THE CONTRIBUTION OF MIGRATION TO POVERTY REDUCTION IN RURAL HOUSEHOLDS IN KOSOVO

Transcription

1 THE CONTRIBUTION OF MIGRATION TO POVERTY REDUCTION IN RURAL HOUSEHOLDS IN KOSOVO JUDITH MÖLLERS +, WIEBKE MEYER +, GERTRUD BUCHENRIEDER ++, AND SHERIF XHEMA + + Leibniz Institute for Agricultural Development in Central and Eastern Europe (IAMO) ++ Universität der Bundeswehr München Contact details of corresponding author: moellers@iamo.de Theodor-Lieser-Str. 2, Halle (Saale), Germany phone: fax: Paper prepared for presentation at the World Bank International Conference on Poverty and Social Inclusion in the Western Balkans WBalkans 2010 Brussels, Belgium, December 14-15, 2010 Copyright 2010 by Moellers et al. This paper is based on a preliminary data set of the project Labour migration from Kosovo to Germany Motivation and Impacts on Remittee Farm Households. Please, do not cite from this paper without the permission of the authors. 0

2 Abstract Kosovo is one of the poorest countries in Europe. Poverty remains persistent and widespread: 45% of the population is living below the national poverty line, and an estimated 17% are extremely poor. Poverty is a strong reason for migration. Currently, between 15-23% of the Kosovar population have migrated, that is one adult in every third family. Subsequently, the wellbeing of an extraordinarily high number of families in depends heavily on their migrated family members. Currently, remittances represent around 11% of gross domestic product (GDP). This paper presents an impact analysis of migration on poverty reduction in rural households in Kosovo based on a 2009/10 survey. We draw on a unique sample of Kosovar migrants in Germany and their rural families in Kosovo as well as a matching sample of Kosovar families without migrants. This paper will use the migrant family and non-migrant family samples in Kosovo. First we determine which of the families are poor based on a 60% of median equivalised income at-the-riskof-poverty threshold as well as based on absolute poverty lines. This allows us to determine, which of the families with migrants would fall into poverty without remittances. In a next step we address the question how the rural income distribution is affected by mixed income structures and especially remittances? We calculate decomposed Gini-coefficients and Gini elasticities to analyse the marginal effect of remittances on inter-household income distribution. Finally, an econometric analysis of the determinants of the rural households wellbeing is applied. We are specifically interested which factors migration related and other factors, contribute to differentiating between income groups. The procedure applied is an ordered probit. Keywords: migration, Germany, rural Kosovo, remittances, poverty

3 1 INTRODUCTION 1 Labour migration is one of the most important livelihood strategies that rural households go for. In Kosovo a saying goes, a family needs one son for the family, who stays at home, one for migration, who works abroad, and one for the mother country. Indeed, the migration cum remittances livelihood strategy plays an outstanding role for Kosovo until today. Unlike migration from Albania, which was in the focus of research for quite some time, relatively little is known about Kosovo. Among the few recent studies available are UNDP-USAID (2010), Mustafa et al. (2007) and Haxhikadrija (2009); the World Bank (2007a) has published a Poverty Assessment based on Household Budget Survey data. Currently it is estimated that about 2.18 million people reside in Kosovo. Albanian people account for 92% of the population (Statistical Office of Kosova 2010a). One outstanding feature of the Albanian population is that about two thirds are younger than 30 years. The overall unemployment rate in 2009 was 40%, but it seems safe to assume that the real unemployment rate is even higher as particularly many rural job-seekers do not register as unemployed. An estimated number of 200,000 young people will enter the labour market within the coming five years. Thus the pressure to migrate will increase and indeed 50% of the young generation say they would emigrate if they could (Haxhikadrija 2009). Between 315,000 and 500,000 Kosovars live abroad, of which about 50% live in Germany. It is estimated that every third household in Kosovo has family members who live abroad (Mustafa et al. 2007, ESI 2006). It is well-known that Kosovo relies massively on capital inflows from abroad. A large share of this money comes from private remittances made by migrated family members. As considerable amounts of the money are transferred informally, there is hardly any official and reliable data on the remitted sums available (Möllers et al. 2010). However, it is believed that the sum of remittances represents 11% of the GDP. The UNDP-USAID (2010) give the absolute amount of remittances in 2009 as 443 million. There is not much doubt that remittances play an important role in overcoming poverty in Kosovo. New official statistics show that nearly one out of ten households in Kosovo indicate that remittances are even their main income source (Statistical Office of Kosova 2010b). The share of the population which receives 1 This paper is based on a recent dataset of the project Labour migration from Kosovo to Germany Motivation and Impacts on Remittee Farm Households (Buchenrieder and Möllers 2010). Some of the results are still preliminary. The data represent a sub-sample of the total survey because data cleaning is still ongoing. Please do not cite from this paper without the permission of the authors. 1

