Statistical Office of Kosovo Enti i Statistikës të Kosovës Zavod za statistiku Kosova 75436 Consumption Poverty in the Republic of Kosovo in 2009 May 2011 Western Balkans Programmatic Poverty Assessment The World Bank Statistical Office of Kosovo Europe and Central Asia Region Social Satatistics Department Poverty Reduction and Economic Management Unit Living Standards Sector Statistical Office of Kosovo Enti i Statistikës të Kosovës Zavod za statistiku Kosova Consumption Poverty in the Republic of Kosovo in 2009 Western Balkans Programmatic Poverty Assessment The World Bank Statistical Office of Kosovo Europe and Central Asia Region Social Satatistics Department Poverty Reduction and Economic Management Unit Living Standards Sector May, 2011 This report is a joint publication of the World Bank and the Statistical Office of Kosovo as part of the World Bank's Western Balkans Programmatic Poverty Assessment work program (P112410). The World Bank team was led by Ken Simler (ECSP3), with the assistance of Sachiko Miyata and Yeva Gyulnazaryan (consultants), under the supervision of Benu Bidani (ECSP3). The Statistical Office of Kosovo team was led by Bashkim Bellaqa, assisted by Emina Deliu and Besa Haqifi. Alexandru Cojocaru, Agim Demukaj, Borko Handjiski, Ranjit Nayak, Anita Schwarz, and Manohar Sharma (peer reviewer) provided helpful comments on earlier drafts of this report. Generous financial support from the United Kingdom Department for International Development is gratefully acknowledged. . The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgement on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. . Cover photo credit: World Bank Cover design: Sanja Tanić Contents Executive Summary ................................................................................................................................ iii 1. Introduction ................................................................................................................................... 1 2. Poverty levels and changes ........................................................................................................... 2 2.1 Concepts and definitions ....................................................................................................... 2 2.2 Geographic poverty profile ................................................................................................... 4 2.3 Inequality .............................................................................................................................. 7 2.4 Relative poverty .................................................................................................................... 8 3. Poverty profile by household and individual characteristics ........................................................ 8 3.1 Socio-demographic characteristics........................................................................................ 9 3.2 Economic and employment status....................................................................................... 11 3.3 Conditional poverty profile ................................................................................................. 14 Probability of being poor ........................................................................................................... 15 Correlates of consumption per adult equivalent ........................................................................ 16 4. Conclusions ................................................................................................................................. 16 References.............................................................................................................................................. 17 Annexes ................................................................................................................................................. 19 List of Tables Table 1: Poverty and extreme poverty by region, 2009 ................................................................................ 6 Table 2: Distribution of the poor, 2009......................................................................................................... 7 Table 3: Inequality indices, 2009 .................................................................................................................. 8 Table 4: Relative poverty, 2009 .................................................................................................................... 8 Table 5: Poverty and distribution of the poor by main income source of the household, 2009 .................. 12 Table 6: Poverty incidence and distribution of the poor by the main activity of individuals, 2009 ........... 12 List of Figures Figure 1: Poverty and extreme poverty headcount, 2009 ............................................................................. 4 Figure 2: Poverty gap index for full and extreme poverty lines, 2009.......................................................... 5 Figure 3: Sensitivity of rural-urban poverty levels to choice of poverty line, 2009 ..................................... 6 Figure 4: Poverty incidence by age group (left panel) and distribution of the poor and total populations by age group (right panel), 2009 .......................................................................................... 9 Figure 5: Poverty incidence by sex and by sex of the head of household, 2009......................................... 10 Figure 6: Poverty incidence and the distribution of poverty by size of household, 2009 ........................... 10 Figure 7: Poverty headcount by highest level of education completed, 2009............................................. 14 Figure 8: Poverty and distribution of the poor by highest education level completed, 2009 ...................... 14 ii Executive Summary Aggregate macroeconomic data suggest that Kosovo suffered less from the economic crisis than most countries in Europe. Kosovo and Albania are the only countries in South Eastern Europe to record growth in GDP per capita during 2009. Nevertheless, Kosovo remains one of the poorest countries in Europe, and the poorest in the Western Balkans. This report updates information on poverty in Kosovo, drawing on the 2009 Household Budget Survey (HBS). The 2009 survey is the first complete round of the HBS to use the new master sample that was developed in 2008. The new master sample provides a more accurate picture of poverty and well-being, however, the change in sample frame also implies that the results from the 2009 HBS are not directly comparable with previous HBSs. As a result it is not possible to determine the degree to which apparent changes are attributable to having a more representative sample versus actual changes in living conditions. In 2009, slightly more than one-third of the population (34 percent) lived below an absolute poverty line of €1.55 per adult equivalent per day, and 12 percent lived below the extreme poverty line of €1.02. The poverty gap index was 9.6, indicating that on average a poor person in Kosovo attains only 72 percent of poverty line. Poverty levels in urban and rural areas are almost equal, but vary widely across Kosovo’s seven regions. Absolute consumption poverty levels are considerably lower than those reported from the 2005–06 HBS in the most recent World Bank poverty assessment. Although the two surveys cannot be compared reliably, it is likely that at least some of the apparent poverty reduction is genuine when one takes into account the modest growth in GDP per capita and the observation that inequality has remained relatively low and unchanged over this period. Kosovo has a young population, and young people in Kosovo are disproportionately poor, such that over 40 percent of poor people are below the age of 20, and 60 percent of the poor are less than 30 years old (compared to 38 and 56 percent of the overall population below those ages, respectively). There is a strong link between education and poverty, as