Is there such a thing as migration of poverty in Albania

Transcription

1 Is there such a thing as migration of poverty in Albania Draft, please do not cite April 2008 Jessica Hagen-Zanker 1, Maastricht Graduate School of Governance Carlo Azzarri, World Bank 1 Corresponding author: jessica.hagenzanker@governance.unimaas.nl, Universiteit Maastricht, Maastricht Graduate School of Governance, P.O. Box 616, 6200 MD Maastricht, The Netherlands 1

2 Abstract While Albania is more known for its massive international migration, internal migration is also noteworthy and understudied. This paper consequently tries to fill a gap in the literature by assessing the impact of internal migration on poverty. Earlier studies on Albanian migration show that internal and international migration flows show distinctive patterns: internal migrants mainly come from the mountain and North-Eastern districts, migrate to the peripheries of big urban centres, and are generally poorer. We make use of detailed information on the residence duration of internal migrants, as well as local economic and social conditions they are exposed to, taking advantage of the 2005 Albania LSMS with a focus on the households oversampled in peri-urban Tirana. We assess whether welfare in terms of health, education, consumption, income, and inequality has changed after internal migration. Our descriptive statistics confirm earlier studies on peri-urban Tirana and show that migrant households live in poor dwelling conditions, work in irregular and unstable employment and have low levels of health and schooling. However, their total percapita income and consumption is higher than for rural households. Further analysis on the impact of internal migration will be based on econometric techniques. The propensity-score-matching (PSM) approach and the instrumental variable (IV) method will be used to compare rural migrant households in the destination communities with their similar counterparts in rural Albania who did not move, taking their pre-migration wealth into account. 2

3 1. Introduction While Albania is more known for its massive international migration flows, internal migration flows are also noteworthy and understudied. Earlier studies on Albanian migration have shown that internal and international migration flows show distinctive spatial patterns, both in terms of geography and poverty: internal migrants mainly come from the mountain and North-Eastern districts, migrate to the peripheries of big urban centres, and are generally poorer. This paper consequently tries to fill a gap in the literature by assessing the impact of internal migration on poverty. Albania was one of the most isolated and authoritarian communist countries. While international migration was almost completely stopped, internal migration was freely permitted until the 1960s when the Albanian government introduced a policy of rural retention and minimal urbanisation (Vullnetari, 2007). This meant that by the fall of communism in 1990, Albania was the most rural society of Europe (Cabiri, 2002). Albanians started migrating internally immediately after the fall of Communism, even though it was not legal to do so until Figure 1 shows the internal migration trends between 1990 and Figure 1. Internal migration flows number of individuals that moved internally 250, , , ,000 50, Source: Own calculations, based on 2005 LSMS household survey data Like international migration, internal migration peaked after the regime change in 1990 and after the collapse of the pyramid savings scheme in Between the two Censuses in 1989 and ,600 people (5.7% of the population) moved between regions and twice that many people moved between prefectures (INSTAT, 2 In several informal pyramid savings schemes collapsed that amounted to about 50% of GDP in nominal terms. About two thirds of Albanians had invested in them and the collapse had immense financial and political effects. Albania was close to a civil war and many people lost all their savings and property. 3

4 2004). Internal migration took place in different forms, but the most dramatic internal migration flows are rural to urban. A major consequence of internal migration is that urbanisation increased greatly, as figure 2 below shows. Figure 2: Urbanisation trends urban share of population in % Source: Cabiri (2002), 2009 is predicted Between rural population declined by 15%, while urban population increased by 14% (INSTAT, 2002). The share of urban population increased from 35% in 1980 to 45% in 2005 (INSTAT quoted in Cila, 2006). In 2001, 60% of internal migration was rural-urban, with almost half of the rural-rural flows going to the Tirana municipality (INSTAT, 2004). Internal migrants mainly come from the North-Eastern remote mountain districts and mountainous Librazhd and Skrapar districts and move to the coastal districts and Albania s capital, Tirana. Tirana is the most important destination in absolute and relative terms, with Durres in second place (Carletto et al 2004). Figure 3 below is a map of Albania that shows the major internal migration trends. The arrows are indicative of actual migration flows. 4

5 Figure 3. Main inter-district migration flows Source: Based on Zezza et al. (2005) The second major consequence of internal migration is the enormous growth of Tirana. Tirana is the financial, economic, cultural and educational centre of country and grew by more than 40% since The population increased from 368,000 to at least 600,000 in 2002 (INSTAT, 2004). Unofficial estimates place the current population at around 800,000 inhabitants (de Soto et al, 2002). More than 60% of population did not live there before 1989 (Zezza et al. 2005). Tirana experienced a building boom as a result of the population increase and 51% of the building stock was constructed after 1990 (Agorastakis et al., 2007). Growth is focused in peri-urban areas. The largest inflows of internal migrants were towards mini-municipalities 11 and 6, they doubled in population between the two censuses (Carletto et al 2004). One of the most important peri-urban settlements is Bathore, in the Kamza municipality. 5

