and Health Transitions Research Unit (Agincourt), School of Public Health, University of the Witwatersrand, South Africa.

Children and Migration in South Africa: A case study from a rural, northeastern district (version 2) by Mark Collinson 1 23 March 2008 A scientific report for the Princeton University/ Rockefeller Foundation Initiative in Children and Migration Table of contents Introduction... 2 Aim of the project... 3 Study population... 4 Methods and data... 6 Measuring migration... 7 Categorizing children s exposure to migration... 8 Analytic plan... 9 Description of migration patterns... 10 Youth, migration and education -1 educational attainment... 15 Youth, migration and education-2 crèche attendance... 18 Migration and child mortality... 23 Conclusion... 27 Acknowledgements... 28 References... 29 1 Any reference to material contained in the report should acknowledge the MRC/Wits Rural Public Health and Health Transitions Research Unit (Agincourt), School of Public Health, University of the Witwatersrand, South Africa. 1

Introduction There are many unanswered questions about migration and its outcomes for children and youth in the developing world, some of which are hard to address due to the limitations of temporal sequencing in available data. The INDEPTH Network hopes to make a contribution in this regard by using demographic surveillance system (DSS) data to understand the dynamics of migration and its outcomes in developing countries [1]. The databases of INDEPTH are well equipped for this task due to the prospective and longitudinal tracking of population events in time, including migration. Demographic surveillance employs fieldwork and information system methodologies to monitor population dynamics in small areas in developing world settings [2, 3]. This project explores the potential of a demographic surveillance system in the Agincourt rural subdistrict population of South Africa to help understand the impact of migration on children and youth in developing world settings. Migration is a frequent strategy used to enhance income opportunities, but it has a range of other driving factors (causes) behind it and different types of migration may have different impacts. The impacts of migration on children and youth have not been comprehensively explored. Health outcomes for the household left behind have been reported as both positive and negative in different parts of the world. The situation is affected by multiple factors operating at different levels of analyses. These include structural factors, e.g. economic opportunities linked to labour market adjustment and affecting rural livelihoods [4-6]; patterns of health and disease, e.g. exposure to sexually transmitted diseases [7-9], non-communicable disease and mental stress [4, 5]; social and cultural norms, e.g. the practice of child fostering [10, 11], or the social ties between migrants and their families [12, 13]; gender norms, e.g. the changing role of women in society [14-16], and urbanisation, e.g. how far along the mobility transition is the society, in terms of migration patterns changes from more temporary to more permanent forms of migration [17]. The socio-political context plays an important role; and when examining the question in South Africa it is vital to consider the impact of Apartheid, and its demise, on mobility patterns and social structures [4, 18]. This history is dominated by the mining industry, the rapid industrialisation following the mineral discoveries of the late nineteenth century and the Apartheid-driven homeland system, which restructured the settlement patterns and livelihood strategies of the African population to provide necessary labour, while forcing unemployed family members to remain in densely settled, rural areas. The infamous and well documented Influx Control, Group Areas Acts, and the pass laws, exerted controls on migration patterns that altered population development in South Africa by creating an impermanence on the urbanisation process of the black population [19-21]. In urban areas, these laws resulted in chronic lack of urban planning and a diversion of urban settlement into sprawling peri-urban areas [19, 22]. In rural areas, people were forced to live in homeland areas, based on a system of ethnic homogeneity. Access to land was further restricted by a process of villagisation [23]. Ultimately, these forces resulted in a transition from an agrarian to a cash-based rural economy [21], but a 2

poor, rural economy that engendered continual labour migrations and large numbers of disunited households split into rural and urban components. Thus, migration has had a key role in the success of the South African economy, while placing an immense burden on the rural population. This double edge of economic success has been expressed as follows: The very process that guaranteed wealth in the economy (i.e. the migrant labour system) simultaneously produced poverty and patterns of unemployment that still hobble South Africa [24]. The last two decades have seen major changes in the arena of politics, with the repeal of the pass laws occurring in 1986, after political struggle was waged against them for over a century (Wilson 2001); the birth of democracy in 1994; and in macro-economics, the liberalisation of markets, which, together with an unstable gold price, have led to changing labour markets. All of these changes are likely to have had an impact on migration, and on the lives of children and youth in both urban and rural areas. Aim of the project The project aims to address the following questions using the Agincourt demographic surveillance system in the rural north east of South Africa. 1. What are the contours of the migration that affects children and youth? 2. How does the exposure to different forms of migration affect the educational attainment and survivorship of children, compared to children who live without these migration exposures? 3

Study population LIMPOPO PROVINCE MPUMALANGA PROVINCE Agincourt Health and Demographic Surveillance System study site Figure 1: Location of the study site, Bushbuckridge, South Africa The Agincourt Demographic Surveillance System field-site is located in the Agincourt sub-district of the Bushbuckridge region, of the Northern Province, about 500 kms northeast of Johannesburg. It lies adjacent to South Africa s north-eastern boundary with Mozambique. The field-site has twenty-one village communities and measures 400 km². The total surveillance population is 70000 people living in some 11500 households, with a population density of 175 persons per square kilometer. The field-site was selected with specific aims of the programme in view, namely, to determine the health status and its determinants in an area typical of South African rural society (some distance from a tar road or township settlement), and to address issues of decentralised health systems development, particularly at health centre, clinic and community levels (Tollman et al 1995). The geo-ecological zone is semi-arid savanna. Average annual rainfall ranges from 700mm in the western part of the site to 550mm in the eastern part, with some eighty percent falling in the summer months of November to March. Seasonal rainfall patterns are variable, however, which render the area vulnerable to drought. The area experiences 4