those who have completed secondary and especially tertiary education are much less likely to be poor than those who have not. The poverty headcount rate in 2009 was 38 percent among those who had only completed primary school, compared to 28 percent among those with secondary education and 11 percent for those with tertiary education. Education alone is not sufficient. More and better jobs are also needed. Nearly one-half of the labor force in Kosovo is unemployed, and unemployment is strongly linked to poverty. The unemployed in Kosovo are disproportionately young and women, with education levels better than the overall population, but not as good as those who have regular employment. Creating good jobs to reduce unemployment—not only for the current cohort of young working-age people, but also for the large population that is still in school—is the primary challenge for reducing poverty in Kosovo. iii 1. Introduction Since the end of armed conflict in 1999, progress on improving living standards in the Republic of Kosovo has been slow and uneven, despite solid macroeconomic performance. In the immediate post-conflict period economic growth was rapid, as GDP grew at double-digit rates, fueled largely by donor-funded reconstruction efforts. After 2005 economic growth slowed to a more moderate pace, with annual GDP growth rates of around 4 percent from 2005 through 2008, driven mostly by donor financing, remittances, and a recovery in economic activity. However, the positive aggregate growth trends had only limited impact on poverty reduction. The Bank’s most recent comprehensive poverty assessment for Kosovo concluded that there was no significant change in poverty levels between 2003–04 and 2005–06 (World Bank 2007b). Kosovo weathered the economic crisis in 2008–09 better than most countries in South Eastern Europe (SEE). Indeed, Kosovo and Albania were the only countries in SEE that registered positive GDP growth in 2009. Even so, this growth was modest, constrained by a worsening investment climate, a drop in foreign direct investment, poor governance and rule of law, and growing rent seeking behavior (World Bank 2010). Remittances from the sizeable out-migrant population also declined following the onset of the crisis. Thus Kosovo remains the poorest country in South Eastern Europe. Kosovo’s GDP per capita in 2009 of €1,760 (World Bank 2010) is the lowest in SEE by a wide margin, and places it as one of the poorest countries in all of Europe. Improving the employment picture is central to reducing poverty and raising living standards through more rapid and more inclusive growth. Kosovo’s unemployment rate is estimated at 46 percent in the 2009 Labor Force Survey (Statistical Office of Kosovo 2010) and 48 percent in the World Bank’s more recent Country Economic Memorandum (World Bank 2010). The employment rate is only 26 percent (World Bank 2010). Creating jobs is even more important because Kosovo’s population is young in comparison to other countries in the region. Approximately one-half of the population is less than 24 years old, and only six percent of the population is older than 65 years old. This note provides an update to the poverty assessment for the Republic of Kosovo that was published in 2007. It is intended to be a concise and timely summary that highlights the key aspects of poverty in Kosovo, rather than a comprehensive analysis of poverty and its determinants. The approach reflects a broader shift in the World Bank’s analytical work on poverty in the Western Balkans, placing greater emphasis on shorter and more frequent poverty updates as new data become available. This poverty update is based on the Household Budget Survey (HBS) conducted in 2009. The new poverty estimates and profile in this update are not directly comparable with estimates from previous years. The adoption of a new master sample for the 2009 HBS confounds comparisons with previous rounds of the HBS, which were based on the increasingly outdated sampling frame from the 1981 census. See Box 1 for more details about issues of HBS comparability over time. This report should be considered as a new snapshot of poverty in Kosovo, and not indicative of trends over time. This report is organized as follows. Section 2 presents estimates of consumption poverty at the national and regional levels, and also describes how inequality has influenced poverty levels in Kosovo. Section 3 develops a detailed poverty profile, examining how poverty is related to numerous individual and household characteristics. Section 4 offers a brief summary of the findings and concluding remarks. 1 Box 1: Poverty estimates in Kosovo are not comparable across years Inferring trends reliably from successive years of HBS data in Kosovo has proven to be problematic because of frequent changes in survey design and because of uncertainty concerning the sampling frame. Volume II of the 2007 Poverty Assessment provides a detailed analysis of the challenges of comparing surveys across years in Kosovo (World Bank 2007b). The Poverty Assessment noted that results from the Living Standards Measurement Study (LSMS) survey in 2000 and successive HBSs from 2002 through 2005 were not comparable because of (a) changes in the HBS from a fortnightly diary to weekly recall questions about consumption, (b) changing definitions of consumption items across successive rounds of the HBS, and (c) differences between the LSMS and HBS in terms of season(s) covered, recall periods for consumption, and length of lists for consumption items. Comparability problems arising from these changes in the design and implementation of the household surveys are compounded by the use of an outdated sampling frame, which through the 2008 HBS was based on the 1981 population census. The outdated sample frame and limited field supervision led to considerable uncertainty about appropriate sample weights, yielding statistical trends in basic demographic data that were hard to reconcile with reality, casting doubt on the representativity of the samples. Comparing the 2009 HBS to previous years is similarly difficult, and such comparisons should be made with caution and considered indicative at best. In an effort to overcome the increasingly unrepresentative sampling frame, in 2008 a new ―master sample‖ was created, based on a new listing of dwellings in selected areas. The new master sample is believed to provide more representative samples, serving as an interim solution until the next census—which is scheduled for April 2011—has been conducted and processed. Unfortunately, one cost of this improvement in representativity is a loss of comparability. Because of the change in sampling, the results from the 2009 HBS should not be compared directly with results from the previous HBSs or Poverty Assessment. This is true even though there were no other major changes in HBS design or implementation in 2009, and identical methods were used for calculating consumption and poverty. The new major changes in the sampling frame of the 2009 HBS confound comparisons with previous rounds of the HBS. Differences observed across years may be because of genuine changes on the ground or because of the 2009 HBS being more representative of the population than the earlier surveys. This applies not only to the overall levels of poverty, but also to the poverty profile. Most likely both of those two factors are responsible for the observed differences, with their relative contributions unknown. The findings in this report are thus best interpreted as updating our understanding of poverty and well-being in Kosovo, and not indicative of trends over time. 2. Poverty levels and changes 2.1 Concepts and definitions As is the case with previous World Bank analyses of poverty in Kosovo, this report focuses on absolute poverty. That is, it compares living standards to a poverty threshold that is held fixed in real terms over time and space. The poverty line is updated over time to account for changes in prices, so that it is sufficient