6 Here the number of inhabitants increased from about 7000 in 1989 to 80,000 in 2005 (Cila, 2006). Peri-urban areas lie on the fringes of urban Tirana, often occupying former agricultural communes or abandoned public industrial areas. About 55% of the population in Greater Tirana lives in informal conditions. There are four different peri-urban areas in Tirana: Selita and Mihal Grameno in the Tirana municipality and Bathore (the largest peri-urban settlement) and Frutikultura in Kamza, which is part of Greater Tirana. Urbanisation took place illegally and informally with no urban planning. This means that most households do not have proper documentation for their houses and often lack connections to basic utilities (e.g. running water). Infrastructure was non-existent at first, but in the meantime international organisations or NGOs have made some investments in infrastructure, for example Bathore has a pioneer waster disposal system (Cila, 2006). Access to schooling is insufficient: Schools are far away and over-crowded, for example 2500 children are attending Bathore s only primary school which has a capacity of 1200 (Deda, 2007). Accordingly many schools have started holding multiple shifts of classes per day; students thus attend school for only a couple of hours each day (World Bank, 2003). Access to health services is also limited, as hospitals are only located in urban Tirana and health centres are running beyond capacity (ibid). Internal migration is often characterised by whole family relocation, which means that the share of women migrants is exceptionally large with 54% (INSTAT, 2004). Migrants are generally of working age and have low education levels. However, earlier studies on internal migration indicate that migrants come from all socioeconomic backgrounds (De Soto et al., 2002, Cila, 2006). The main motivation behind internal migration seems to be economic, for example work-related (Carletto et al. 2004). After being a very isolated and rigid central planning economy, Albania emerged as one of the poorest countries in Europe in Economic growth was high during transition (though marked by two recessions in 1991 and 1997), but the reduction in poverty has not matched economic growth (World Bank, 2003). Poverty in Albania has remained high with large spatial and regional differences: the rural and mountain areas are much poorer than the rest of the country, with the rural headcount in 2005 at 6

7 24.5% (down from 30% in 1994) and the urban headcount at 11% (down from 15% in 1996) (World Bank, 1997, 2006). Between poverty declined in all major areas of Albania. The most recent Albania Poverty Assessment of the World Bank shows a remarkable drop in poverty in rural northern areas, comparatively higher than in urban areas. The mountain areas experienced both a decrease in inequality and high growth, resulting in a large poverty decline (World Bank, 2006). Overall, transition resulted in even deeper regional inequalities with rural areas lacking behind in infrastructure investments and growth of employment possibilities. Therefore the poverty decrease in rural mountain areas comes unexpected and resulted in the hypothesis that poverty has spatially migrated from rural to urban areas. Zezza et al. (2005) investigate the link between migration in and outflows and poverty rates and find that unlike for international migration, poverty is a push factor for internal migration. Further evidence for the spatial migration of poverty hypothesis is that the incidence of poverty is lower amongst old residents than more recent residents and poverty is higher in areas where new residents live (Zezza et al. 2005). Our study will try to detangle the impact of two possible effects: on the one hand, rural-urban migration of individuals from poor households may have simply affected poverty indicators through its effect on household size; alternatively, it could have impacted regional poverty levels via the relocation of entire poor households which, despite the move, have remained in poverty. We will make use of detailed information on the duration of residence of internal migrants, as well as local economic and social condition they are exposed to, as drawn from the 2005 Albania Living Standards Measurement Survey. The analysis will be enriched by the focus on the oversampled households in peri-urban Tirana We will assess whether welfare conditions in terms of health, education, consumption and income changed after internal migration. The study apply econometric techniques, namely the propensity-score-matching approach and the instrumental variable method, to compare rural migrant households in the destination communities with their similar counterpart in rural Albania who did not move, taking their pre-migration asset endowments into account. 7

8 This study goes beyond previous studies on the wellbeing of internal migrants in periurban Tirana (Cila, 2006 and Deda, 2007) by directly comparing the wellbeing of migrant households to other groups in the population, in particular rural households. Through econometric techniques and by comparing migrant households to similar households who have not migrated we will show the impact internal migration has had on migrant households. We are able to do so since we have a fairly big sample size due to an oversample of peri-urban households in the 2005 ALSMS. The next section will review the theoretical and empirical literature on internal migration, with a focus on papers that explicitly discuss the impact of internal migration and previous papers on Albania. Section 3 will discuss the data and gives a descriptive overview on peri-urban migrants, while section 4 discusses the methodology and gives the empirical results. In the last section we conclude. 8