hot summer and mild winter months, with temperature range of 12-40 degrees Celsius in Summer and 5-27 degrees Celsius in Winter (Collinson, et al. 2000) Population structure Agincourt Study Site 2004 Age Groups 100-104 95-99 90-94 85-89 80-84 75-79 70-74 65-69 60-64 55-59 50-54 45-49 40-44 35-39 30-34 25-29 20-24 15-19 10-14 5-9 0-4 Male De Facto Male De Jure Female De facto Female De Jure 5,000 4,000 3,000 2,000 1,000 0 1,000 2,000 3,000 4,000 5,000 Population Figure 2: The age sex population structure of the Agincourt study site in 2004, showing the de facto population (permanently resident) and the de jure population (which includes the linked temporary migrants) [25] De jure population 69749 Temporary migration female 6253 De facto population 52984 Temporary migration male 10512 De jure population female 36145 Households 11737 De jure population male 33604 Table 1: Agincourt study population numbers in 2004 (Source: Agincourt DSS database) Household context Housing ranges from traditional mud huts to brick structures with tin or tiled roofs. Stands are generally too small to support subsistence agriculture. Water is collected manually by women or children from communal taps or shallow wells in riverbeds. It is transported in 25litre drums and usually carried on the head. Levels of household sanitation are poor, with pit toilets, of varying effectiveness, the norm. All roads are unpaved. Public transport is limited to privately owned mini-bus taxis. Electricity and telephone services are limited 5

Methods and data Overall study design of the demographic surveillance system A key characteristic of a Demographic Surveillance System is the continuous demographic monitoring of an entire geographically-defined population. In the case of Agincourt, this involves a multi-round, prospective community study, with systematic recording of all birth, death and migration events, covering the whole population of the Agincourt sub-district (see figure 1). [25] Mapping the study area In 1992, accurate and detailed village maps were hand-drawn by project staff. Each existing dwelling was represented and allocated its own identification number, allowing return visits to each household. Maps were updated regularly to incorporate new dwellings and other infrastructural changes. In 1997 village maps were digitised and geo-referenced through an innovative technique that made it possible to conduct queries at household level. In 2003, a full geographic information system (GIS) with geo-referencing of all households was introduced. Maps are updated and edited annually to take account of spatial changes. The coordinate data, aerial photographs and DSS household identifiers were developed into a GIS database that has strengthened spatial analyses and fieldwork management. Current uses include: navigation (to locate sampled households); fieldwork management (more efficient workload allocation through measuring distance, assessing mobility, and determining household size); and spatial analyses (to map clustering of events such as migration or deaths, patterns of natural resource use, distribution of risk factors or health outcomes, and uptake of interventions). Baseline census and update rounds The baseline census was conducted in 1992 during which each household was visited and information on every resident recorded. The primary tool of health and demographic surveillance in Agincourt is a rigorous annual update of household membership along with individual status variables (relation to household head, nationality, marital-, residence- and education status), followed by enquiry into vital events (pregnancy outcome, death, in- and out-migration) and full maternity histories in women 15-54 years. Enquiry into key variables relevant to each vital event is undertaken. Since the baseline census, twelve census updates have been conducted, with the last nine at strictly annual intervals (1999 to 2007). There are four data collection teams, each composed of five fieldworkers and a supervisor. Carrying a populated census form and vital event forms, members of each team visit allocated households within pre-assigned villages and interview the most senior responsible adult. Up to two revisits are carried out, following which a neighbour may serve as proxy informant for basic information; more detailed information on vital 6

events would be obtained the following year unless the family had permanently outmigrated. Additional data: Special modules and status observations Additional data is collected in the form of special census modules or status observations nested within each update. These provide limited information relevant to particular lines of investigation. Status observations involve an additional two or three questions on individuals or households. Examples include education status update (conducted every five years, with a special module conducted in 2006) and questions that screen for conditions of public health importance such as chronic cough (the basis for active pulmonary tuberculosis case-finding (1999)), and one-sided weakness (the basis for a stroke prevalence study (2001)). Special census modules, introduced from 2000, provide explanatory variables on topics pertinent to understanding and monitoring transitions and thus may be repeated to assess changes over time. Examples include a labour participation module (2000, 2004), a temporary migration module (2002), child social grant uptake (2002, 2005), health care utilisation (2003, 2006), food security (2004) and marital status (2005). A measure of household socio-economic status is gained through an asset survey conducted in each household every two years (2001, 2003, and 2005). Measuring migration Definition of a temporary migrant A temporary migrant is a household member who is away the majority of time, but retains a significant link to their base household. In analysis, a six month per year cut off point was chosen to differentiate temporary migrants from local residents. Thus, people who are referred to as temporary migrants were absent from the household for more than six months of the year preceding observation, but who considered the index household to be their home base. Definition of a permanent migrant The Agincourt definition of permanent migrant is a person who enters or leaves a household with a permanent intention of entering or leaving. This definition closely follows the classic definition that migrants are people who experience a change in residence. This includes people who leave the index household and establish a household or join a household elsewhere. A key feature is that the destination household becomes the new home base for the migrant. 7