4 remittances (around 20%) is substantially higher than the fraction receiving social assistance (about 13%); the average amount of remittances is three times higher than average values from social protection programs in recipient households. There is also evidence that especially rural areas, where poverty is more pressing than in urban areas, enjoy higher average amounts of remittances (UNDP-USAID 2010, World Bank 2007a). This paper gives some new, empirically based insights on the effects of remittances on poverty and inequality by making use of a 2009/10 household survey conducted by the authors. We present results on the impact of remittances on poverty reduction in rural households based on poverty and inequality indicators as well as an analysis of welfare determinants by employing an ordered probit regression model. 2 RESEARCH OBJECTIVES AND METHODS This paper is based on a unique micro-database analysing migration-cum remittances strategies of farm households in Kosovo. The project takes a rural development perspective and is thus mainly interested in the remittance-sending behaviour and the impact of migration on the labour-sending households. Our sample is derived from the population of Kosovar migrants who currently live in Germany. The survey included only migrants who left Kosovo with an economic motivation. It therefore draws on a sub-group of the whole migrants population 2. An original random sample of 200 Albanian migrants from Kosovo has been interviewed by means of a structured questionnaire in 2009/10 in Germany. This survey was a first step of a two stage empirical approach, which first interviewed migrants and then in the second stage, the corresponding households in Kosovo. In the second stage, a matching non-migrant household sample was identified too. In summary, we can draw on primary data from three sub-groups: (1) rural emigrants from Kosovo in Germany, (2) their sending farm households in Kosovo, and (3) matching non-migrant households in Kosovo. This paper focuses on the data from the sub-groups (2) and (3). The results refer to a subsample of 170 rural households (with and without access to remittances) with complete income data. The econometric analysis is done for 134 households. The analysis focuses on the effect of remittances on poverty and inter-household income inequality. Poverty is measured by three standard poverty measures, (1) the headcount index, (2) the poverty deficit index, and 2 UNDP-USAID (2010) estimates the number of migrants from Kosovo who have left for economic reasons as 43%. Another big group (24%) left for political reason, 18% are refugees. 2

5 (3) the poverty severity index (FOSTER et al. 1984). The three poverty measures by Foster et al. (1984) are described by m 1 z c α = i P max, 0 n i= 1 z (1) ( ) α, where z is the poverty line, c i is the income of the individual i, n is the total number of individuals and m is the number of poor individuals. The parameter α changes depending on the poverty measure. If α is set equal to 0, we obtain P(0), that is, the headcount index indicating the share of poor below the poverty line. P(1) displays the poverty deficit, a measure that takes into account how far the poor, on average, fall below the poverty line. Finally, if α is set equal to 2, we obtain P(2), called the poverty severity measure, which captures the difference in the severity of poverty by giving more weight to the poorest. Thus, poverty severity considers income differences better (World Bank 2000, Coudouel et al. 2000). Poverty analyses often refer to adjusted household sizes which are used to calculate per capita incomes that consider economies of scale. Economies of scale arise in many ways in a family, for example by sharing certain expenditures such as housing or a car. There are different methods for estimating equivalence scales. Here we use figures that reflect the OECD equivalence scale. It assigns the coefficient 1 to adult household members, 0.5 to elderly adults in the household, and 0.3 to children under the age of 16. The effect of certain income sources on income distribution can be determined by Gini coefficients 3 in two ways (Reardon et al. 2000). The most common method is a comparison of a Gini coefficient for all incomes with another Gini coefficient that is calculated excluding the income source of interest (e.g. remittance income). If the latter is smaller (bigger) than the Gini based on total income, this income source has a negative (positive) effect on the income distribution. In addition, decomposition according to different income sources allows conclusions regarding the relative distribution effect of certain sources. The contribution of each income source is the product of a concentration coefficient for that income source and the fraction of that income source in total income (Shorrocks 1982, World Bank 2000). Formally Error! Bookmark not defined. G, the concentration coefficient for income component k, is given by * k 3 The Gini coefficient is the most widely used measure of income distribution. Its value varies between zero and one, with zero indicating a perfectly equal income distribution. The higher a Gini coefficient is, the more unequally the incomes are distributed. Gini coefficients between 0.25 and 0.35 are considered reasonable, while coefficients higher than 0.5 indicate that income distribution is seriously unbalanced (ELLIS 2000). 3

6 n n (3) G 2 * + 1 k = ri y 2 k, i μ n i= 1 2, where y k,i is the component k of the income of household i. The mean total income is denoted by μ; r i is the household s i rank in the ranking of all incomes. The Gini coefficient is a weighted sum of the concentration coefficients G *, K K μk * (4) G = Gk = SkG μ k= 1 k= 1 * k, where S k = μ k /μ is the share of component k in total income. The percentage contribution of income source k to total income equality is found to be * Gk (5) PK = Sk 100% G. The marginal contribution of each income source k to inequality can be described by an elasticity of the Gini coefficient, which is given by Lerman and Yitzhaki (1994) as (6) k S = * ( G G) k k G, S. G For the econometric analysis of welfare determinants we use an ordered probit. The ordered probit model differs from a univariate probit model in that the dependent variable is no longer a dummy variable, but an ordered variable taking values 1, 2, 3 according to the income of the household. Thresholds partition the households into groups of similar wealth corresponding to the three ordinal categories. Specifically, the total equivalised household income is considered in two different ways: first, including all sources of income, i.e. farm and non-farm income, social transfers and other income, and remittances, and second, the sum of income sources excluding remittances. An ordered probit model measures the probability that the dependent variable (Yi, for the i-th household) falls in one of the discrete categories conditioned on levels of the independent variables (Xj). Suppose the income of the sample households (y*) is the unobserved variable (latent continuous variable). y* is expressed in the following equation: 4