to acquire the same set of basic food and nonfood needs. The absolute poverty approach is distinct from the relative poverty approach employed in much of Europe, and particularly in the European Union. Under a relative poverty approach the poverty threshold increases as the median income of the country rises, and therefore is not fixed in real terms. If all incomes or expenditures doubled (after accounting for inflation), there would be a substantial decline in absolute poverty, whereas relative poverty would remain completely unchanged. The absolute and relative poverty approaches both provide useful and complementary information. However, the 2 absolute poverty perspective is more useful at this time for Kosovo, a lower-middle-income country1 in which a substantial share of the population still struggles to meet basic consumption needs. In this report consumption is used as the measure of individual well-being, or welfare. Household consumption is calculated as the total value of a household’s expenditure on food and nonfood items as recorded in the HBS, including valuation of any home-produced food items that were consumed by the household. In keeping with past practices in Kosovo, expenditures on consumer durable items and rent are excluded from the consumption measure. The adequacy of a given value of household consumption depends on the size and demographic composition of the household, so household consumption is divided by the number of adult equivalents in the household to arrive at the welfare measure, which is consumption per adult equivalent, expressed in Euros at 2009 prices. Consumption is preferred to other monetary measures such as income for several reasons. Consumption is generally a better indicator of current well-being, as it reflects households’ efforts to smooth out income fluctuations. Survey respondents are often engaged in multiple income- earning activities over the course of a year, and the process of recalling and netting out costs can be difficult, especially in settings with large informal economies such as Kosovo. In such settings, respondents are also usually more willing to report expenditures than incomes. We explicitly acknowledge that poverty is a multidimensional concept, and that monetary measures such as consumption only capture one of those dimensions, albeit an extremely important one. Ideally the profile of consumption poverty presented in this report should be complemented by analysis of other dimensions of poverty and deprivation, such as lack of access to public services, safe living conditions, political voice, labor and product markets, and other aspects of living conditions. These dimensions are addressed in other World Bank reports, as well as documents from other development partners, such as the UNDP’s Human Development Report for Kosovo, which focuses on social inclusion (UNDP 2011). Two poverty lines are used in the analysis that follows: a poverty line that is considered adequate to meet basic needs and a lower extreme poverty line. Both poverty lines are based on the poverty lines that were calculated for earlier poverty assessments (World Bank 2005).2 After adjusting for inflation, the poverty line and extreme poverty line are €1.55 and €1.02 per adult equivalent per day (using 2009 prices).3 The two measures of consumption poverty used in this report are the poverty headcount ratio and the poverty gap. The poverty headcount ratio measures how many people are poor, and is simply the percentage of the population whose consumption per adult equivalent is less than the applicable poverty line. The poverty gap measures not only how many people are poor, but also the depth of poverty among the poor, and therefore can detect changes in welfare that occur below the poverty line, such as households becoming less poor, but not enough to cross the poverty line. The poverty gap may be calculated as the product HC, where H is the headcount ratio and C is the ―consumption gap,‖ which is the difference between the poverty line and the average consumption level of poor households, expressed as a proportion of the poverty line. 1 The World Bank grouping of countries by income has four categories, based on gross national income per capita (World Bank Atlas method). At 2009 prices the categories are low-income (US$995 or less), lower- middle-income (US$996–3,945), upper-middle-income (US$3,946–12,195), and high-income (US$12,196 or more). 2 Kosovo’s poverty lines were established using the Cost of Basic Needs method (Ravallion 1994). The food component of the poverty line is anchored to calorie intake of 2,100 kilocalories per person per day, based on the average consumption patterns of households near the poverty line. The nonfood component of the poverty line is based on the share of total expenditures that poor households allocate to nonfood items. The poverty line is the sum of the food and nonfood components. The extreme poverty line is set equal to the food poverty line, meaning that a household that is extremely poor has a total consumption level below the amount required to acquire the poverty reference food basket. 3 The CPI from the IMF’s World Economic Outlook was used to inflate the poverty line to 2009 values. Alternative estimates using the CPI from the Statistical Office of Kosovo produces almost identical poverty lines. 3 2.2 Geographic poverty profile Based on the 2009 HBS, it is estimated that 34.5 percent of Kosovo’s population lives below the poverty line, with 12.1 percent of the population living below the extreme poverty line (Figure 1). As explained in the introduction, the results from 2009 are not strictly comparable with the results from previous surveys because of major differences in the sampling frames of the two surveys. Across a number of non-comparable surveys the poverty headcount has routinely oscillated between mid-30 percent and mid-40 percent (World Bank 2007a).4 That said, the difference in the poverty rates between 2005–06 and 2009 are generally consistent with Kosovo’s GDP growth over that period, which was modest but steady through 2009 despite the economic crisis.5 Figure 1: Poverty and extreme poverty headcount, 2009 35.3 34.5 30 33.1 Population (%) 20 10 13.0 12.1 11.6 0 Urban Rural Overall Urban Rural Overall Poverty Extreme poverty Source: World Bank staff estimates based on 2009 HBS. Poverty rates in 2009 are almost the same in urban and rural areas—poverty is slightly higher in rural areas, whereas extreme poverty is slightly higher in urban areas. Like the overall poverty rate, the rural-urban poverty differential has fluctuated widely over successive non- comparable surveys, from 12 percentage points higher in urban areas in 2002 –03 to 12 percentage points higher in rural areas in 2005–06 (World Bank 2007b). Although it is not possible to disentangle the possible causes, it is clear that poverty in 2009 is as much an urban phenomenon as it is rural. The depth of poverty—as measured by the poverty gap index—and extreme poverty are nearly equal in rural and urban areas in 2009, with poverty levels slightly higher in urban areas (Figure 2). In 2009, on average a poor person’s consumption level was 72.2 percent of the poverty line. Among those below the extreme poverty line the corresponding figure is 77.7 percent. 4 The poverty headcounts reported in the previous poverty assessment are 37.7 in 2002 –03, 43.7 in 2003– 04, 34.8 in 2004–05, and 45.1 in 2005–06. 5 GDP per capita grew 20.6 percent between 2005 –06 and 2009, and 13.1 percent in per capita terms (World Bank 2010). A change in the poverty rate from 45.1 to 34.5 percent is a 23.5 percent reduction, implying a growth elasticity of poverty of -1.8, which is well within that observed for other countries (Ravallion 2001). Also note that even though survey and national accounts estimates of consumption typically differ (Ravallion 2003), the estimate of growth in consumption per capita between the two surveys (7.6 percent) is reasonably close to the corresponding estimates from the national accounts (11.3 percent). 