Categorizing children s exposure to migration To examine the impact of migration on children a range of migration exposure categories were identified and isolated in the database. This provided the ability to explore how different migration exposures have different impacts on children s outcomes. The following categories were defined: 1. Mother was a temporary migrant This implies that a child s mother is absent from the rural household for a majority of the time, but retains a significant link to the household. Usually this migration is for work or work seeking (see table 2 for reasons given for different migration types). In the database, a temporary migrant woman is not removed from the household roster, but the residence status is updated on an annual basis. 2. Father was a temporary migrant As with the previous category this implies that a child s father was absent from the rural household for a majority of the time, but retained a significant link to the household, usually for work purposes or work seeking. 3. Mother was a permanent migrant This implies that the child s mother was a permanent migrant. Thus, the woman was either added or removed from the household roster. The reasons include union formation or dissolution, or the whole household moving to a better place (see table 2 for reasons for migration). This type of move may leave a child behind or the child may accompany the mother. The child s migration is categorized below. 4. Father was a permanent migrant This is the same as category 3, above, but for fathers. In this society only 45% of children s fathers are co-resident with their children. 5. Child migrated unaccompanied A child migrated but was not accompanied by either the mother or father. This indicates a foster situation whereby a child relocates for a variety of reasons, usually moving to stay with a relative. For younger children this is usually for child care in the event of both parents working and with older children for access to better schooling. This kind of migration can also occur if a parent dies. In the database it could either reflect a permanent or temporary move, but the key feature is that the child is unaccompanied by a parent. 6. Child migrated permanently, accompanied by a parent This is the child version of the exposure described in categories 3 and 4, whereby a child and parent move together permanently, due to union formation or dissolution or migration to a better dwelling. 7. Child accompanied a parent who was migrating temporarily 8

This usually means the parent is a temporary migrant and the child accompanies the parent to their place of work. 8. Child is second generation settled former refugee This category of exposure relates to the former refugee experience of the child s parents. It implies that at least one of the parents migrated from Mozambique at the time of Mozambican civil war in the late 1980 s. While some former refugees still reside in designated refugee villages, others are more assimilated into South African villages. 9. Maternal orphanhood This means the child mother died. It is included to contrast the case of a child s mother being absent due to migration. 10. Paternal orphanhood This means a child s father died. Since fathers are not as likely to reside with the child as mothers this may have less disruption to child care arrangements, but may have household income implications 11. Father does not co-reside with child A large proportion (~55%) of fathers are not co-resident with children, so this category is included to examine the impact of this state on child outcomes. It is intrinsically a migration exposure because the father has moved away at some point. It usually implies there is little contact through communication or support between the father and the household. If the father was a temporary migrant and remained linked to the household then he would fall into category 2 (father is a temporary migrant) and would not be included here. On the mother s side, exposure to a situation where the mother is alive, but neither co-resident nor temporary migrant is rare so it is not included as a specific category. In a foster situation where the child lives separately from the mother it is captured in category 5, i.e. child migrated unaccompanied. Analytic plan 1. Description of migration patterns Permanent and temporary migration patterns are explored giving the age-sex patterns of migration streams, trends and reasons for move. 2. Youth, migration and educational attainment. 2.1 Educational Attainment by age This section gives cross tabulations of Educational attainment at ages 7, 11 and 17 by categories of child migration exposure. T-tests are conducted to compare the average educational attainment of the exposed versus the unexposed group. 2.2 Crèche Attendance This section gives cross tabulations of formal pre-school (crèche) attendance by categories of child migration exposure. Chi-squared tests are conducted to 9

compare the proportion of crèche attendees in the exposed versus the unexposed group. 3. Migration and under-five child mortality. This section gives bivariate logistic regressions of child mortality by categories of child migration exposure. A discrete time event history analysis was conducted whereby the child s exposure time was divided into child person quarters starting at birth and ending at each three month interval after birth. Migration exposures were assessed during each child person quarter, as well as a dummy variable indicating whether or not the child died in that quarter. Description of migration patterns This section discriminates permanent and temporary migration and shows how they differ by age, sex and volume of migration. In Figures 3 to 6, the solid curves depict the prevalence of migration in the period 2000 to 2004. The first three profiles are types of permanent migration, namely external in-migration (permanent moves into the field-site from outside the site), external out-migration (permanent moves out of the site), and permanent migration within the site (also known as local mobility). The fourth category is the proportion of the population who were temporary migrants during the observation period. The dotted profile line is the earlier period, 1995 to 1999, and is further discussed in the section Migration trends below. The circular and permanent migration profiles are very different. External in-migration, presented in Figure 3, was highest for young female adults aged 15 to 30 at the level of 4.2 per cent per year. In contrast, after age 9, males never exceeded a rate of 2 per cent per year. Children under age 5 of both sexes showed a reasonably high rate of 3.5 per cent per year, which dropped off over the 5- to 14-year age group to around 2 per cent per year for girls and slightly lower for boys. Figure 4 shows a similar pattern for the out-migrants leaving the study population. The rates were higher for females aged 15 to 30 at a level of 5 per cent of the population per year. Male out-migration rates were again lower in this age group at around 2 per cent of the population per year. For children under age 5 the rates were identical for males and females at around 3 per cent per year. Figure 5 shows who is moving within the study area, i.e. local mobility. Local movers showed much higher rates than external migrants, with a prevalence of 6.5 per cent of 0- to 4-year-olds and 8 per cent of 20- to 24-year-old females in each year; there was a relative increase in the mobility of 25- to 35-year-old men. Figure 6 shows the age/sex prevalence of temporary migration, which is clearly different from the demographic profile for permanent migrants. Males, from the late teens onward, were very likely to be circular migrants, and this high prevalence (60 per cent probability) continued from ages 15 to 50. This adult temporary migration remained high 10