9 2. Literature review This paper falls in between two major strands of the literature. The majority of the literature on internal migration is on the determinants of migration. While the impact of internal migration has not been discussed much, there is a vast body of literature on the impact of international migration. This literature is not relevant in this context, because it focuses on the family left behind, whereas our focus is on the internal migrant households in the new destination. Below we outline some of the important theories on internal migration and highlight any predictions on the outcomes of internal migration. Where possible we illustrate the theories with empirical applications on Albania. Then we discuss the literature on the impact of internal migration and categorise the different effects. 2.1 Determinants of internal migration The neoclassical macro-economic migration theories explain migration as part of economic development. Internal migration occurs as a result of geographical differences in the supply and demand of labour, mostly between the rural traditional agricultural sector and the urban modern manufacturing sector. The basic dual economy model (Lewis, 1954 and Ranis & Fei, 1961) that grew out of trade theory, assumes perfect markets and a labour surplus in the traditional agricultural sector that is absorbed by the modern sector. The modern sector grows through capital accumulation and by poaching labour from the traditional sector, where there is surplus labour. Rural workers are attracted by the positive wage differential and migrate to the urban sector. In these models migration occurs until wage equalisation has occurred. Empirical models look at whether there is a rural surplus that can supply potential migrants. Cabiri et al. (2000) and INSTAT (2004) argue that internal migration in Albania takes place due to a surplus of rural labour which migrates to urban areas until equilibrium is reached. Todaro and Harris (Todaro, 1969 and Harris & Todaro, 1970) augmented the dual economy model to account of the significant urban unemployment that was found in many less developed countries. In the Harris & Todaro model rural-urban migration occurs, as long as the expected real income differential is positive. Expected income is a function of the rigid, institutionally determined urban wages and the urban employment rate. The employment rate is the probability of finding a job, i.e. being 9

10 selected from the pool of labour, which increases over time, for example due to wider networks of the migrants. Migration thus increases if urban wages increase or the urban employment rate increases (ceteris paribus). The authors show that it can be perfectly rational to migrate, despite urban employment, due to a positive expected income differential. Extensions of this model generally aim at making the labour market more realistic, for example by endogenising the urban wage. Isaah et al (2005) include urban infrastructure as an additional factor that positively affects utility of potential migrants (together with expected income). The Harris & Todaro model is usually tested with macro-level data. In this type of analysis inter-regional migration flows are compared to inter-region wage differentials and employment differentials. Cattaneo (2006) applies the model to Albania using household survey data and finds that migrants more likely to move when wages differentials are higher and unemployment differentials are lower. Zelinsky s functionalist hypothesis of mobility transition (1971) argues that migration is part of the economic and social changes inherent in the modernisation process. He hypothesizes that patterns and rates of migration can be closely linked to the stage of modernisation (e.g. industrialisation) and demographic factors (e.g. high birth rates). Zelinsky emphasizes that the preference for more personal freedom is part of the modernisations process. The theory of mobility transition broadly make sense when looking at past migration patterns in industrialised nations and could also be applied to transition in Albania. However, it is vague and does not allow differentiation of different types of migration. The human capital approach is the neoclassical micro-level migration theory. Based on the work of Sjaastad (1962) migration is treated as an individual investment decision to increase the productivity of human capital. Individuals make a rational cost-benefit calculation of the expected discounted returns of migration over future time periods, migrating only when the expected returns are positive. The expected returns depend on the expected benefits (discounted earnings at origin and destination and non-monetary returns) and migration costs (travel costs, opportunity costs and psychological costs). This model predicts that the young and educated migrate are the first to migrate. The human capital approach is interesting and useful in explaining the selectivity of migrants, but it is very hard to test empirically. It has not been 10

11 specifically applied to Albania, but previous studies show what while internal migrants in Albania are general young, they are not well educated (see section 3). The job search approach models migration and job search as joint decision (Harris & Sabot, 1982, Vishwanath, 1991). In these models migrant s decision to move depends on the migrant s job search mechanism (e.g. already look for job while still in rural area or move to a certain location in order to observe the different types of jobs offered). It thus goes beyond the Harris & Todaro assumption that rural migrants go uninformed to the cities and work in the informal sector while searching for a formal sector job and is also more specific than the human capital approach. This approach has not been applied in Albania, but Banerjee (1983) finds that in India most internal migrants already have a job in mind when they move, due to their urban contacts or due to information from previous migrants from the same community. 2.2 Effects of internal migration The effects of internal migration that have been studied can be classified into five categories that will all be discussed in turn. They are monetary returns of internal migration, impact on social mobility, impact on poverty, impact on labour-market success, impact on education and health. The neoclassical migration models predict that migrants move internally due to an expected income gain. Some empirical papers test the ex-post income gain by comparing the current urban income to the previous rural income or comparing income of migrants in urban areas to income of rural non migrants. Yap (1977) reviews a number of papers and finds that generally migrants experience income gains, even taking higher living expenses and in-kind rural income into account. Rye (2006) also finds that all rural-urban migrants in Norway benefit, even though to different extents (see below). Cila (2006) finds that income of peri-urban migrants in Tirana has increased compared to their pre-migration income. Tunali (2000) on the other hand shows that not everybody benefits from internal migration in Turkey in the 1960s and that a sizeable fraction of internal migrants experiences income losses. When rural-urban migration takes place it means that the reference group changes and social mobility may take place. Mohaditi (1986) measures income inequality as a result of internal migration in Iran and finds that rural inequalities are reproduced in the urban areas (using census data on a provincial level). If migrants come from a 11