7 (7) y* = β x + ε where y* is the dependent variable (income tertiles), β is the vector of the estimated parameters and x is the vector of explanatory variables. ε is the error term. We observe that the thresholds for the tertiles are for equivalised income with remittances and without remittances y* = 1 if y < 1,304 y* = 1 if y < 630 = 2 if 1,304 <= y < 2,324 = 2 if 630 <= y < 1,311 = 3 if y >= 2,324 = 3 if y >= 1,311 Coefficients (β) of the ordered probit model give an indication of positive or negative impact of an independent variable on the probability of diversification, but do not relay information concerning the magnitude of the effect. Like the models of the binary data, we are concerned with how changes in the predictors translate into the probability of observing a particular ordinal outcome. Using a transformation function, the model creates a linear index of the probabilities with a cumulative standard normal distribution. Consider the probabilities of each ordinal outcome: For equivalised income with remittances: Pr(Y i = 0) = Pr(y* < μ 1 ) = Φ(1.304 β`x i ) Pr(Y i = 1) = Pr(μ 1 y* < μ 2 ) = Φ(2,324 β`x i ) - Φ(1,304 β`x i ) Pr(Y i = 2) = Pr(μ 2 y*) = 1- Φ(2,324 β`x i ) For equivalised income without remittances: Pr(Y i = 0) = Pr(y* < μ 1 ) = Φ(630 β`x i ) Pr(Y i = 1) = Pr(μ 1 y* < μ 2 ) = Φ(1,311 β`x i ) - Φ(630 β`x i ) Pr(Y i = 2) = Pr(μ 2 y*) = 1- Φ(1,311 β`x i ) where μ i represent the threshold or cut-off parameters for placement of y* in the discrete diversification categories, and Φ(.) is the standard normal cumulative distribution function such that the sum total of above probabilities is equal to one. 5

8 As the survey data used do not include longitudinal developments, but only a snapshot of the households, their income situation and remittances transfers, endogeneity may be a problem in the econometric analysis. Statements about causality are therefore not made. 3 POVERTY AND INCOME INEQUALITY IN RURAL KOSOVO HOW IMPORTANT ARE REMITTANCES? There are three (complementary) pathways out of rural poverty: (1) farming (intensification, specialisation), (2) non-farm labour, and (3) migration (World Bank 2007b). Furthermore, the level of social transfers clearly influences the vulnerability of rural households towards poverty risk. Our analysis of farm household data focuses on the role of migration and the contribution of remittances to household welfare and interhousehold income distribution in rural Kosovo. Among the most important effects of remittances that are discussed in the literature is their potential to significantly reduce acute poverty (Adams and Page 2005). A large and growing number of multidisciplinary micro-studies show that temporary migration helps to smooth seasonal income fluctuations, to provide extra cash to meet contingencies or to increase disposable income (Haberfeld et al. 1999; Rogaly et al. 2001; Mosse et al. 2002; Deshingkar and Start 2003; Deshingkar and Anderson 2004; International Organisation for Migration (IOM) 2005). One of the much debated issues is about the causality between migration and the wealth status of the labour sending household. On the one hand, people are in a position to migrate because they are better-off. On the other hand, migration improves the economic position of those who migrate, and possibly of the remittee household, and as a consequence increases inequality in the communities (de Haan 2000; de Haan and Rogaly 2002; Vathi and Black 2007; Daume et al. 2008; Sterbling 2008). Whether or not migration increases income inequality of the sending households due to the receipt of remittances is not yet fully clarified. Some argue that remittances contribute to a more equal income distribution (e.g. Stark et al. 1986, Stark 1988, Taylor and Wyatt 1996), others expect an increased inequality. Barham and Boucher (1998), who find depending on the chosen methodology - both directions, argue that inequality outcomes are sensitive to the choice of method and have therefore to be treated with care. Recent empirical studies that show that remittances usually reduce Gini coefficients and poverty risk are for example Gianetti et al. (2009) and 6