and did not drop below 20 per cent until men reached ages of 65 to 69 years. Young adult women were also engaged in circular migration, but, unlike the permanent migrants, the likelihood of migration remained around 25 per cent for a shorter age interval, from 20 to 44 years, and then gradually declined. Temporary migration rates were also higher among young children (around 7 per cent), compared to permanent migrants. Children of both sexes were involved in all types of moves. Around 20 per cent of children per year made a permanent or temporary move. They were also affected by the high levels of adult temporary migration. Migration trends Trends indicate how migration streams are changing over time, and can be referenced to a mobility transition linked to social transition occurring during the previous decade. The longitudinal data allowed a computation of period rates, whereby moves are pooled over the period 1995 to 1999, as a function of the population resident at that time, and then computed again for the period 2000 to 2004. The comparison of these two profiles shows the trend in migration rates by age and sex. The following four trends are notable for the 1995 to 2005 period: 1) Population mobility increased, with increases particularly apparent among children aged 0 to 4, young adult men aged 15 to 34 and adult women aged 15 to 49. 2) There was a shift to migration at a younger age for women. 3) Temporary migration age profiles changed for both sexes. Men aged 15 to 25 were over 20 per cent more likely to migrate temporarily in the later period compared with the earlier. Adults older than age 50 showed a slight decrease in temporary migration over the period, though the incidence of migration remained high. 4) Women aged 15 to 49 showed a strong increase in the likelihood of temporary migration. Within this trend, the proportion of younger adult women (15 to 29) was growing faster than the proportion of older women making such moves. 11

1995-1999 Fem ales 1995-1999 Males 2000-2004 Fem ales 2000-2004 Males 90.00 Prevalence of migration per 1000 person years 80.00 70.00 60.00 50.00 40.00 30.00 20.00 10.00 0.00 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 Age 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80+ Figure 3: The age-sex profile of permanent in-migration into the Agincourt field-site from outside the site. Prevalence of migration per 1000 person years 90.00 80.00 70.00 60.00 50.00 40.00 30.00 20.00 10.00 0.00 1995-1999 Fem ales 1995-1999 Males 2000-2004 Fem ales 2000-2004 Males 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 Age 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80+ Figure 4: The age-sex profile of permanent out-migration out of the Agincourt field site 12

90.00 1995-1999 Fem ales 1995-1999 Males 2000-2004 Fem ales 2000-2004 Males Prevalance of migration per 1000 person years 80.00 70.00 60.00 50.00 40.00 30.00 20.00 10.00 0.00 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80+ Age Figure 5: The age-sex profile of permanent migration within the Agincourt field site (local mobility) 1995-1999 Fem ales 1995-1999 Males 2000-2004 Fem ales 2000-2004 Males 0.900 Proportion of the population temporary migrant 0.800 0.700 0.600 0.500 0.400 0.300 0.200 0.100 0.000 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80+ Age Figure 6: The age-sex profile of temporary migration, i.e. migrants who circulated into and out of the Agincourt field site population. 13

Local Mobility In-migration Out-migration n in 2002 Column % n in 2002 Column % n in 2002 Temporary migration Column % n in 2002 Reason for move union formation or dissolution 581 17% 188 18% 228 18% 0 0% 997 6% to live with another 341 10% 206 20% 201 16% 766 6% 1514 8% new dwelling for household 569 17% 55 5% 224 17% 0 0% 848 5% Work 25 1% 43 4% 24 2% 8005 66% 8097 45% Looking for work 0 0% 0 0% 3 0% 812 7% 815 5% Health 2 0% 2 0% 0 0% 21 0% 25 0% school/study 12 0% 8 1% 6 0% 1446 12% 1472 8% child accompanies parent move 1293 38% 289 28% 414 32% 876 7% 2872 16% other/unknown 558 17% 249 24% 191 15% 210 2% 1208 7% Total 3381 100% 1040 100% 1291 100% 12136 Colum n % 100 % n in 2002 1784 8 Total Colum n % 100 % Table 2: Reasons given for migration in the different migration streams in one year only, namely 2002. It can be seem in Table 2 that the main reason for local mobility (within or between villages) is households moving to a new dwelling, with a large number of children making these moves accompanying the adults involved. Union formation or dissolution also produces a large proportion of local moves. There is virtually no permanent migration for work, schooling or health reasons. The reason for moving into a rural village from a more urbanized setting is primarily to stay with a member of the family network, or for union formation or dissolution, or a child accompanying an adult migration. The reasons for temporary migration on the other hand differ completely from permanent migration. This type of move is largely for employment, education or looking for work. A small proportion move to live with another family member, or are children accompanying a parent who is a temporary migrant. 14