12 landless family, urban inequality is increased. Rye (2006) looks at the effects of internal migration in Norway and shows that migrants from a higher social class benefit disproportionally more from internal migration and the intergenerational reproduction of social class is stronger in urban areas. The effect of internal migration on poverty has not been studied much empirically. Beegle et al. (2008) show using a unique tracking survey of individuals that internal migration is a powerful way of moving out of poverty, regardless of the destination. Migrants experienced 36% consumption growth compared to those that stayed behind. Kundu & Sarangi (2007) show with descriptive statistics for different districts in India that migrants are less poor than rural households but poorer than urban nonmigrant households. They also show that rural-urban migration has a strong and negative effect on the likelihood of being poor, even for casual workers. Cila (2006) uses the livelihoods framework to assess the change in wellbeing for rural migrants in Bathore. She finds that in general their livelihoods have improved. Income is higher, but more volatile and living expenses are also greater. Households are generally satisfied and their political, physical and social capital has improved. Furthermore their human capital in terms of number of years of education and awareness has increased (even though there are gender differences). Deda (2006) finds that peri-urban households are poor despite higher incomes because they need to pay for basic public services (e.g. connecting to electricity etc.). The dual-labour market theory (Piore, 1979) describes economic dualism between migrants and citizens in developed destination countries. There is a primary sector providing well-paid jobs and a secondary sector, which is where migrants work in unskilled jobs. The same occupational segregation could take place between rural migrants and urban workers, in addition to intra-occupational group discrimination, both of which could result in lower wages for rural workers. There is mixed evidence on the labour-market success of internal migrants. Knight, Song & Huaibin (1999) and Meng & Zhang (2001) show that for internal migrants in China there is labour market segregation and that they receive lower wages in the same occupation and with the same educational background. Davin (1999) confirms that the working conditions for migrant workers are tough and that they enjoy less economic and social security benefits than other workers. Yamauchi 12

13 (2004) also finds that returns to schooling are lower for Thai internal migrants than urban non-migrants, but that the accumulation of destination experiences raises migrant s wages. Banerjee (1983) finds that mobility between the informal and formal sector in India is low, but that within sectors education and urban experience of migrants and non-migrants is valued equally. In the studies that Yap (1977) reviewed migrants have a lower unemployment rate than non-migrants, their education levels are adequately valued and they are not disproportionally more likely to be in the informal sector. Labour market outcomes for Albanian internal migrants have not been studied extensively, but descriptive statistics reveal that most of the employed workers are employed in casual, informal and low-paid work (see section 3). Other effects of internal migration on the migrants or migrant households have not been studied much. The International Organization of Migration (2005) argues that migrants face greater health risks due to hazardous working and living conditions, peer pressure and easier access to drugs and the sex industry and because many migrants do not have access to the public health system. Shaokang et al. (2002) find that female internal migrants in Shanghai utilise antenatal care services to a much lower degree and show worse pregnancy outcomes compared to non-migrant women as a result of their illegal status and consequently higher costs for health services. Davin (2002) and confirm that in China temporary residents have inferior access to education and health services. Cila (2006) show that health status of peri-urban migrants is fine in general, but that there are some health problems due to bad sanitary conditions and that the health infrastructure is insufficient (travel distances to hospital are far and under-the-table payments are often required). However, she does not directly compare health status to the rural situation. Cila also finds that school enrolment in peri-urban areas has improved, but that there are still high drop-out rates amongst girls. The next section briefly discusses the data and compares peri-urban households to other households. 13

