural dummy variable should reflect spatial exclusion and the smaller chances of finding employment in rural areas. Thus, the variable should relate to lower expenditure levels. Lastly, the Roma and IDPs/refugees dummy variables are expected to be with negative sign, that is, Roma and IDPs/refugees have lower levels of welfare than the majority sample serving as the baseline.

Table 1 presents simple means and shares of the variables. The data show that for all households (Roma, majority, R&IDPs) the share of households with university education is the smallest; that one-third of households are dependent on benefits as their main source of income; and that most households are in urban areas.

Table 1 - Household characteristics.

Full table

Table 2 summarises the results obtained running the linear regression with the characteristics of each household.¹¹ The most striking result is undoubtedly the large, negative coefficient on Roma. Controlling for other household characteristics, the equivalent expenditures of Roma households is 31 per cent lower than that of the majority households (the base group). R&IDPs face a similarly large differential, their equivalent expenditures are 29 per cent lower than that of the majority households. The other explanatory variables tend to have their expected signs. There is a strong positive association between education and household welfare. But there are noticeable differences in the returns to education across different education levels. Obviously, higher education has most added value relative to elementary education (the base group). Returns to primary, secondary and university education in terms of higher household expenditures are 19, 39 and 68 per cent, respectively, higher than returns to elementary education level. The household size is negatively associated with household expenditures. The benefits dummy is negative and shows the expenditure vulnerability of those households that are dependent on benefits as the main source of income in contrast to wages (21 per cent differential to base group). The informal activities variable has an even stronger impact. It estimates a 32 per cent differential between regular wage and informal income. Households living in the capital experience 16 per cent higher equivalent expenditures than those living in other urban areas (the base group). The higher vulnerability among rural households is reflected in the 15 per cent differential to urban households in terms of welfare.

Table 2 - Determinants of equivalent expenditures, regression results.

Full table

The multivariate analysis confirms the existence of the bivariate relationships outlined in the previous section, that is, that poverty is positively correlated with low education, unstable earnings or state benefits are the major source of income for all groups. The multivariate analysis provides an explanation for the ambiguous observation that Roma tend to show less disparity between certain characteristics regardless their poverty status. Controlling for all these determinants of poverty at the same time allows us to partial out the effect of each individual characteristic as well as the ethnic or group component.

CONCLUSION

This brief analysis has addressed the differences in human development of the Roma compared to the majority living in close proximity to Roma and refugees/IDPs. In particular, similarities as well as differences in poverty correlates and coping strategies between the three groups have been examined. Poverty tends to be more dominant for the following groups: (1) the poorly educated, (2) households that rely on benefits or informal income as major source of income, (3) children, (4) those that have been out of labour force for a long time across all three groups. But the association between poverty and these correlates seem less significant for the Roma than for the other groups. As the multivariate analysis has shown, taking into account all these correlates simultaneously, a strong association between education, source of income and household size nevertheless exists. However, regardless of education or other characteristics, the probability of being poor is substantially higher if one is a Roma or, to a lesser extent, a refugee or IDP, than if one is a member of the majority population.

This paper has also revealed major differences but also similarities in the coping strategies of the Roma versus those of other groups. For example, all vulnerable groups – especially the Roma – engage in small-scale agriculture, which helps to reduce household expenditures. However, only the majority population and R&IDPs sell agricultural products to increase household income. Access to loans and other financial services is severely limited for the poor across all groups. Worryingly, poor Roma households face extremely large outstanding payments that will be extremely difficult to eliminate in the future. These findings suggest that while the Roma need priority attention, refugees and internally displaced persons are also vulnerable groups who face greater-than-average risks of poverty. Unless these problems are addressed in a comprehensive way, according to general welfare (rather than solely ethnic) criteria, poverty reduction strategies are unlikely to be effective or enjoy wider community support.

Notes

¹ In this paper, the term 'vulnerable groups' refers to the three samples 'Roma', IDPs/refugees and 'majority living in close proximity to Roma'. All three groups are predefined as being more vulnerable than the total national population. For the sake of simplicity, throughout this paper, the majority sample will be classified as 'majority' and IDPs and refugees as 'R&IDPs'.

² Equivalent expenditures refers to the OECD equivalence scale, which takes into account economies of scale when calculating expenditures per capita.

³ The survey was conducted in November 2004 in Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Macedonia, Serbia and Montenegro (with separate samples for Serbia, Montenegro and Kosovo) and Romania. In total, 8,273 households were surveyed (3,534 Roma, 3,537 representing majority populations living in close proximity to Roma and 1,202 representing IDPs and refugees) with 34,116 household members' individual profiles recorded (17,270 Roma, 12,548 from the majority populations living in close proximity to Roma and 4,298 IDPs and refugees). In all countries, the survey used identical questionnaires and followed identical sampling and methodological guidelines.

⁴ One explanation for this situation in Croatia could be that a high share of R/IDPs are Roma. Another one could be since refugees most likely fled from a poorer country than Croatia, they face higher levels of poverty in the more developed Croatian environment than the rest of the population.

⁵ Throughout the paper, the data are presented for whole Southeast Europe instead by country.

⁶ 'Poor' household members are those who live in a household whose equivalent expenditure is less than the threshold of $4.3 PPP a day.

⁷ The relationship between unemployment, returns to education and discrimination in the labour market is explored in more detail in O'Higgins and Ivanov (2006).

⁸ The numbers reflect a subjective assessment by those respondents that declared to be unemployed (subjective assessment) or being housekeeper (out of the labour force). Thus, they do not provide an unemployment estimate based on official or ILO methodology and do not imply that respondents did not work or did not earn income. However, the high percentages can have three explanations. People use the agricultural sector, the informal sector and the non-paid homework as survival strategy, because of different type of barriers into the formal labour market (see O'Higgins and Ivanov, 2006).

⁹ A similar analysis for Roma and non-Roma in Bulgaria, Hungary and Romania has been performed by Revenga et al. (2002).

¹⁰ For the purpose of the paper, a linear ordinary least-squares (OLS) model will be sufficient. Expenditures are used to determine welfare, as they better reflect living standards of households because respondents tend to understate their income, especially if a large share of it is earned in the informal sector.

¹¹ The model's explanatory power is above 50 per cent and all household characteristics are significant at the 1 per cent level. Country dummies' coefficients have been estimated but have been omitted from the table for simplicity, Kosovo and Romania turned out to be insignificant.

References

1. O'Higgins, N. and Ivanov, A. 2006: Education and employment opportunities for the Roma. Comparative Economic Studies 48(1): 6–19. | Article |
2. Revenga, A., Ringold, D. and Tracy, W.M. 2002: Poverty and ethnicity: A cross-country study of Roma poverty in Central Europe. The World Bank: Washington, DC.
3. UNDP. 2002: The Roma in Central and Eastern Europe: Avoiding the dependency trap. Regional Human Development Report. UNDP: Bratislava.