Youth, migration and education -1 educational attainment This section examines the impact of migration on the average educational attainment at ages 7, 11 and 17 years. The next section looks at crèche attendance. Educational attainment is examined, by age, for each category of child migration exposure, and compared to the educational of children at the same age who were not exposed to that migration category. The relation of the observed migration exposure to the educational attainment outcome was that the current education status was measured in 2006 in a special census module and the exposure variables were computed for each child during the period 2001 to 2005. Thus, the exposure to the migration category was observed prior to the educational outcome in 2006. Table 3 shows that if a child s mother was a temporary migrant in the four years prior to measurement of educational status no impact was observed on educational attainment. Table 4 shows that if a child s father was a temporary migrant there was a positive impact on educational attainment in the older school going ages. This advantage was most likely due to the benefit of remittances, but also a positive selection whereby a labour migrant father has a better understanding of returns to education than a non labour-migrant father. Table 5 shows that if a child s mother was a permanent migrant during the four years prior to measurement there was a lower educational attainment at all ages of schooling, but especially the older school going ages. This effect may be due to the disruption caused by the migration. Table 6 shows that if a child s father was a permanent migrant during the four years prior to measurement there was no impact on educational attainment. Table 7 shows that if a child had migrated unaccompanied in the four years prior to measurement there was no impact on the 7 or 17 year olds, but the 11 year olds showed a modestly negative impact. This may have been due to the disruption created by the move but also weaker supervision in the non-parental household. Table 8 shows that if a child had migrated permanently accompanied by a parent in the four years prior to measurement there was a lower educational attainment by age, especially for the older children. This may have been due to the disruption created by the move. Table 9 shows there was no impact on educational attainment for children who accompanied a parent when they made a temporary move. Table 10 shows that older children of second generation settled former Mozambican refugees were strongly impacted by the migration. The cross-border migration occurred in the late 1980s therefore these are the children that experienced the forced migration and had a major disruption in access to schooling. 15

Table 11 shows that maternal orphanhood was negative for educational attainment, especially in eleven year old children, which may reflect the time required by young girls to undertake household chores in the absence of their mothers. Table 12 shows no impact of paternal orphanhood. Table 13 shows that if the father does not co-reside with child there is a lower educational attainment, especially in older children. By definition this implies that fathers are not contributing to the well-being of the family. Without a male breadwinner the mother may be forced to work which implies that older children have larger burdens of household chores that impact on their progress at school. Migration category: Mother was a temporary migrant 7 years old 11 years old 17 years old N avg.ed.attain. n avg.ed.attain. N avg.ed.attain. No 1144 1.74 1164 4.87 1224 10.00 Yes 370 1.71 407 4.85 391 9.99 1514 t = 0.6734 ns 1571 t = 0.2688 ns 1615 t = 0.0920 ns Table 3: Average educational attainment by age and whether or not the child s mother was a temporary migrant. Migration category: Father was a temporary migrant 7 years old 11 years old 17 years old N avg.ed.attain. n avg.ed.attain. n avg.ed.attain. No 941 1.72 977 4.80 1005 9.91 Yes 573 1.77 594 4.97 610 10.14 1514 T = -1.1957 ns 1571 t = -2.5340 ** 1615 t = -2.7918 *** Table 4: Average educational attainment by age and whether or not the child s father was a temporary migrant. Migration category: Mother was a permanent migrant 7 years old 11 years old 17 years old N avg.ed.attain. n avg.ed.attain. n avg.ed.attain. No 1150 1.76 1281 4.89 1450 10.05 Yes 364 1.65 290 4.75 165 9.52 1514 T = 2.4830 ** 1571 T = 1.7201 * 1615 t = 4.0764 *** Table 5: Average educational attainment by age and whether or not the child s mother was a permanent migrant. Migration category: Father was a permanent migrant 7 years old 11 years old 17 years old N avg.ed.attain. n avg.ed.attain. n avg.ed.attain. No 1388 1.74 1452 4.88 1549 10.01 Yes 126 1.67 119 4.76 66 9.70 1514 T = 1.0854 ns 1571 T = 0.9647 ns 1615 t = 1.5450 ns Table 6: Average educational attainment by age and whether or not the child s father was a permanent migrant. 16

Migration category: Child migrated unaccompanied 7 years old 11 years old 17 years old n avg.ed.attain. N avg.ed.attain. n avg.ed.attain. No 1288 1.75 1290 4.90 1251 10.01 Yes 226 1.68 281 4.70 364 9.96 1514 T = 1.2860 ns 1571 T = 2.3567 ** 1615 t = 0.4684 ns Table 7: Average educational attainment by age and whether or not the child migrated unaccompanied. Migration category: Child migrated permanently, accompanied by a parent 7 years old 11 years old 17 years old N avg.ed.attain. N avg.ed.attain. n avg.ed.attain. No 1176 1.76 1315 4.89 1476 10.04 Yes 338 1.66 256 4.73 139 9.55 1514 T = 2.2070 ** 1571 t = 1.8333 * 1615 t = 3.4598 *** Table 8: Average educational attainment by age and whether or not the child migrated permanently, accompanied by a parent. Migration category: Child accompanied a parent on a temporary migration 7 years old 11 years old 17 years old n avg.ed.attain. N avg.ed.attain. n avg.ed.attain. No 1342 1.74 1472 4.87 1563 9.99 Yes 172 1.72 99 4.82 52 10.17 1514 T = 0.3943 ns 1571 t = 0.3879 ns 1615 t = -0.8059 ns Table 9: Average educational attainment by age and whether or not the child accompanied a parent on a temporary migration. Migration category: Child a second generation settled former refugee 7 years old 11 years old 17 years old n avg.ed.attain. N avg.ed.attain. n avg.ed.attain. No 997 1.742227 972 4.900206 1147 10.1456 Yes 517 1.725338 599 4.814691 468 9.630342 1514 T = 0.4132 ns 1571 t = 1.2578 Ns 1615 T = 5.9038 *** Table 10: Average educational attainment by age and whether or not the child was a second generation settled former refugee. Maternal orphanhood 7 years old 11 years old 17 years old N avg.ed.attain. N avg.ed.attain. n avg.ed.attain. No 1358 1.74 1336 4.92 1322 10.03 Yes 156 1.71 235 4.59 293 9.86 1514 T = 0.4359 ns 1571 t = 3.5189 *** 1615 t = 1.5635 ns Table 11: Average educational attainment by age and whether or not the child s mother had died. Paternal orphanhood 7 years old 11 years old 17 years old n avg.ed.attain. N avg.ed.attain. n avg.ed.attain. No 1454 1.74 1500 4.86 1443 9.98 Yes 60 1.77 71 4.96 172 10.10 1514 t = -0.3166 ns 1571 t = -0.5937 ns 1615 t = -0.8849 ns 17