14 3. Data and Descriptive Statistics 3.1 Data We use the Living Standards Measurement Survey (LSMS) data for Albania that was collected by the Albania Institute of Statistics (INSTAT) with the technical assistance of the World Bank for the year of Data was collected on a community level, as well as on a household level, where detailed information on housing, health, migration, education, labour and consumption was compiled. A nationally representative sample of 3640 households was collected, supplemented with an additional 200 households oversampled in peri-urban Tirana in order to allow detailed analysis of peri-urban households. A brief comparison between rural and urban households is made in the next section. We compare households by migration status of the household head because in Albania internal migration is often commenced by an initial move of the household head. After the household head is settled, the rest of the family follows or the household head founds a family at the new location. We compare the households in peri-urban Tirana where the household head moved internally since 1990, to rural households where the household head did not move internally as well as to peri-urban households where the household also did not move internally post 1990s. 3 This will allow us to compare peri-urban households to two control groups when analysing the reasons for and impact of internal migration. Out of the 320 peri-urban households, 155 have a household head that did not move post 1990 and 165that have a household head that migrated after There are 1599 rural households where the household head did not migrate since The differences between these three groups of households will be further explored in section 3.3 and the peri-urban migrant group will be examined more closely in section 3.4. However, we first show some descriptive statistics for the whole sample. 3 We look at household heads that come from both urban and rural origins because this allows us to keep our sample as large and representative as possible. The differences between the groups are not large and they follow the same trend. 4 The following descriptive statistics are drawn on a slightly smaller sample, as overall 34 households show missing information for consumption and income. 14

15 3.2 Urban and rural households In this section we first give an overview of the differences between rural and urban households for the whole population, as a basis of comparison for our focus on periurban migrant households. Table 1 presents some general characteristics of urban and rural households Table 1. General characteristics Urban households Rural households Peri-urban households Whether head married 0.85*** 0.91*** Average age of members 34.32*** 32.30*** 29.57*** Number of children 0.86*** 1.20*** 1.30*** 1.05 Household size 3.83*** 4.46*** 4.74*** 4.19 Number of members living abroad *** 0.31*** 0.61 Whether single family house 0.49*** 0.94*** 0.85*** 0.73 Whether housing quality good 0.92*** 0.87*** 0.84*** 0.89 Whether running water 0.92*** 0.53*** 0.46*** 0.71 Whether someone has sudden illness 0.21*** 0.26*** 0.35*** 0.24 Distance to closest doctor in minutes 13.49*** 23.82*** Average years of education of adults 10.14*** 7.95*** 8.45*** 8.91 Unemployment ratio 0.15*** 0.04*** Monthly per capita income *** *** Monthly per capita consumption *** *** ** Poverty headcount 0.11*** 0.25*** 0.12** 0.18 Poverty gap 2.30*** 5.67*** 1.64*** 3.97 Poverty severity 0.80*** 1.89*** 0.34*** 1.32 Number of observations 1,938 1, ,806 Whether all children attend primary school *** 0.84** Whether all children attend secondary school *** 0.34*** Time to school in minutes *** ** Total Stars indicate whether the mean for each group is statistically different from the rest (* significant at 10%; ** significant at 5%; *** significant at 1%) 1 Number of observations are 800, 823, 156and 1779 respectively, as not all households have primaryschool age children 2 Number of observations are 608, 605, 110 and 1323 respectively, as not all households have secondary-school age children 3 Number of observations are 1114, 891, 190 and 2195 respectively, as not all households have school age children Source: Own calculations on ALSMS 2005, including oversample. Rural households have a greater incidence of married heads, are younger on average, are larger and with a higher number of dependant children. They are much more likely to have any kind of migration exposure, as well as a higher number of migrants 15

16 per household. Peri-urban households are in between urban and rural households for most categories. Table 1 shows that rural households have lower wellbeing in most categories. Urban households are less likely to live in a single family house, as apartment buildings are more common in urban areas. However, urban households live in accommodation that is of better quality and are significantly more likely to have running water. Peri-urban households have even lower achievements than rural households in terms of housing. Peri-urban household members are significantly more likely to have a sudden illness and rural households have to travel much further to the closest doctor. Rural households have much lower education levels, but nevertheless have a significantly lower unemployment rate, as more than 90% of households in rural areas are active in agriculture. Peri-urban adults are much closer in terms of education to rural adults. Urban households are much richer than rural households in terms of income and consume about 30% more. They get most of their income from wage and self employment, as opposed to rural households whose largest share is from agricultural activity, which does not result in high incomes. Peri-urban households are richer compared to rural households, but do not consume much more. 24% of rural individuals are poor, compared to 11% of urban individuals whose poverty is also less deep (measured by the poverty gap, see Foster, Greer & Thorbecke, 1984). Primary school attendance is significantly lower for rural and periurban households and the difference is enormous for secondary school attendance: Only 35% of rural households send all their children to secondary school. One reason for this might be worse infrastructure: Rural and especially peri-urban children have to travel further to school. 3.3 Rural and peri-urban households Now we explore some of the descriptive differences between the following three groups: Households in rural areas whose head did not migrate post 1990s (RNM), households in peri-urban Tirana whose head did not migrate post 1990 (PNM) and households in peri-urban Tirana whose household head migrated post 1990 (PM). Table 2 below shows some differences in the demographic composition of households. 16