4 Figure 2: Poverty gap index for full and extreme poverty lines, 2009 10 9.8 9.5 9.6 Poverty gap index (%) 5 2.9 2.6 2.7 0 Urban Rural Overall Urban Rural Overall Poverty Extreme poverty Source: World Bank staff estimates based on 2009 HBS. The ranking of rural and urban poverty changes depending on whether the full poverty line or the extreme poverty line is chosen. This is illustrated more generally for a wide range of poverty lines in Figure 3. The curves are the empirical cumulative density functions (CDF) of consumption per adult equivalent, also known as poverty incidence curves. The vertical line in the center of the figure is the extreme poverty line (€1.02 per day) and to its right is the full poverty line (€1.55 per day). One may read the poverty headcount ratio from the vertical axis, depending on where the CDF intersects the poverty line. From Figure 3 it may be seen that for most poverty lines below €1.35 per day, the poverty headcount ratio is slightly higher in urban areas than in rural areas. Conversely, for poverty lines higher than €1.35 per day, the headcount is higher in rural areas, with the difference increasing considerably as one moves to poverty lines of €2.00 per day and higher. 5 Figure 3: Sensitivity of rural-urban poverty levels to choice of poverty line, 2009 80 2009 rural 2009 urban 60 Population (%) 40 Extreme poverty line 20 Poverty line 0 0.50 1.00 1.50 2.00 3.00 Consumption per adult equivalent per day (2009 Euros, log scale) Source: World Bank staff estimates from 2009 HBS Poverty rates across Kosovo’s seven regions vary widely, as shown in Table 1. According to the 2009 HBS, the poverty headcount ratio ranges from a high of 54 percent in Ferizaji and Gjakova to a low of 18 percent in Gjilani. The poverty gap and poverty measures using the extreme poverty line show similarly wide ranges and fairly consistent ranking of the regions across the different measures. The regional poverty profile in 2009 has some major differences from the 2005 – 06 regional profile. It is expected that some of the changes in the regional poverty estimates are attributable to the use of the new master sample in 2009, especially because the smaller sample sizes at the regional level imply that the estimates from any single survey are less precise. Table 1: Poverty and extreme poverty by region, 2009 Full poverty line Extreme poverty line Region Headcount Poverty gap Headcount Poverty gap Ferizaji 53.8 14.6 17.7 3.7 Gjakova 54.0 16.3 22.4 5.9 Gjilani 18.0 3.8 3.6 0.9 Mitrovica 38.0 12.2 17.0 4.2 Peja 37.2 10.9 13.1 3.8 Pristina 21.8 6.2 9.0 1.3 Prizreni 33.9 7.6 6.8 1.0 Overall 34.5 9.6 12.1 2.7 Source: World Bank staff estimates based on 2009 HBS. Even though the rural and urban poverty rates in Kosovo are almost equal in 2009, nearly two-thirds of poor people live in rural areas, simply because Kosovo is predominantly rural. Table 2 combines information on poverty rates with the geographic distribution of the population to 6 show which areas have larger or smaller numbers of poor people. At the full poverty line there are remarkably equal numbers of poor persons in five of the seven regions (Ferizaji, Gjakova, Gjilani, Mitrovica, Pristina, and Prizreni), with each holding 16 to 17 percent of Kosovo’s poor. The remaining 17 percent are split between Peja (11.0 percent) and Gjilani (5.8 percent). Extreme poverty in 2009 is more geographically concentrated. Gjakova and Mitrovica are each home to more than 20 percent of the extremely poor population, and Pristina holds almost as many (18.7 percent). In contrast, Gjilani, Peja, and Prizreni combined account for only 23.4 percent of the extremely poor population. Table 2: Distribution of the poor, 2009 Distribution of the Distribution of the extreme Area/Region population Distribution of the poor poor Urban 49.8 35.6 39.8 Rural 50.3 64.4 60.2 Overall 100.0 100.0 100.0 Ferizaji 20.0 17.1 16.0 Gjakova 19.6 17.3 20.4 Gjilani 6.8 5.8 3.3 Mitrovica 15.3 16.8 21.4 Peja 14.0 11.1 11.0 Prishtina 10.6 15.9 18.7 Prizreni 13.8 16.0 9.1 Overall 100.0 100.0 100.0 Source: World Bank staff estimates based on 2009 HBS. 2.3 Inequality Overall inequality in Kosovo is relatively low. Table 3 presents an array of inequality measures at the national level and disaggregated by rural-urban and regional areas of residence. Each of the inequality indices places greater weight on inequality at different parts of the consumption distribution. The Gini coefficient is the most widely used summary statistic of inequality, and gives more weight to inequality around the median of the distribution. The generalized entropy (GE) measure is a parameterized index, with lower values of the parameter assigning higher weight to inequality in poorer parts of the distribution, and higher values of the parameter assigning greater weight to inequality in richer parts of the distribution.6 Consumption inequality is considerably higher in urban areas than in rural areas. This is a common finding across the world. All four of the inequality measures reported in Table 3 are consistent in showing higher inequality within urban areas, and lower inequality within rural areas. Among regions, Mitrovica, Pristina and Gjakova have the highest levels of consumption inequality, while Prizreni and Ferizaji have the lowest. 6 When the parameter is set to zero (GE(0)) the index is also known as the mean log deviation index, or Theil L. When the parameter is set to one (GE(1)) the index is also known as the Theil T index. 7 Table 3: Inequality indices, 2009 Area/Region Gini GE(–1) GE(0) GE(1) Urban 33.9 22.3 19.0 19.7 Rural 27.2 14.2 12.4 12.7 Ferizaji 24.4 11.0 10.0 10.4 Gjakova 29.2 16.8 14.4 14.7 Gjilani 25.5 12.1 10.9 11.6 Mitrovica 32.4 20.3 17.4 17.6 Peja 28.0 16.4 13.5 13.0 Pristina 31.4 19.5 16.7 17.0 Prizreni 19.1 6.6 6.1 6.0 Overall 30.2 17.3 15.2 15.9 Source: World Bank staff estimates from 2009 HBS. 2.4 Relative poverty Although this report focuses on absolute poverty, it is nonetheless instructive to examine changes in relative poverty, which is effectively another measure of inequality. Unlike absolute poverty lines, the real value of a relative poverty line—that is, the minimum acceptable standard of living—changes over time instead of remaining fixed. The Eurostat standard for ―risk of poverty‖ is 60 percent of median national income, which we adapt here to consumption to arrive at a daily poverty line €1.13 per adult equivalent in 2009. Note also that this is significantly less than the absolute poverty line used in the rest of the report, and corresponds to a standard of living much closer to the extreme absolute poverty line of €1.02 per day. At this poverty line, both the poverty headcount and the poverty gap are higher in urban areas than in rural areas. Table 4: Relative poverty, 2009 Poverty headcount ratio Poverty gap Urban 17.1 4.1 Rural 16.1 3.7 Overall 16.4 3.8 Source: World Bank staff estimates from 2009 HBS. 3. Poverty profile by household and individual characteristics This section examines how poverty is related to household and individual characteristics in two ways. One way is by comparing poverty rates across different population sub-groups, which tells us the likelihood that someone is poor given certain characteristics such as age, sex, or employment status. The second way describes the characteristics of those who are below the poverty line. For example, the poverty rate for adults who have not completed even primary education is very high. However, such individuals make up a small part of the overall population, and also a small part of the poor population. By looking at these two perspectives together it is possible to see not only which groups are most likely to be poor, but also which groups make up a large proportion of the poor population, which are not always the same thing. One limitation of the analysis of individual characteristics is that poverty status is defined at the household level, based on household consumption divided by the number of adult equivalents in the household. Like most household surveys, the HBS does not collect sufficiently detailed consumption data at the individual to distinguish different standards of living among individuals living in the same household, although there is no doubt that such differences do exist. The impact of this limitation is 8 that the analysis of poverty by individual characteristics is not as sensitive to differences across groups as it would be if actual individual-level consumption data were available. 