9 Acosta et al (2008). Gianetti et al. (2009) point out that the impact depends on the share of households receiving transfers, the average amounts received, and the distribution of remittances among the population. 3.1 WHAT DISTINGUISHES REMITTANCES RECIPIENTS FROM NON-RECIPIENTS? Typical farm households in Kosovo are relatively big. Our sample has an average household size of six persons. The dependency ratio is comparatively high indicating that there are 0.70 dependent household members per person of active age. Due to our interest in the impact of remittances and the survey structure, our sample is somewhat biased in the sense that it includes more households, 60%, with access to remittances (in 2009) than the overall population (around 20% according o UNDP-USAID 2010, see above). Table 1 shows that indeed, there are significant differences between households of these two groups. While the age of the household head is 53 years on average for both groups, the family size is smaller in households without remittances and the dependency ratio is also significantly smaller with only The educational level is surprisingly high in both groups. We look at the highest level of education in the household and find that in more than 80% of the sample households, there is at least one person who has finished higher education (meaning some type of university degree). All households have members with at least a secondary school degree or vocational education. Households without access to remittances seem to have a slightly better educational attainment with regard to higher education (Table 1). Contrary to our findings, national data imply that lower educated households are more dependent on remittances than others (Statistical Office of Kosova HBS 2009). Although Kosovo is a country with an extraordinary agro-ecological potential for agricultural activities, rural households rely to a relatively low degree on farm incomes. Only around 6% of Kosovar households indicate farm incomes as their major income source and the contribution to overall individual incomes is only 1% on the national level (Statistical Office of Kosova HBS 2009). Also for the rural sample analysed here the contribution of farm incomes to total incomes is rather low (see below). The average farm size of the sample households is only 2.3 hectares. Therefore it can be assumed that farming is mostly subsistence based. 4 These numbers will, however, be checked once more by the authors, to ensure that no systematic bias is causing this effect. 7

10 Indeed, the major share of households (62%) sell 10% or less of their farm produce. Only 7% sell more than 50% of their produce and could thus be termed market orientated farmers. In our sample, the average annual household income is around 12,500 (Table 1). Both household and per capita incomes are significantly higher for households who do not receive remittances. This could be an indicator that it is, at least for some part of the households, a voluntary choice to not engage in migration because the family income is high enough to sustain the livelihood of the household without remittances. However, the standard deviation of this variable is very high, indicating that income differences are high in this group. Per capita incomes are on average around 2,700. They are calculated based on the equivalised household size (see above). The low median in the group without remittances indicates once more that the income gap is bigger within this group. Table 1 Socio-economic characteristics of households with and without remittances, 2009 All households With remittances=1 Test statistics 1 0 χ 2 / p Number of HH HH size /0.000 Dependency ratio /0.000 Age of HH head /0.876 Number of migrants /0.000 Highest level of education in HH - Elementary or lower / Vocational education / Secondary school / Higher education /0.050 Farm land (ha) /0.663 Number of cattle /0.957 Share of subsistence > 50% /0.241 Household income ( ) 12,475 11,839 13, /0.001 PC income, equivalised ( ) 2,686 2,631 2, /0.000 Median of PC income, equivalised ( ) 1,799 2,021 1, Income shares (%) - Farm income / Non-farm income / Remittances / Other (unearned) income /0.955 Source: Note: Own calculation N=170 rural households; HH=household, PC = per capita Test statistics refer to a Kruskal-Wallis-Test. 8

11 Table 1 further shows that the income portfolios are mainly determined by non-farm incomes (44%) and remittances (36%). However, obviously, there are differences between the two groups, mainly caused by the availability or lack of remittances. In households with remittances, this income type makes up almost 50% of total income 5. Another big share is derived from local non-farm employment (34%). In our sample, households without remittances rely even more strongly on non-farm employment (68%), but have another relatively important income source in farming (18%). Farming plays generally a minor role; the share is even lower (11%) for households with remittances. Often, remittances have been brought forward as capital source for farm investments. If this has taken place in the household group with migrants, then it does not yet show. Both groups have relatively low income shares from other income sources, which are mainly pensions or social payments. On the national level, social welfare benefits account for 2% of individual incomes, pensions from Kosovo for 6% and pensions from abroad for 4% of the incomes (Statistical Office of Kosova HBS 2009). 3.2 POVERTY AND INEQUALITY IN KOSOVAR FARM HOUSEHOLDS Table 2 offers information on socio-economic characteristics in three income classes (tertiles), where the first tertile is the income class with the lowest per capita income and the third tertile is the richest group. While household sizes do not differ much, the dependency ratio increases with the income. This means that despite a larger number of dependents, the third tertile reaches higher per capita incomes. Another surprising although not significant result is related to the educational attainment: it seems higher in poorer households. The wealthier income classes have a higher share in secondary and vocational education. We believe that one explanation for this could lie in the use of remittances particularly for education purposes. This however, contrasts World Bank (2007a) results, which show no evidence that remittances are used to invest in education. 5 The national average of remittance income in total household income is given with 12% by the Statistical Office of Kosovo HBS (2009). For recipient households UNDP-USAID (2010) finds that remittances contribute approximately 40% to household income. 9