17 Table 2. Demographics Rural Peri-urban RNM PNM PM Total Average age of members 32.63*** *** Age of head 52.76*** 49.48** 46.31*** Whether head married Ratio of males/females 1.21** 1.60*** Household size 4.47** ** 4.5 Number of children ** 1.19 Whether household has permanent migrant 0.37*** 0.16*** 0.26* 0.35 Whether household has temporary migrant Number of members living abroad 0.69*** 0.21*** 0.43** 0.65 Number of observations 1, ,803 Percentage of total households in ALSMS % 4% 4% 47% RNM = Rural household whose head did not migrate internally; PNM = Peri-urban household whose head did not migrate internally; PM = Peri-urban household whose head did migrate internally from rural area Stars indicate whether the mean for each group is statistically different from the rest (* significant at 10%; ** significant at 5%; *** significant at 1%) Source: Own calculations on ALSMS 2005, including oversample Individuals and household heads in PM households are significantly younger on average compared to the two non-migrating groups. This is consistent with the human capital approach to migration. There is no significant difference in terms of marital status. There are significantly fewer men per household in the RNM households compared to PNM households; it is often the women that are left behind as a result of international migration. Furthermore an internal move might lead to the extended family (e.g. elderly relatives which are more likely to be female) being left behind. It seems like families that moved internally mostly move as complete families, therefore there is no gender bias in favour of men. This also explains the significantly larger family size of PM households. The latter also have the highest number of children per household. RNM households are much more likely to have any type international migrant and have the most international migrants per household. It is likely that rural households have such a high rate of international migrants as they opted for an international, not internal migration strategy. PNM households are much more likely to have a migrant than PM households, so it seems that some households are more prone to any kind of migration. 17

18 Table 3 below shows the differences in education and employment for the different types of households. Table 3. Education and employment Rural Peri-urban RNM PNM PM Total Average years of education of adults 7.92*** 8.73*** # of education years of head 7.98* 8.79** Whether all working age members work 0.26*** *** 0.25 Whether head unemployed 0.01*** *** 0.02 Unemployment ratio 0.04*** *** 0.04 Average weekly hours worked by adults 25.17*** *** Number of observations 1, ,803 Percentage of total households in ALSMS % 4% 4% 47% RNM = Rural household whose head did not migrate internally; PNM = Peri-urban household whose head did not migrate internally; PM = Peri-urban household whose head did migrate internally from rural area Stars indicate whether the mean for each group is statistically different from the rest (* significant at 10%; ** significant at 5%; *** significant at 1%) Source: Own calculations on ALSMS 2005, including oversample Household heads and adults of households in PNM have significantly higher education levels compared to rural households. Adults and household heads in PM households have a lower (not significant) education level, but in terms of education they are more like RNM households than PNM households. Individuals in peri-urban households are much more likely to be unemployed and on average work fewer hours per week than rural individuals. Amongst the peri-urban households this is even more the case for the PM households, who have a significantly higher unemployment rate for household heads and individuals of working age in general. Peri-urban individuals mostly work in casual and informal employment and other surveys confirm that it is hard for rural migrants to find employment in Tirana (Cila, 2006). Another reasons for low employment ratios in PM households, could be low employment rates for women. Almost all women in rural households work in agriculture, whereas women coming from rural are to peri-urban areas are likely to be subject to traditional values and consequently become housewives and do not look for wage employment. Indeed, only one woman in our sample of PM household is working. 18

19 Table 4 and Figure 4 below show different indicators of wellbeing for the different households. we are able to compare some indicators to the 1990 situation, i.e. to the pre Transition and migration situation. Table 4. Wellbeing indicators Rural Peri-urban RNM PNM PM Total Monthly per capita income *** *** *** Monthly per capita consumption Morris score index *** 0.43*** 0.38*** Whether life improved over past three years 0.27*** 0.54*** 0.46*** 0.29 Change in self-declared economic status Worsened 0.14*** 0.07* 0.08* 0.13 Unchanged 0.57*** 0.73*** 0.69** 0.59 Improved 0.29** 0.20** Number of observations 1, ,803 Percentage of total households in ALSMS % 4% 4% 47% Peri-urban households have a significantly higher per-capita income than rural households, with PM households having the highest per capita income. Rural households have less than 75% of PM per capita income. This is in line with the neoclassical macro theory and means that migration has been beneficial in terms of monetary wellbeing. However, if you look at differences in consumption level, you get a different picture. PM households have lower consumption levels than the PNM group, as living expenses are high in the city (especially for new migrants whose status is illegal) and PM families do not have free food products from agriculture or nature anymore. The Morris score index is a asset score index index and it shows that in terms of asset ownership, PM migrants were not the poorest group in By 2005 the Morris score index was lowest for the PM group, but much higher for all groups. While PM households feel on average much more strongly about their life having improved in the past three years, their level of wellbeing has not reached the levels of PNM households. RNM households are the most pessimistic. Looking at self-declared financial status PM households do relatively well, especially compared to the RNM 5 The Morris score index is a weighted asset index that weighs each asset owned by the household by the reciprocal of the number of households that own the asset, see Morris et al,