3.1 Socio-demographic characteristics Kosovo’s youth are disproportionately poor. Kosovo has an exceptionally young population by European standards, with an estimated one-third under the age of 15, and approximately one-half younger than 24. The left panel of Figure 4 shows that poverty rates are highest among those less than 20 years of age, and then taper off slowly before rising again among the elderly. As young people have both high poverty rates and constitute a large share of the overall population, they are a very large share of the poor population. The right panel of Figure 4 shows the composition of the poor population by age group (shaded area) compared to the age composition of the total population. In 2009 more than 40 percent of the poor were below age 20, and 60 percent were below age 30. This compares to 38 and 56 percent of the total population in those age groups, respectively. Providing adequate education and employment opportunities for this large and growing segment of the population is one of the major challenges confronting Kosovo now and in the medium- term. Figure 4: Poverty incidence by age group (left panel) and distribution of the poor and total populations by age group (right panel), 2009 50 25 Poverty headcount ratio (%) Poor population Share of population (%) 40 20 Total population 30 15 20 10 10 5 0 0 Age group (in years) Age group (in years) Source: World Bank staff estimates based on 2009 HBS. Persons living in female-headed households tend to be slightly poorer than those in male- headed households. Figure 5 shows the relationship between gender and poverty. The left side displays the poverty headcount for males and females from the 2009 HBS, showing that the poverty rate is one percentage point higher among females than it is among males. The right side shows the poverty headcount ratios by the sex of the head of household. According to the 2009 HBS approximately five percent of the population lives in households headed by women, and poverty is 3 percentage points higher in those households than in households headed by men. 9 Figure 5: Poverty incidence by sex and by sex of the head of household, 2009 40 Male 37.2 Poverty headcount ratio (%) 35.0 Female 34.3 33.9 30 20 10 0 Sex of individual Sex of head of household Source: World Bank staff estimates based on 2009 HBS. Large households tend to be poorer, and account for a large proportion of the poor population. Households in Kosovo tend to be larger than those in other European countries, averaging six people per household. As seen in the left panel of Figure 6, poverty rates are highest for large households. About 11 percent of households in Kosovo have 10 or more persons, accounting for approximately 23 percent of the population. Thus given their higher rates of poverty it is not surprising that approximately one-quarter of poor people live in households of ten or more people, as shown in the right panel of Figure 6. Substantial shares of poor people also live in households of five to seven people. Figure 6: Poverty incidence and the distribution of poverty by size of household, 2009 50 25 Poor population Poverty headcount ratio (%) 40 Total population Proportion of population 20 30 15 20 10 10 5 0 0 2 3 4 5 6 7 8 9 10+ 1 1 2 3 4 5 6 7 8 9 10+ Number of people in household Number of people in household Source: World Bank staff estimates based on 2009 HBS. 10 3.2 Economic and employment status Consumption poverty is closely related to the economic activities of a household, especially employment status and, if employed, the sector and type of employment. In turn, employment status and returns to employment are heavily influenced by a person’s educational level, as those with higher levels of education are typically more likely to find employment and to have access to higher paying jobs. Not surprisingly, poverty rates are highest among those whose main source of income is social assistance. The HBS asked households what their main source of income was, and the responses are summarized in Table 5. For most recipients the social assistance payments reduce the depth of their poverty, but are insufficient to raise them above the poverty line. Other groups with higher-than-average poverty rates are households that depend on casual employment (per diem work) and those that depend on pensions. The standard pension of €45 per month is almost equal to the monthly poverty line of €46.50, and pensioners with a contribution history receive more. Therefore households consisting solely of pensioners are unlikely to be poor, but those in larger households without other significant sources of income are likely to be poor because the pensions are not adequate to support the larger family. One implication of this is that the most efficient way to reduce poverty among pensioners may not be to increase pensions, but rather to improve income-earning opportunities for the working-age population. Poverty rates are lowest among households that depend primarily on remittances from abroad, public sector wage employment, and household businesses (self-employment). Out- migration of labor from Kosovo has traditionally been high, with the main destination countries being Germany and Switzerland. It is estimated that in 2009 remittances from migrants totaled approximately €500 million, making remittances the largest source of external income, equaling 13 to 15 percent of total GDP (World Bank 2010). According to the 2009 HBS, 14 percent of households receive remittance income, averaging €316 per month among receiving households, although the median is only €200 per month. Although households receiving remittances are less likely to be poor, the extent to which this can be attributed directly to remittances is not clear. The HBS data indicate that most nonpoor households receiving remittances would not be poor even if they did not have remittances, but it is also widely recognized that remittances are likely to be undercounted by this type of survey. Nearly 40 percent of the poor live in households where the main source of income is wages and salaries (Table 5). This is not inconsistent with the high rates of unemployment and low rates of employment noted earlier. An average household in Kosovo has six members, four of whom are adults. In many cases one or two members may have wage employment, but that income is not sufficient to meet the needs of the entire household, especially when the other adults do not have regular employment. It is also important to note that these households with high rates of adult unemployment are not eligible for social assistance under the current rules of the system, which exclude households in which any person is employed, regardless of how many adults in the household are not employed. This explanation is supported by Table 6, which presents information at the individual level on the relationship between poverty and a person’s main activity, limited to persons 15 years of age or older. The highest rates of poverty are found among per diem workers and the unemployed. Turning to the composition of the poor population we see that nearly 30 percent of poor adults are unemployed.7 In 2009 the next most common principal activities among poor adults are housekeepers (19 percent), students (16 percent), and the retired or disabled (12 percent). 7 See Box 2 for a description of the characteristics of unemployed workers in Kosovo. 