12 Table 1 Socio-economic characteristics according to income classes, 2009 All Income class (tertile) households Test statistics χ 2 / p Number of HH HH size ,448/0.799 Dependency ratio ,125/0.028 Highest level of education in HH - Elementary or lower ,000/ Vocational education ,199/ Secondary school ,920/ Higher education ,833/0.243 Farm land (ha) ,297/0.317 Household income ( ) 12, ,411/0.000 PC income, equivalised ( ) 2, ,224/0.000 Median of PC income, equivalised ( ) 1, Share in all household incomes (%) Income shares (%) - Farm income ,001/ Non-farm income ,747/ Remittances ,046/ Unearned income ,926/0.382 Share of HH without remittances Source: Note: Own calculation N=170 farm households; Tertile 1 = Income class with the lowest per capita income etc. The test statistics refer to a Kruskal-Wallis-Test. Income differences between the three groups are considerable. The income of the middle tertile is almost double that of the first, while the richest group has almost three times as much as the second and six times as much as the first. Differences in per capita incomes are even slightly higher. Indeed, the richest income group earns more than two thirds of all incomes, while the share of the poorest tertile in all household incomes is only 11%. This can be interpreted as a sign of the existence of income inequality. Poor households, according to the results in Table 2, rely slightly more on social transfer and farm incomes than wealthier households. Their share in farm incomes (19%) is around twice as high as in the other groups; unearned incomes have a share of 14% compared to only around 5% in the second and third tertile. Nonfarm incomes (38%) and remittances (29%), on the other hand, are still extremely important also for poor households, but the shares are lower than in the other tertiles. The better-off households rely mainly on local non-farm employment (53%). Remittances are found to be most important in the middle income group (42%). In this group the share of households without access to remittances is the lowest (18%), while the 10

13 poorest tertile has a higher share of non-migrant households. Thus, it seems that although the group of nonremittee households has on average similar incomes to the remittee households, remittances do play an important role in terms of poverty reduction. Table 3 presents three common poverty indicators based on three different poverty lines. We have chosen two absolute poverty lines, and a relative poverty line (Table 3). Clearly, poverty levels look different depending on the choice of the poverty line. The very low poverty line of 1.41 per day in 2002 prices that is used by the World Bank (2007a) seems from our point of view too low. In our sample, according to that poverty line, only 7% of the rural population would be considered poor. Compared with a rating done by our enumerators, this is indeed in line with their perception: they considered only 4.3% of the visited households as visibly poor. On the other hand, this poverty line is below the social assistance level that a person gets in Kosovo ( 40 per month according to the Statistical Office of Kosova (2010b)), an amount, which is, according to what we saw, hard to live of in the country. Furthermore it is also far below the current average expenditure that UNDP-USAID (2010) indicates with 410 per month. Therefore, we mainly refer to an absolute poverty line reflecting USD 4.30 PPP dollars or to the relative poverty line of 60% of the median of the equivalised per capita income. The latter should ideally be linked to the national median income. In our case, due to lack of data, we used the sample median income. This should be somewhat below the national median because we look only at rural households. The US$ 4.30 PPP per day is used by the World Bank, usually to reflect the different conditions in EU applicant countries. According to these two measures the poverty incidence is 25% or even 39% in the case of the US$ 4.30 PPP line. The poverty deficit, defined as the average distance of the poor to the relative poverty line, is 8% or 14%. The measure of poverty severity considers income differences by giving more weight to the poorest. In other words, greater weight is given to households that are further away from the poverty line. This indicator shows relatively low figures for the sample households, meaning that there the inequality in income distribution amongst the poor is fairly low. The impact of remittances on poverty is depicted in the last column of Table 3. Again, results depend a lot on the chosen poverty line. However, it gets clear that remittances are a decisive income source and that poverty levels would rise substantially if households were not supported by their migrated family members. Almost 30% would fall even below the 1.41 /day line. The relative indicator would raise to more than 50% and the 11

14 US$ 4.30 PPP line yields a headcount index of almost 70%. However, poverty reduction is generally overestimated here because absent migrants can be expected to earn some income if they had not migrated. 6 Table 3 Poverty in rural Kosovo (2009) Yearly income ( ) Headcount -index Poverty deficit Poverty severity Headcount index without considering remittances Absolute poverty line 4.30 PPP$ per day line 1, per day line in 2002 prices* Relative poverty line 60% of median** 1, Source: Own calculation. N=170 farm households from both subgroups with and without remittances *Absolute poverty line used by the WORLD BANK (2007a) for Kosovo on the basis of a cost of basic needs approach for **This poverty line corresponds to 60% of the median equivalised income within the sample. Finally, we raise the question, if remittances also play a role with regard to income inequality. Table 4 depicts Gini coefficients. The income distribution was calculated for total equivalised per capita incomes and for incomes excluding remittances. The national Gini coefficient for the year 2005 was about 0.30 (World Bank 2007a). Compared to this relatively modest Gini coefficient, the Gini coefficient of 0.48 for the sample indicates that income distribution is markedly more unequal in our sample. The Gini coefficient, which was calculated without considering remittances income displays a notable increase to This implies that remittances contribute to a more equal income distribution in rural areas. However, it should be noted that the number of remittances receiving households is higher in our sample than in the total population. This should increase the equalising effect that remittances have. The examination of partial coefficients calculated on the basis of decomposed Gini coefficients confirms the equalising effect (lower part of Table 4). It is local non-farm income that influences income distribution negatively, while international remittances have a positive effect on income equality. The elasticity for nonfarm income is positive (0.165) and negative for remittances (-0.102), farm incomes (-0.031) and other incomes (-0.032). A negative elasticity indicates a positive impact on income distribution, i.e., the Gini 6 Acosta et al (2008) confirm this on an empirical basis for a number of Latin American countries. They point out that when no imputations are made fort he income of remittance senders, countries where recipients are concentrated at the bottom of the distribution of non-remittance income exhibit much higher reductions in poverty headcounts attributable to remittances. Nonetheless, it is clear that effects of remittances are far from negligible. 12