Table 12: Average educational attainment by age and whether or not the child s father had died. Father did not co-reside with child 7 years old 11 years old 17 years old n avg.ed.attain. N avg.ed.attain. n avg.ed.attain. No 818 1.75 869 4.99 992 10.12 Yes 696 1.72 702 4.71 623 9.79 1514 t = 0.6549 ns 1571 t = 4.2899 *** 1615 t = 4.0472 *** Table 13: Average educational attainment by age and whether the child s father was not co-resident. Youth, migration and education-2 Formal pre-school (crèche) attendance This section examines the impact of migration on crèche attendance. Attending crèche has two possible implications. Firstly, when children attend crèche it may imply that the mother is working. In this setting there is little subsistence or peasant farming conducted and women s employment usually means working on nearby game farms or commercial fruit farms, or else migrating temporarily to nearby towns or the city. The other implication is that children attending crèche may experience a benefit in terms of learning elementary skills that will later facilitate learning at primary school. The crèche teachers in this setting are seldom trained in foundational learning, but an advantage may still be created in terms of numeracy and language development. Crèche attendance is reported as a dummy variable and cross tabulated with each category of child migration exposure. As with educational attainment the current education status was measured in 2006 in a special census module and the exposure variables were computed for each child during the period 2001 to 2005. Thus, the exposure to the migration category was observed prior to the observation of crèche attendance. Table 14 shows that if a child s mother was a temporary migrant in the four years prior to observation there was no impact on the likelihood of crèche attendance. This is somewhat surprising and implies that children are looked after by relatives when their mothers are temporarily away. Table 15 shows that if a child s father was a temporary migrant there was less likelihood of crèche attendance. We have seen a higher likelihood of remittances which may imply that mothers are more likely to be home bound and caring for their children on a daily basis. Table 16 shows that if a child s mother was a permanent migrant during the four years prior to observation there was a higher likelihood of crèche attendance. This may be caused by a two factors. Firstly, the mother s permanent migration can imply that social 18

networks are disrupted and there are fewer options for child care with relatives. Secondly, women who migrate may be positively selected for education and see more benefit in crèche attendance. Table 17 shows that if a child s father was a permanent migrant during the four years prior to observation there was a positive impact on crèche attendance. As with mother s migration this may be influenced by positive selection and disruption in social networks. Table 18 shows that if a child had migrated unaccompanied in the four years prior to observation there was less likelihood of crèche attendance. This may imply that there is less investment in fostered children by the foster family. Table 19 shows that if a child had migrated permanently accompanied by a parent in the four years prior to observation there was a positive impact on crèche attendance. As with mother s migration this may be influenced by positive selection and disruption in social networks. Table 20 shows that if a child had accompanied a parent on a temporary migration in the four years prior to observation there was a positive impact on crèche attendance. As with mother s migration this may be influenced by positive selection and disruption in social networks. Table 21 shows that children of second generation settled former Mozambican refugees were no more or less likely to attend crèche. Table 22 shows that maternal orphanhood was negative for crèche attendance which may be due to poverty or less investment in the orphaned children. Table 23 shows that paternal orphanhood was negative for crèche attendance which may be due to poverty. Table 24 shows that if the father does not co-reside with child there is a higher likelihood of crèche attendance. This is somewhat surprising because these households don t have fathers contributing income and therefore may be worse off. However, this may imply that mothers are forced to work resulting in higher levels of crèche attendance. 19

Mother was a temporary migrant Child in crèche 0 1 Total 0 15,761 1,263 17,024 92.58 7.42 100 1 5,413 408 5,821 92.99 7.01 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 1.0747 ns Table 14: Cross tabulating whether the child was in crèche when his/her mother was a temporary migrant. Father was a temporary migrant Child in crèche 0 1 Total 0 13,336 1,123 14,459 92.23 7.77 100 1 7,838 548 8,386 93.47 6.53 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 11.8847 *** Table 15: Cross tabulating whether the child was in crèche when his/her father was a temporary migrant. Mother was a permanent migrant Child in crèche 0 1 Total 0 17,530 1,206 18,736 93.56 6.44 100 1 3,644 465 4,109 88.68 11.32 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 118.3673 *** Table 16: Cross tabulating whether the child was in crèche when his/her mother was a permanent migrant. Father was a permanent migrant Child in crèche 0 1 Total 0 19,812 1,510 21,322 92.92 7.08 100 1 1,362 161 1,523 89.43 10.57 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 25.5288 *** Table 17: Cross tabulating whether the child was in crèche when his/her father was a permanent migrant. 20