11 Table 5: Poverty and distribution of the poor by main income source of the household, 2009 Household’s main source of Poverty headcount Distribution of the Distribution of the income ratio poor population Public sector wages and salaries 23.6 17.3 25.2 Private sector wages and salaries 36.4 20.7 19.6 Farming 36.3 8.2 7.8 Per diem work 55.2 17.5 10.9 Other household business 20.6 8.7 14.5 Pensions 43.4 7.6 6.1 Remittances from abroad 21.9 5.5 8.6 Other remittances 32.9 0.3 0.3 Social assistance 83.1 12.6 5.2 Other 32.5 1.6 1.7 Overall 34.5 100.0 100.0 Source: World Bank staff estimates from 2009 HBS. Table 6: Poverty incidence and distribution of the poor by the main activity of individuals, 2009 Poverty Distribution Distribution of headcount ratio of the poor the population Main Activity 2009 2009 2009 Employer 6.9 0.2 0.8 Employed with salary 20.5 10.1 16.2 Subsistence farmer 33.0 3.4 3.4 Per diem worker 50.1 6.1 4.0 Other self-employed 20.1 2.0 3.2 Unpaid family worker 15.9 0.4 0.8 Pupil/student 32.3 16.2 16.5 Retired or disabled 32.3 11.8 12.0 Unemployed 39.0 29.7 25.0 Housekeeper 35.8 18.8 17.2 Other 48.1 1.5 1.0 Overall 32.8 100.0 100.0 Source: World Bank staff estimates from 2009 HBS. The cross-sectional data from the HBS show a clear relationship between education and poverty in Kosovo. Education is one of the most reliable routes out of poverty, as it is associated not only with improved access to better job opportunities and often higher productivity in those jobs, but also with welfare-enhancements outside the workplace. Figure 7 shows the poverty headcount ratios for individuals 15 years or older by the highest level of education completed. In both surveys the reduction in poverty rates increases as one moves up the education ladder. That is, completing primary education leads to slightly better poverty outcomes than not completing primary, and the incremental gain is larger when going from complete primary to complete secondary, and even larger when going from complete secondary to complete tertiary. 12 Box 2: Who is unemployed? As noted in the introduction, Kosovo’s unemployment rate is nearly 50 percent, the highest in South Eastern Europe and one of the highest anywhere in Europe. Expanding employment opportunities is essential for reducing poverty in Kosovo, and it is equally important to know the target group for employment creation efforts. This box uses the 2009 HBS data to sketch a brief profile of the unemployed. Table 6 shows that among poor people age 15 and over, more people are unemployed than any of the other main activity categories listed in the HBS, and by a wide margin. The table below shows that this is true for the overall adult population as well. In fact, the number of unemployed is almost as large as all of the employed categories combined (employer, salaried employee, subsistence farmer, other self-employed, per diem worker, and unpaid family worker). Salaried Other House- Overall Unemployed employees employed* keepers Population share (17+ years old) 100.0 26.3 17.2 12.9 18.3 Mean age (years) 40.1 32.5 39.6 38.5 44.1 Less than primary education (%) 13.7 4.0 0.5 3.9 19.7 Completed primary education (%) 38.2 40.1 9.7 35.9 33.9 Completed secondary education (%) 42.6 51.4 74.5 56.8 13.8 Completed tertiary education (%) 5.5 4.5 15.8 3.4 0.0 Females (%) 50.5 59.6 23.7 5.5 99.8 Urban (%) 38.2 37.5 52.5 33.2 31.6 Household size (persons) 7.2 7.5 6.5 7.8 7.3 Female-headed household (%) 5.6 6.2 4.7 3.9 6.0 *Note: ―Other employed‖ includes employers, subsistence farmers, other self -employed, per diem workers, and unpaid family workers. Source: World Bank staff estimates from 2009 HBS. The unemployed tend to be considerably younger than those who are employed, with an average age of 32.5 years, compared to 39.6 among salaried employees and 38.5 for other employed persons. Despite having slightly more education than the general adult population, the unemployed have much less education than salaried employees, and slightly less education than other employed persons. Among the unemployed, 40 percent have completed only primary education, 51 percent secondary, and 4 percent tertiary, compared to 10, 74, and 16 percent, respectively, among salaried employees. Approximately 60 percent of the unemployed are women, whereas only 24 percent of salaried employees are women. Women make up an even smaller percentage (6 percent) of other employed workers. That women are such a large proportion of the unemployed is even more striking when one considers that almost all housekeepers are women, who make up 18 percent of the population (36 percent of women). If suitable employment opportunities were available and other conditions permitted, many of the women who are currently housekeepers would likely enter the labor force. On other characteristics, the unemployed are very similar to the general population, with approximately equal rates of urbanization, sizes of households, and percentages residing in female-headed households. In contrast, those working in salaried positions are more likely to be urban, come from smaller households, and have a male head of household. 13 Figure 7: Poverty headcount by highest level of education completed, 2009 Poverty headcount ratio (%) 40 39.9 37.9 30 28.0 20 10 11.0 0 Less than Primary Secondary or Tertiary complete primary Vocational Source: World Bank staff estimates based on 2009 HBS. Another perspective on education and poverty, based on the 2009 HBS, is shown in Figure 8. The line shows the same relationship between poverty headcount and education level as the preceding figure. The vertical bars show that approximately 40 percent of Kosovo’s adult population h as completed only primary school, and an additional 40 percent have completed as far as secondary or vocational education. About 15 percent have not completed primary and only about five percent have completed tertiary. When comparing the distribution of the poor (light bars) with the distribution of the overall adult population (dark bars) it is clear that those with primary education or less are disproportionately poor, and that conversely, those with secondary or higher education are less likely to be poor. Figure 8: Poverty and distribution of the poor by highest education level completed, 2009 50 40 30 20 10 0 Less than Primary Secondary or Tertiary complete primary Vocational Distrib. of the poor Distrib. of the population Poverty headcount Source: World Bank staff estimates based on 2009 HBS. 3.3 Conditional poverty profile The preceding sections have presented what may be called an ―unconditional‖ poverty profile, which describes the bivariate associations between poverty and a particular individual or household characteristic, without controlling for other characteristics. For example, it was noted that female- 14 headed households tend to be poorer (Figure 5) and that people with more education are less likely to be poor (Figure 8). However, female heads of households have less education, on average, than their male counterparts, so it is not clear if the poverty outcome should be attributed to education, or gender, or both. In this section we present a ―conditional‖ poverty profile that uses econometric methods to examine the relationship between an individual or household characteristic and poverty, while controlling for other characteristics. The first set of analyses uses probit models to estimate the relationship between a change in one characteristic and the probability of being below the poverty line, while holding other characteristics constant. The second set of analyses takes a slightly broader view, and uses ordinary least squares (OLS) regression to analyze the percentage change in consumption per adult equivalent associated with a change in a given characteristic, again holding other variables constant. The main advantage of the first approach is that it shows a clear link between characteristics and the probability of being below the specific poverty line used in Kosovo. The main advantage of the second approach is that it allows for gradations of well-being by using the full