15 coefficient will decrease when the respective income increases. In the sample, a 1% increase in remittances would lead to a decrease in income inequality of 10%. Table 4 Income distribution and remittances (2009) Gini coefficient on the basis of equivalised per capita incomes 0.48 remittances excluded 0.66 Decomposed Gini coefficients (elasticity in brackets) on the basis of farm incomes 0.32 (-0.031) on the basis of non-farm incomes 0.61 (0.165) on the basis of remittances 0.30 (-0.102) on the basis of other (unearned) incomes 0.11 (-0.032) Source: Own calculation. Note: N=170 farm households from both subgroups with and without remittances 3.3 DETERMINANTS OF RURAL INCOME: WHAT DIFFERENTIATES POOR FROM BETTER-OFF HOUSEHOLDS? This section presents the econometric results. The ordered probit regression models are specified with three income groups (tertiles) as dependent variable. We hypothesise that remittances receiving households are better-off than other rural households in Kosovo and that the amount of remittances received is one of the key variables that separates poorer from better-off households. However, as mentioned above, the direction of causality is not always clear. This loop of causality between the independent and dependent variables of a model could lead to an endogeneity problem. We therefore make only very cautious interpretation at this stage. The main aim of this econometric analysis is to show how a set of key variables is correlated to the well-being of farm households in Kosovo. We set our results in perspective with insights for Kosovo on the national level and studies in other countries as well as with a model that does not consider income from remittances in the tertile calculation. This serves as a first attempt to deal with the possible endogeneity problem. By comparing the results for tertiles with and without remittances as a source of income, the robustness of the results is confirmed. Beside remittances, there are a number of other factors that should influence the economic standing of a household within a rural society. Variables that are typically used in poverty analyses are either household specific (including the household structure and household related economic variables such as farm characteristics) or they represent individual characteristics including the education of household members or the age of the household head). 13

16 Both models presented in Table 5 and 6 have an overall good fit and a satisfactory coefficient of determination. Descriptives of the variables used in the model can be found in the Annex. The first interesting and surprising result is that key personal characteristics such as age and education turn out insignificant in the first model which includes all income sources in the tertile calculation. Other studies, such as the Kosovo Poverty Assessment by the World Bank (2007a) find that poor families are characterised by a low education level. UNDP-USAID (2010) find a slightly slower level of education in remittances receiving households. The figures provided by UNDP-USAID for the mean years of education in Kosovo as a whole (11.3 years) lie, not surprisingly, above the average of our farm households (9.6 years). In our analysis, the variability of education as shown in Table 2 is low. Nonetheless, the second model (Table 6) shows significant results. We can therefore conclude that education is more important for locally derived incomes than for remittances. The age of the household head does not seem to make any difference. Remittances went into the model with their absolute value. A dummy for remittances access or the share of remittances in household incomes yielded similar, significant results. Clearly, as already shown above, remittances play a major role with regard to household welfare. The second model shows that the higher the absolute amount of remittances is, the lower are total incomes from other sources. Thus, the households that tend to be well-off in the first model are wealthy exactly because they have access to this income source. Of course one could rightly argue that if their remittances stem from core household members that local incomes would rise if this labour force would contribute to local incomes. Nonetheless, there are results that confirm that this effect does not fully outweigh the income effect of remittances (see footnote 6). In the Kosovar context it is also doubtful if many migrants could be adequately employed in the local economy. Another income source, local rural non-farm employment, seems also decisive. The variable, a dummy that turns to one if the income share from non-farm sources is below 25%, is significant with reverse signs in both models. Thus, if remittances are considered, a large share of non-farm income leads to a higher probability to be part of the poorer income groups (probably because it means that the share in remittances is comparatively lower); if remittances are not considered, however, a high share of non-farm becomes a key determinant of the welfare level. 14