20 group. Only 8% of households find that their status has worsened since 1990 and almost 70% find that their status is unchanged. Figure 4 below shows different poverty indicators. Figure 4. Poverty indicators RNM PNM PM Consumption PHC Income PHC RNM PNM PM Poverty gap Poverty severity RNM = Rural household whose head did not migrate internally; PNM = Peri-urban household whose head did not migrate internally; PM = Peri-urban household whose head did migrate internally from rural area Source: Own calculations on ALSMS 2005, including oversample Rural households have a higher risk of being poor if poverty is based on either income or consumption. The poverty rates based on consumption are much lower than other studies, for example Cila (2006) finds that 32% of migrant families in Bathore are poor. If we base the poverty head count on income, so disregarding consumption smoothing, we find much higher poverty levels and actually see that PM migrants are significantly less poor than PNM households. Furthermore RNM households are also significantly poorer when looking at higher order poverty measures, for example the mean poverty gap is 5.63 compared to a mean poverty gap of 1.10 for PNM households. This means that poor rural households are much further away from the poverty gap compared to peri-urban households. In conclusion, if income levels and poverty rates of RNM and PM households are compared, households have benefited from the move from rural areas with a monetary gain. Figure 5 below shows the participation of and the income shares of agricultural income, wage and self employment, transfers and other income generating activities for the different households. 20

21 Figure 5. Participation in income-generating activities and income shares RNM PNM PM Agricultural Transfer Non agricultural wage Other Non agricultural self-employment RNM PNM PM Agricultural Transfer Non agricultural wage Other Non agricultural self-employment RNM = Rural household whose head did not migrate internally; PNM = Peri-urban household whose head did not migrate internally; PM = Peri-urban household whose head did migrate internally from rural area Source: Own calculations on ALSMS 2005, including oversample As expected rural households have the highest share of income from agricultural income. 96% of rural households participate in agriculture. PNM households have a larger share of agricultural income compared to other peri-urban households and almost 80% participate in agriculture. It could be that those households have always been farmers since many peri-urban areas used to be agricultural communes or part of the agricultural university of Tirana. The share of income from both wage and selfemployment is the highest for PM households and the most significant share of income for all peri-urban households. Wage income is the most important activity for PM households. While self-employment is most common for PM households, as you would expect from migrant households, the difference is not significant. More than 50% of households in all categories receive some kind of transfer, either public or remittances. RNM households have the highest share of income from transfers and PM households a much lower share, despite the high share of international of migrants for the latter group. This could be related to the fact that rural households have a high share of international migrants (thus receiving remittances) and that many PM households are illegal, so probably not eligible for public transfers. Cila (2006) confirms that despite the high poverty rates amongst peri urban migrant households only 10% receive state economic assistance. Table 5 compares the differences in housing and infrastructure for the different households. Again we are able to compare some indicators to the 1990 situation. 21

22 Table 5. Housing outcomes Rural Peri-urban RNM PNM PM Total Whether single family house *** *** 0.84*** 0.86*** 0.93 Whether good quality house in * 0.87 Number of rooms per household member *** 0.43** 0.39*** *** 0.50*** 0.49*** 0.7 Whether running water * *** 0.52 Whether indoor toilet *** 0.61*** 0.39*** *** 0.84*** 0.79*** 0.52 Number of observations 1, ,867 Percentage of total households in ALSMS % 4% 4% 50% RNM = Rural household whose head did not migrate internally; PNM = Peri-urban household whose head did not migrate internally; PM = Peri-urban household whose head did migrate internally from rural area. Stars indicate whether the mean for each group is statistically different from the rest (* significant at 10%; ** significant at 5%; *** significant at 1%) Source: Own calculations on ALSMS 2005, including oversample The differences in housing between the different groups are interesting. Rural households have significantly better housing conditions than peri-urban households, especially when compared to PM households. RNM are the most likely to live in a high quality freestanding house and have a significantly larger number of rooms per person. In terms of house type, RNM and PM households lived in a very similar situation in 1990, both mainly lived in single family houses. PM are now slightly less likely to live in single family houses, since they moved to a peri-urban area, where apartment blocks are more likely than in rural areas. PM had the lowest number of rooms per household member in 1990 and they still do. Rural households gained the most space per household member, probably due to high migration flows from rural areas. All groups were unlikely to have running water in The situation improved for all groups except for PM households. PM households have the lowest rate of running water (only 33% have access to running water) and lowest rate of drinkable water. However, peri-urban households are now more likely to have an indoor toilet than in 1990, both compared to their own improvement and to the improvement of rural households. Differences in access to other utilities (e.g. 22