consumption distribution rather than the discrete poor/nonpoor dependent variable of the probit analysis. For both approaches the models are estimated separately for rural and urban areas. Probability of being poor When controlling for other variables, the probit analysis shows that household size only has a statistically significant effect in urban areas, in which there is an inverted-U pattern where the probability of being poor peaks around 11–12 people (Table A1). Holding other variables constant, the sex of the head of household is not a significant predictor of whether or not a household is poor. The age of the head of the household is significantly correlated with poverty status of the household only in rural areas, following an inverted-U pattern that peaks at 44 years. After controlling for other variables, the age and sex composition of the household does not have a strong relationship with the probability of being poor in rural or urban areas. In both areas the only significant household composition coefficient is that for men from 17 to 64 years old—other things being equal, households with more men in this age group are less likely to be poor. Education of adult household members (defined here as age 17 or older) is significantly related to the probability of being poor in both rural and urban areas and in both HBSs. As expected, the coefficients become consistently more negative (i.e., associated with a lower probability of being poor) as one moves to higher levels of education. In urban areas the coefficient for completed primary education is not significant in either survey, indicating that increasing the proportion of adults with secondary or higher education is needed to reduce the probability of being poor in urban areas. In contrast, in rural areas there are significant poverty reduction advantages from completed primary education, although the advantages are greater for secondary and tertiary education. Urban households that depend primarily on social assistance and pensions have higher probabilities of being poor than those depending on other sources, other things being equal. Reliance on income from self-employment and remittances is associated with lower probabilities of being poor in urban areas. In rural areas the patterns are similar, although pensions are not statistically significant. When it comes to the main activities of adult members of the household it is not surprising that households with larger shares of working-age members who are unemployed are more likely to be poor, although the coefficients are only significant in urban areas. Those with higher proportions of wage or salaried workers are significantly less likely to be poor. There are pronounced differences in poverty across regions even when one controls for other variables. The last variables in the probit equations are a set of regional binary variables, showing the differences in the probability of a household in the indicated region being poor relative to a household with the same characteristics in Pristina region. For example, urban households in Gjakova region are 25 percent more likely to be poor than similar households in Pristina region, whereas households in urban Gjilani region are 12 percent less likely to be poor than their 15 counterparts in urban Pristina. Regional patterns are also statistically significant in rural areas, with rural households in all regions except Gjilani more likely to be poor than similar households in rural Pristina region. Correlates of consumption per adult equivalent The consumption regressions use the same explanatory variables as the probit analysis, but the dependent variable is the natural logarithm of consumption per adult equivalent. Thus the coefficients from the regressions (Table A2) may be interpreted as the percentage change in consumption from a unit change in an independent variable, holding other variables constant. Because the dependent variable is log consumption, the interpretation of the signs of the coefficients is the opposite of probit analysis, in that positive coefficients are associated with higher levels of welfare rather than higher probabilities of being poor. After controlling for other variables, the age and sex of the head of household are not significant predictors of welfare levels. In the consumption regressions neither the sex of the household head nor the head’s age are significantly correlated with consumption in any of the regressions. Household size is a significant correlate in urban areas only, where consumption per adult equivalent decreases (at a decreasing rate) as household size increases. The effects of the age and sex composition of the household are stronger in rural areas than in urban areas. The education levels of household members are positively related to consumption even after controlling for other factors. All of the coefficients on the education variables are statistically significant. They are qualitatively very similar to those from the probit analysis, with the size of the coefficients increasing at higher levels of education, and with stronger results in rural areas for completing only primary education. Households that depend primarily on social assistance have significantly lower consumption than other households. Remittances are strongly associated with higher consumption in rural areas, but not in urban areas. Conversely, dependence on pensions, per diem work, and wage employment are associated with lower consumption in urban areas, but are not significant in rural areas. The unusual negative association between wage employment and consumption shown in the main household income variables is better understood in combination with the significant positive coefficients for the share of household adults with wage or salary employment. As described earlier in section 3.2, many poorer households have one or two members who have wage or salaried employment, but also have several other adult members who are unemployed or have less remunerative employment. This is also consistent with the negative coefficients for the proportion of working-age adults who are unemployed and the share of adults who are unpaid family workers. The results for the regional variables are largely similar to the findings in the probit analysis, with most regions having lower consumption levels than Pristina region, the sole exception being urban Gjilani. 4. Conclusions Even though Kosovo appears to have suffered less from the economic crisis than most countries in Europe, it remains one of the poorest countries in Europe. Analysis of the 2009 HBS indicates that more than one-third of the population lives below the poverty line of €1.55 per adult equivalent per day, and 12 percent are below the extreme poverty line of €1.02 per adult equivalent per day. Although these levels are lower than those reported from the 2005–06 HBS (45.1 and 16.7 percent, respectively) it is not possible to draw firm conclusions about poverty trends because of changes in the HBS sample frame that were introduced in 2008. Kosovo’s GDP growth since 2005– 06 has been slow but consistently positive, so it is plausible that at least some of the apparent reduction in poverty is real and not merely an artifact of the change in sampling methods, especially if the HBS’s indication that inequality has not increased significantly is correct. Once it is available, the 16 sampling frame from the 2011 census will help provide a stronger basis for comparison across future surveys, although comparability of survey implementation will still be required. The 2009 HBS shows remarkably similar rates of poverty in rural and urban areas, although this finding is sensitive to the choice of poverty line. Urban poverty is slightly higher than rural poverty for poverty lines lower than €500 per adult equivalent per year. For higher poverty lines rural poverty is greater, with the rural-urban difference growing as one approaches poverty lines of €750 or €1000 per year. Kosovo’s population is exceptionally