17 Table 5 Ordered probit regression on income tertiles including remittances (2009) Independent variables Coefficient Std. Err. z < P 95% Conf. Interval Farm size in ha ** Number of migrants in household Dependency ratio * Years of education (household head) Age of household head Age of household head squared Remittances received by household in *** Share of non-farm income < 25% ** Strength of migration related network Pseudo R² = LR chi2(9) = Prob > chi2 = Source: Own calculation. Note: Dependent variable: tertiles of total per capita equivalised annual income, N = 134 Table 6 Ordered probit regression on income tertiles excluding remittances (2009) Independent variables Coefficient Std. Err. z < P 95% Conf. Interval Farm size in ha *** Number of migrants in household Dependency ratio Years of education (household head) ** Age of household head Age of household head squared Remittances received by household in ** e-06 Share of non-farm income < 25% *** Strength of migration related network Pseudo R² = LR chi2(9) = Prob > chi2 = Source: Own calculation. Note: Dependent variable: tertiles of total per capita equivalised annual income without remittances N = 134 The household structure also seems to play a role. However, it is not easy to find a straightforward interpretation of the results. A high dependency ratio, meaning a higher share of dependent members in the family, is found in better-off households in the first model. This is rather counterintuitive and is also not in line with World Bank (2007a) results. However, in the traditional family structure of Kosovo the support of needy family members like children and elderly plays a major role (Meyer et al., 2010). Consequently, migrants whose parents, grandparents, children or nephews and nieces live at the origin feel stronger obliged to remit, leading to an increase in overall household income. This also explains why the dependency ratio 15

18 plays no role in the second model. The number of migrants in the household is not significant in both models. The same is true for the strength of the migration related network that a household has. As we are dealing with farm households, one would expect that the involvement in farming should be decisive for the welfare of households. However, as already shown above, the share of farm income in overall income is low and other income sources, notably remittances and non-farm income, seem to be much more important. Nonetheless, the farm size variable is significant in both models. The larger the arable land available for the household the higher is the probability that it is found in the higher income tertiles. This probably indicates that a certain economic asset base is important for paving the way out of poverty. In line with this, the World Bank (2007a) finds that the landless are especially affected by poverty. In summary, we find that access to specific income sources makes the differences between poor and betteroff households. Typical individual and household related characteristics play no major role (such as age and the number of migrant). Only education seems an important door-opener for rural non-farm employment. Based on these first results, we plan to extend and improve the model by linking it with data of the reference migrant. Furthermore we will include further variables describing the location of the farm and consider imputation of income for remittance senders. 4 CONCLUSIONS Is one son for migration really needed in rural Kosovo to be able to make ends meet? Are remittances helping to find the pathway out of poverty? We can not give a definite answer to these questions, because this would need longitudinal data and counterfactual scenarios. However, our results strongly indicate that indeed the migration-cum-remittances strategy is important in overcoming poverty in Kosovar farm households. First of all, we can show that the share of remittances in the income of recipient households is very high (almost 50%). In fact, remittances play a much bigger role in the income portfolio than for example social transfer payments. Second, a simple comparison of income with and without remittances shows that a large number of families are lifted above the poverty line due to their access to remittances. Third, we can confirm that remittances contribute to a more equal income distribution based on decomposed Gini indices. 16

19 The role of farming is worth discussing here. At the outset in Germany, we included only migrants in the survey that grew up in farm households. It turned out when visiting these farm households in Kosovo, that the farm sizes were small and their production mainly subsistence oriented. Subsequently, farming has a surprisingly low share in total incomes especially in remittances receiving households. Nonetheless, we find that the farm size has an impact on the welfare level. We assume that this is rather an indirect effect in which the farm constitutes an economic basis that facilitates also other economic activities. Education is certainly a key to household welfare and rural development in general. The educational level in the surveyed households seems, at least on paper, surprisingly high. However, we get no significant relationship between the educational level and the income group of the household. Only incomes which do not consider remittances seem to be influenced (positively) by education. We interpret this as an indication that migration related activities are less dependent on education than local employment. It seems plausible that given the high pressure of young people on the local labour market, especially local non-farm employment is more easily accessed by better educated persons. 17