Child moved unaccompanied Child in crèche 0 1 Total 0 17,107 1,458 18,565 92.15 7.85 100 1 4,067 213 4,280 95.02 4.98 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 42.4606 *** Table 18: Cross tabulating whether the child was in crèche after he/ she had made a migration unaccompanied by a parent. Child moved permanently, accompanied Child in crèche 0 1 Total 0 17,943 1,261 19,204 93.43 6.57 100 1 3,231 410 3,641 88.74 11.26 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 99.487 *** Table 19: Cross tabulating whether the child was in crèche after he/ she had made a permanent migration, accompanied by a parent. Child accompanied a parent on a temporary migration Child in crèche 0 1 Total 0 19,665 1,469 21,134 93.05 6.95 100 1 1,509 202 1,711 88.19 11.81 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 55.0346 *** Table 20: Cross tabulating whether the child was in crèche after he/ she had accompanied a parent on a temporary migration. Child a second generation former Mozambican refugee Child in crèche 0 1 Total 0 13,949 1,111 15,060 92.62 7.38 100 1 7,225 560 7,785 92.81 7.19 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 0.2558 Ns Table 21: Cross tabulating whether the child was in crèche when the child was a second generation former Mozambican refugee. 21

Child in crèche 0 1 Total 0 18,001 1,538 19,539 Maternal 92.13 7.87 100 orphanhood 1 3,173 133 3,306 95.98 4.02 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 61.7715 *** Table 22: Cross tabulating whether the child was in crèche after his/her mother had died. Child in crèche 0 1 Total 0 19,855 1,631 21,486 Paternal 92.41 7.59 100 orphanhood 1 1,319 40 1,359 97.06 2.94 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 40.7243 *** Table 23: Cross tabulating whether the child was in crèche after his/her father had died. Father did not coreside with child Child in crèche 0 1 Total 0 11,704 776 12,480 93.78 6.22 100 1 9,470 895 10,365 91.37 8.63 100 Total 21,174 1,671 22,845 92.69 7.31 100 Chi Square: 48.787 *** Table 24: Cross tabulating whether the child was in crèche when the father did not co-reside with child. 22

Migration and child mortality This section examines the impact of migration on mortality of children under five years old. A child cohort was constructed from health and demographic surveillance system data for the observation period August 1997 to July 2005. A discrete time event history analysis was conducted whereby each child s exposure time was divided into units of three months (child person quarters) starting at birth and ending at each three month interval after birth. Migration exposures were assessed during each child person quarter, and dummy variables updated to indicate whether or not the child died before their fifth birthday. As the household dynamics unfolded for each child the health and demographic surveillance system recorded the status of who was present, who temporary migrant, and who had moved into or out the household. Bivariate logistic regressions were conducted for each migration exposure category testing whether the migration exposure influenced the odds of the child dying before they turned five. The unit of analysis was child person quarters which required an adjustment of the confidence intervals (using Stata s cluster option) to reflect that child person quarters were not independent for each child. Adjustment was made for child s sex and age, mother s age, household socio-economic status and time period. The household socio-economic status was constructed as an absolute index representing the privately owned assets at a household level. The asset status data were obtained from repeated cross-sectional surveys in 2001, 2003 and 2005. The findings presented are robust because these confounders were accounted for in the statistical model. Table 25 is discussed below row by row a. This row shows that if a child s mother was a temporary migrant after the child s birth but prior to the start of a child person quarter there was a significantly negative impact on child mortality. Although only 2.3% of child time had a temporary migrant mother, controlling for the other variables, this exposure had almost double the risk of mortality compared to the unexposed group. This is probably due to decreased parental support while the mother was away. b. This row shows that if a child s father was a temporary migrant after the child s birth but prior to the start of a child person quarter this was positive for child mortality. This is probably due to an improved financial situation due to remittances with consequent better access to health services for sick children. c. If a child s mother was a permanent migrant after the child s birth but prior to the start of a child person quarter this was negatively oriented to child mortality, but not significant. This was somewhat surprising because parental permanent migration had been showing positive outcomes. The situation was examined further with some additional cross-tabulations, given in rows (l) to (o), which explored different combinations of mother and child migrating. Line (l) assessed the likelihood of child mortality if the mother and child had both migrated after the child s birth, a situation experienced in 9% of child time. This was positively inclined, but not significant. In other words, compared to children who did not migrate with their mothers most children who migrated with their mothers were less likely to die. On the other hand row (m) showed that if the mother migrated and the child did not migrate then this was worse for child mortality, although it 23