young, and children and youths are disproportionately poor. Related to this, large households (which tend to have a larger share of young members) have higher poverty rates, and constitute a large proportion of the poor population in Kosovo. Females are only slightly more likely to be poor than males, although the difference in poverty rates between female- and male-headed households is more pronounced. Education and employment are intimately related to poverty status. Education has intrinsic value in improving well-being, and it also has instrumental value in that it expands opportunities. Those who have completed secondary education or especially higher education are much less likely to be poor than those who have only completed primary school. Part of the explanation for this is that those who have more education are more likely to be employed, and to receive higher earnings from their employment. However, even among the unemployed, those with higher levels of education tend to be less poor than the average. In the aggregate, Kosovo’s extremely high unemployment rate is another important driver of its high poverty levels. Poverty rates are especially high among the unemployed and casual day laborers. Large households are very common in Kosovo. Often these households must rely on the income from only one or two household members who are employed, and their income is not adequate to provide for the entire household. More rapid employment growth is needed not only to help today’s adults, but also to accommodate the rapidly growing youth population that is entering the labor force, and will continue to do so for many years to come. References Ravallion, Martin. 1994. Poverty Comparisons. Fundamentals of Pure and Applied Economics Volume 56. Chur, Switzerland: Harwood Academic Publishers. Ravallion, Martin. 2001. ―Growth, inequality, and poverty: Looking beyond averages.‖ World Development 29(11): 1803–1815. Ravallion, Martin. 2003. ―Measuring aggregate welfare in developing countries: How well do national accounts and surveys agree?‖ Review of Economics and Statistics 85(3): 645–652. Statistical Office of Kosovo. 2010. Results of the Labour Force Survey 2009. Pristina: Statistical Office of Kosovo. UNDP (United Nations Development Programme. 2011. Kosovo Human Development Report 2010. Pristina: UNDP. World Bank. 2005. Kosovo: Poverty Assessment. Report No. 32378-XK. Washington DC: The World Bank. World Bank. 2007a. Kosovo Poverty Assessment. Volume I: Accelerating Inclusive Growth to Reduce Widespread Poverty. Report No. 39737-XK. Washington, DC: The World Bank. 17 World Bank. 2007b. Kosovo Poverty Assessment. Volume II: Estimating Trends from Non- comparable Data. Report No. 39737-XK. Washington, DC: The World Bank. World Bank. 2010. Kosovo Country Economic Memorandum. Unlocking Growth Potential: Strategies, Policies, Actions. Report No. 53185-XK. Washington, DC: The World Bank. 18 Annexes Table A1: Probability of being poor, 2009 Dependent variable: poor 0/1 Urban Rural Household size 0.042* -0.008 (0.016) (0.014) Household size squared -0.002** 0.001 (0.001) (0.001) Female-headed household 0.017 -0.024 (0.040) (0.054) Head's age 0.008 0.012*** (0.006) (0.006) Head's age squared -0.000 -0.000** (0.000) (0.000) Household composition (relative to share of children 0 –6 years old) Share of children 7-16 years 0.070 -0.050 (0.087) (0.122) Share of females 17-64 years -0.122 -0.204 (0.105) (0.146) Share of males 17-64 years -0.255** -0.330** (0.102) (0.136) Share of elderly (65+) -0.039 -0.286 (0.140) (0.186) Education of HH members (relative to no education or less than complete primary education) Share with complete primary -0.111 -0.158** (0.075) (0.062) Share with complete secondary -0.252* -0.192* (0.076) (0.065) Share with complete tertiary -0.596* -0.352** (0.102) (0.169) Main source of household income (relative to “other�) Main HH income: Public wages 0.147*** -0.013 (0.083) (0.112) Main HH income: Private wages 0.207* 0.023 (0.078) (0.116) Main HH income: Farming -0.139 -0.088 (0.130) (0.101) Main HH income: Per diem work 0.167** 0.057 (0.081) (0.103) Main HH income: Self-employment -0.070 -0.193** (0.086) (0.095) Main HH income: Pensions 0.291* 0.097 (0.066) (0.096) Main HH income: Remittances -0.075 -0.152*** (0.083) (0.092) Main HH income: Social assistance 0.352* 0.178*** (0.057) (0.096) Main activity of individual household members (relative to “other�) Share wage/salary workers -0.356* -0.356* (0.123) (0.123) Share unpaid family workers 0.063 0.013 (0.082) (0.104) Share retired/disabled -0.172 -0.142 (0.105) (0.143) Share unemployed 0.184** 0.121 (0.077) (0.075) Region (relative to Pristina region) Ferizaji 0.157* 0.348* 19 Dependent variable: poor 0/1 Urban Rural (0.057) (0.052) Gjakova 0.250* 0.265* (0.041) (0.054) Gjilani -0.119* 0.027 (0.041) (0.059) Mitrovica 0.020 0.150* (0.037) (0.055) Peja 0.094*** 0.229* (0.056) (0.051) Prizreni 0.052 0.102*** (0.058) (0.053) Number of observations 1,280 1,102 Notes: Dependent variable: 1 if poor, 0 if nonpoor. Coefficients are average marginal effects (margeff). Robust standard errors, clustered at PSU level, in parentheses. Significance levels: *** p<0.01, ** p<0.05, * p<0.10 Source: World Bank staff estimates from 2009 HBS. 20 Table A2: Consumption regressions, 2009 Dependent variable: natural log of consumption per adult equivalent Urban Rural Household size -0.104* -0.012 (0.019) (0.016) Household size squared 0.005* 0.000 (0.001) (0.001) Female-headed household -0.073 0.017 (0.048) (0.064) Head's age 0.007 -0.005 (0.006) (0.007) Head's age squared -0.000 0.000 (0.000) (0.000) Household composition (relative to share of children 0-6 years old) Share of children 7-16 years -0.019 -0.098 (0.102) (0.123) Share of females 17-64 years 0.058 0.174 (0.130) (0.138) Share of males 17-64 years 0.049 0.215 (0.111) (0.131) Share of elderly (65+) -0.109 0.395*** (0.199) (0.209) Education of HH members (relative to no education or less than complete primary education) Share with complete primary 0.083 0.193* (0.083) (0.072) Share with complete secondary 0.426* 0.310* (0.089) (0.083) Share with complete tertiary 0.865* 0.717* (0.107) (0.222) Main source of household income (relative to “other�) Main HH income: Public wages -0.314* -0.071 (0.094) (0.130) Main HH income: Private wages -0.315* -0.096 (0.089) (0.136) Main HH income: Farming 0.082 0.091 (0.138) (0.128) Main HH income: Per diem work -0.415* -0.077 (0.103) (0.130) Main HH income: Self-employment -0.043 0.199 (0.082) (0.133) Main HH income: Pensions -0.521* -0.148 (0.106) (0.129) Main HH income: Remittances -0.029 0.280** (0.100) (0.129) Main HH income: Social assistance -0.705* -0.315** (0.114) (0.129) Main activity of individual household members (relative to “other�) Share wage/salary workers 0.370* 0.633* (0.095) (0.117) Share unpaid family workers -0.208** -0.003 (0.097) (0.120) Share retired/disabled 0.164 0.035 (0.122) (0.150) Share unemployed -0.413* -0.262* (0.078) (0.078) Region (relative to Pristina region) Ferizaji -0.348* -0.452* (0.062) (0.061) Gjakova -0.423* -0.336* 21 Dependent variable: natural log of consumption per adult equivalent Urban Rural (0.056) (0.066) Gjilani 0.011 -0.147** (0.055) (0.068) Mitrovica -0.086 -0.198* (0.063) (0.066) Peja -0.217* -0.359* (0.057) (0.062) Prizreni -0.237* -0.300* (0.059) (0.056) Constant 6.956* 6.488* (0.246) (0.275) Number of observations 1,280 1,102 R2 0.474 0.321 Notes: Dependent variable: log consumption per adult equivalent. Robust standard errors, clustered at PSU level, in parentheses. Significance levels: *** p<0.01, ** p<0.05, * p<0.10 Source: World Bank staff estimates from 2009 HBS. 22 Statistical Office of Kosovo Enti i Statistikës të Kosovës Zavod za statistiku Kosova The World Bank Country Office in Kosovo Statistical Office of Kosovo Mujo Ulginaku 3 Rr. Zenel Salihu 4 10000 Prishtina, Kosovo 10000 Prishtina, Kosovo P: (381-38) 249-459 P: (381-38) 235-111 / 235-474 F: (381-38) 235-033 http://www.worldbank.org/kosovo/ http://www.ks-gov.net/ESK