20 REFERENCES Acosta, P., Calderón, C., Fajnzylber, P. & Lopez, H. (2008). What is the impact of international remittances on poverty and inequality in Latin America? World Development, 36, Adams, R. H. J. & Page, J. (2005). International migration, remittances and poverty in developing countries. In: MAIMBO, S. M. & RATHA, D. (eds.) Remittances - Development Impact and Future Prospects. Washington DC, USA: The World Bank. Barham, B. & Boucher, S. (1998). Migration remittances and inequality: Estimating the net effect of migration on income distribution. Journal of Development Economics, 55, Coudouel, A., Hentschel, J. S., and Wodon, Q. T. (2000). Poverty measurement and analysis, in: Poverty Reduction Strategy Sourcebook. World Bank, Washington, DC. Daume, H., Bauer, S. & Schüttler, K. (2008). Migration und nachhaltige Wirtschaftsentwicklung. Eschborn, D. De Haan, A. (2000). Migrants, livelihoods, and rights: The relevance of migration in development policies. Department for International Development (DFID). Social Development Working Paper No. 4. London, UK. De Haan, A. and Rogaly, B. (2002). Introduction: Migrant workers and their role in rural change. Journal of Development Studies, 38, Deshingkar, P. and Anderson, E. (2004). People on the move: New policy challenges for increasingly mobile populations. Overseas Development Institute (ODI). Natural Resource Perspectives No. 92. London, UK. Deshingkar, P. and Start, D. (2003). Seasonal migration for livelihoods, coping, accumulation and exclusion. Overseas Development Institute (ODI). ODI Working Paper No London, UK. ESI (2006). Cutting the lifeline: Migration, families and the future of Kosovo. European Stability Initiative (ESI). Berlin, D, and Istanbul, TR. Giannetti;M., D. Federici; and M. Raitano (2009). Migrant remittances and inequality in Central-Eastern Europe. International Review of Applied Economics, , 23 (3): Haberfeld, Y., Menaria, R. K., Sahoo, B. B. and Vyas, R. N. (1999). Seasonal migration of rural labour in India. Population Research and Policy Review, 18: Haxhikadrija, A. (2009). Diaspora as a driving force in the development of Kosovo: myth or reality? < Foster, J., Greer, J. and Thorbecke, E. (1984). A Class of Decomposable Poverty Measures. Econometrica 52 (3): IOM (2005). Internal migration and development: A global perspective. International Organisation for Migration (IOM). IOM Migration Research Series No. 19. Geneva, CH. Lerman, R. I. and Yitzhaki, S., Effect of marginal changes in income sources on US income inequality. Public Finance Quarterly 22 (4): Meyer, W., Traikova, D., Buchenrieder, G., Möllers, J., & Xhema, S. (2010). Remitting from Germany to Kosovo - An empirical study based on the theory of planned behaviour. Paper presented at the World Bank International Conference on Poverty and Social Inclusion in the Western Balkans", Brussels, BE. Möllers, J., Xhema, S., Meyer, W. and Buchenrieder, G A socio-economic picture of migrants from rural Kosovo in Germany. Contributed paper at the conference Activating the sources of economic growth in Kosovo, May 14, Pristina. Riinvest Institute and AAB- Riinvest University. Mosse, D., Gupta, S., Mehta, M., Shah, V. and J. Rees (2002). Brokered livelihoods: Debt, labour migration and development in tribal Western India. Journal of Development Studies, 38, Mustafa, M., Kotorri, M., Gashi, P., Gashi, A. and Demukaj, V. (2007). Diaspora and migration policies. In: Forum 2015, Prishtina, KS. Riinvest. 18

21 Reardon, T., Taylor, J.E., Stamoulis, K., Lanjouw, P. and Balisacan, A., Effects of nonfarm employment on rural income inequality in developing countries: An investment perspective. Journal of Agricultural Economics 51 (2): Rogaly, B., Biswas, J., Coppard, D., Rafique, A. & Rana, K. (2001). Seasonal migration, social change and migrants' rights: Lessons from West-Bengal. Economic and Political Weekly, 36, Shorrocks, A. F. (1995). Revisiting the Sen Poverty Index. Econometrica, 63, Stark, O., Taylor, J.E., Yitzhaki, S. (1986). Remittances and inequality. Economic Journal 96: Stark, Migration, remittances, and inequality: a sensitivity analysis using the extended Gini index. Journal of Development Economics 28: Statistical Office of Kosova (2010a) data archive, < Statistical Office of Kosova (2010b). Household Budget Survey Series 5: Social Statistics Sterbling, A Konturen eines europäischen Migrations- und Sozialraums in Südosteuropa. In: Berger, P. A. and Weiss, A. (eds.) Transnationalisierung sozialer Ungleichheit. Wiesbaden, D: Verlag für Sozialwissenschaften. Taylor, J. E. and Wyatt, T. J. (1996). The shadow value of migrant remittances, income and inequality in a household-farm economy. Journal of Development Studies, 32, UNDP-USAID (2010). Kosovo remittance study Prishtina, KS. Vathi, Z. & Black, R. (2007). Migration and poverty reduction in Kosovo. Development Research Center (DRC) of Migration, Globalization and Poverty DRC on Migration, Globalisation and Poverty Working Paper No. C 12. Brighton, UK. World Bank (2000). Making transition work for everyone: Poverty and inequality in Europe and Central Asia. World Bank, Washington, D.C. World Bank (2007a). Kosovo poverty assessment. Report No XK. World Bank (2007b). World Development Report 2008: Agriculture for Development. World Bank, Washington D.C. Annex Table: Descriptive statistics of variables in the ordered probit models Variables Mean Std.Dev. Min Max Dependent variables Tertiles of total per capita equivalised annual income including remittances (1=poor, 2=medium, 3=better-off) Tertiles of total per capita equivalised annual income excluding remittances (1=poor, 2=medium, 3=better-off) Independent variables Farm size in hectares Number of migrants in household Dependency ratio Years of education of household head Age of household head Age of household head squared Amount of remittances received by household in Network of household concerning migration (1=very small, =below average, 3=average, 4=above average, 5=very large) Household earns less that 25% of income from nonfarm sources (dummy variable, if yes=1) Source: Own calculations Note: N=134 19