did not quite reach significance. Row (n) examines the case when a child migrated and mother did not. This showed no relationship with child mortality. Row (o) examines the case when neither the mother nor the child migrated, which was not significant but inclined positively for child survival. d. Row (d) shows that if a child s father was a permanent migrant after the child s birth but prior to the start of a child person quarter this had a positive impact on child mortality (when controlling for the confounding variables). e. Row (e) shows that if a child had migrated unaccompanied after birth but prior to the start of a child person quarter this had a positive orientation for child survival but was not significant. f. Row (f) shows that if a child had migrated accompanied by a parent there was no relationship with child mortality. g. Row (g) shows that if a child was a second generation of self-settled Mozambican former-refugee then this was negatively oriented but not significant. h. Row (h) shows that maternal orphanhood was negative for child mortality, probably due to a disruption in the mother s care, and also possibly poverty. i. Row (i) shows that paternal orphanhood did not have an impact of child mortality. j. Row (j) shows that if a mother lived in a different household to the child there was a negative orientation but the result was not significant. k. Row (k) shows that if a father lived in a different household to the child there was a negative relation with child survival. This was case for more than half of child time and relates to the poverty experienced by households where the child and father did not co-reside, but the father was not a temporary migrant. 24

Table 25. Bivariate odds ratios for under-5 mortality in relation to parents and child s migration status, all with adjustment for sex, child s age, mother s age, household socio-economic status and time period, Agincourt DSS. Factor Level Person years (% of total) Odds Ratio (95% CI) a. Mother is a [ref] temporary migrant 0 65264 (97.7% ) 1 1550 (2.3% ) 1.97 (1.12-3.46)** b. Father is a temporary 0 55040 (82.4% ) [ref] migrant 1 11774 (17.6% ) 0.73 (0.55-0.97)** c. Mother is an in-migrant 0 46896 (70.2% ) [ref] 1 19918 (29.8% ) 1.15 (0.94-1.41) d. Father is an inmigrant 0 59744 (89.4% ) [ref] e. Child has migrated unaccompanied f. Child has migrated accompanied 1 7070 (10.6% ) 0.62 (0.45-0.85)*** 0 64755 (96.9% ) [ref] 1 2059 (3.1% ) 0.48 (0.16-1.47) 0 58528 (87.6% ) [ref] 1 8286 (12.4% ) 0.91 (0.63-1.33) g. Child is second 0 41422 (62% ) [ref] generation Mozambican 1 25392 (38% ) 1.17 (0.96-1.43) h. Maternal orphan 0 66399 (99.4% ) [ref] 1 415 (0.6% ) 3.54 (1.65-7.62)*** i. Paternal orphan 0 66168 (99% ) [ref] j. Mother lives in a different household k. Father lives in a different household 1 645 (1% ) 1.03 (0.35-2.99) 0 64968 (97.2% ) [ref] 1 1846 (2.8% ) 1.18 (0.46-3.04) 0 32408 (48.5% ) [ref] 1 34406 (51.5% ) 1.78 (1.44-2.21)*** 25

Combinations of mother and child migration status l. Mother and child have both migrated m. Mother has migrated and the child has not n. Child has migrated and the mother has not o. Neither the mother nor the child have migrated 0 60694 (90.8% ) 1 6120 (9.2% ) 0.89 (0.58-1.35) 0 53016 (79.3% ) 1 13798 (20.7% ) 1.2 (0.97-1.48) 0 64648 (96.8% ) 1 2166 (3.2% ) 1.01 (0.48-2.14) 0 22084 (33.1% ) 1 44730 (67% ) 0.87 (0.71-1.07) 26

Conclusion The demographic surveillance system afforded two advantages in assessing the impact of migration on child and youth outcomes. Firstly, due to the prospective follow up of the whole population it was possible to examine a range of different migration exposures that children and youth experienced. This proved useful because the outcomes varied by type of migration exposure, which helped us refine our understanding of the impact of migration. Secondly, the temporal nature of the surveillance data was exploited to ensure that the migration exposure occurred earlier in time and the educational or mortality outcome was measured subsequent to it. With the education status variables this was done by taking a cross sectional measure (in 2006) and examining the migration exposures in the period prior to this measurement, namely 2001 to 2005. The mortality data was more longitudinal in nature and enabled a discrete time event history analysis whereby the migration exposures could be assessed prior to each child s death, no matter when the death occurred over the observation period. Examining different outcome measures enabled us to observe that the impacts of migration are not uniform at all ages. Also, some exposures may be good for mortality but poor for education. The outcomes examined included mortality of children under age five years, pre-school (crèche) attendance for children aged four to six years, and educational attainment by age, at ages 7, 11 and 17 years. As expected, migration was sometime positive and sometimes negative for children. A mother s temporary migration had little impact on education but, when controlling for confounding variables, like household socio-economic status, had a negative impact on child mortality. When parents conducted permanent migration this was generally positive for crèche and school outcomes, but negatively oriented for child mortality when the mother migrated and the child did not. The situation of a child and mother living apart is complex and needs a more detailed analysis to unravel the relationships. A child migrating unaccompanied by a parent was generally negative for educational outcomes, especially the likelihood of attending crèche, but showed a mixed impact on child mortality. Children s migration when accompanied by parents was generally positive and most likely influenced by a positive selection, whereby families with the means to migrate are the same families with the resources and values that promote better education and health outcomes. Second generation Mozambican immigrant children tended to live in the poorer quintiles of the population, but, when controlling for socio-economic status, a child born to parents who were self settled refugees was negative for child mortality, but not quite significant in the bivariate analysis. On the education front the effect was significantly worse for older youth, but younger children fared better. The younger the child the closer the 27