23 electricity) are not significant. Generally, in terms of housing and access to basic utilities, rural-to-peri-urban migration does not lead to an improvement. Table 5 below outlines health outcomes and infrastructure for the different households. Table 6. Health outcome and infrastructure Rural Peri-urban RNM PNM PM Total Whether someone in household has sudden illness 0.26*** ** 0.27 Whether health of household members good Distance to closest doctor in minutes 23.87** * Number of observations 1, ,867 Percentage of total households in ALSMS % 4% 4% 50% RNM = Rural household whose head did not migrate internally; PNM = Peri-urban household whose head did not migrate internally; PM = Peri-urban household whose head did migrate internally from rural area. Stars indicate whether the mean for each group is statistically different from the rest (* significant at 10%; ** significant at 5%; *** significant at 1%) Source: Own calculations on ALSMS 2005, including oversample In terms of health outcomes, not all differences between the groups are significant and there is no clear pattern. There are no significant differences in terms of self-declared health. There are significantly more PM household where a household member has a sudden illness. This could be related to the fact that only 70% of PM households have access to drinking water. Cila (2006) confirms that there are health problems in periurban Tirana due to bad sanitary conditions. In terms of health infrastructure Cila (2006) finds that peri-urban households have to travel far to secondary health care providers. Our data does not fully confirm this. We see that rural households actually travel significantly further to the closest doctor. Table 5 below outlines differences on schooling outcomes and expectations for the different groups. Table 7. Schooling outcomes and expectations Rural Peri-urban RNM PNM PM Total Whether all of household s children attend primary school *** 0.83 Whether all of household s children attend secondary 0.33** 0.57*** school 2 Average weekly hours worked 14.75** 8.09**

24 by teens of the household 3 Time to school in minutes by members * Whether members satisfied with school *** 0.88** 0.85* 0.74 Whether head considers education very important *** 0.90*** Number of observations Percentage of total households in ALSMS % 2% 2% 26% RNM = Rural household whose head did not migrate internally; PNM = Peri-urban household whose head did not migrate internally; PM = Peri-urban household whose head did migrate internally from rural area Stars indicate whether the mean for each group is statistically different from the rest (* significant at 10%; ** significant at 5%; *** significant at 1%) 1 Number of observations are 782, 74, 82 and 938 respectively as not all households have primaryschool age children 2 Number of observations are 586, 54, 56 and 696 respectively as not all households have secondaryschool age children 3 Number of observations are 572, 61, 57 and 690 respectively as not all households have teenage children 4 not all households have school age children or answered these questions Source: Own calculations on ALSMS 2005, including oversample The number of children of compulsory school age is highest for PM households as Table 2 showed. Despite the high number of school age children, primary school attendance is significantly lower for PM households. Secondary school attendance is equally low as for rural households, where one would expect low secondary school attendance. There are several reasons why school attendance could be lower for PM households. Maybe they cannot afford school and their children have to work. Teens in PM periurban households work significantly more hours than teens in PNM households. Rural teens work the most hours per week on average, as they are probably expected to help out in agriculture. Another reason could be that school infrastructure is not sufficient. Evidence for that is that children in PM households travel significantly further to school. However peri-urban households are much more satisfied with the quality of the education (apparently despite the capacity problems described earlier). Finally it could be that migrant households just do not see the value of education. This is unclear: PM households value education somewhat more than RNM households, but the difference is not significant. In the previous figures and tables we showed that PM households are worse off than the other two categories of households for most wellbeing dimensions and are only 24

25 better off in terms of income. We now look at the 1990 situation for a number of variables in order to see whether this change resulted from migration. Overall we conclude that migration has had a mostly negative impact on households that moved to peri-urban Tirana looking at the combined peri-urban migrant group. 3.4 Peri-urban migrant households The last section showed some puzzling and possible contradicting results for periurban migrant households. Migrant households do not seem to do well in many aspects of their lives, but migration flows continue, although at a lower level, as figure 1 showed. In this section we will analyse migrant households in peri-urban Tirana in more detail in order to explain these puzzling results. More specifically we will compare selected wellbeing outcomes for three different migrant cohorts, which are households arriving between , and The first group can be seen as the pioneers, the second group moved around the pyramid savings scheme crisis in 1997, which had a huge effect on many households and the last group can be considered the followers. Figure 6 shows the education level of household heads for these groups. Figure 6. Education level and age of the household head Average education years of hh head Moved Moved Moved Average age of hh head Moved Moved Moved Source: Own calculations on ALSMS 2005, including oversample Figure 6 clearly shows that there is a positive selection process in place among migrant households. The pioneer group consists of younger and more educated household heads, while the followers have much lower levels of education and are older on average. Figure 7 below, shows the unemployment ratios and hours worked for household members of the different groups. 25