Orphanhood , Vulnerability and Primary School Attendance : Evidence from a School-Based Survey in Two Regions of Tanzania By

This paper addresses two frequent overgeneralisations in the orphanhood literature in Africa: about the 'vulnerability' of children and about 'orphans'. It specifically examines school attendance, given the common presumption that orphans are less likely to attend school than non-orphans. Using survey data from two regions in Tanzania, analysis by primary school attendance categories (regular attenders, irregular attenders, dropouts, never attenders) shows that orphans should not be compared only with non-orphans since there are other vulnerable groups of children, all with different levels of social and spatial disadvantage. Both orphans and a second large and potentially vulnerable group of children, children who have not lost a parent, but who live with only one or neither of their parents, are less likely than other children to attend school in urban and roadside settlements, but there is no clear relationship for rural areas between vulnerable groups and attendance and dropout.


Introduction
Among the wide range of studies of the social impacts of HIV/AIDS, substantial attention has been drawn to the difficulties that orphans, and specifically AIDS-orphans, face.Many of these studies examine the effect of orphanhood on education and schooling.Most argue that c h i l d r e n i n H I V / A I D S -a f f e c t e d households are more likely to drop out of school temporarily or permanently than other children, and for a range of economic and social reasons associated with growing or sudden poverty and changing relationships within the household.H o w e v e r , o t h e r s t u d i e s h a v e demonstrated a mix of negative and positive associations between orphanhood and school enrolment for different age and sex groups, that orphanhood is not necessarily associated with lower school enrolment, and that the relationship between these two variables is far from straightforward.Thus paper seeks, firstly, to clarify the nature and extent of the controversy and how and why is has arisen, and, secondly, to shed some further light on the issues raised by the literature review in a study of primary school enrolments in two regions of Tanzania which compares the primary school attendance status of not only orphans, but also non-orphans and children in other vulnerable categories, and notably children who are not living in the same households as one or other of their parents, though they are both still alive.Below, first, there is a literature review identifying some of these major controversies and lacuna in our knowledge of the relationships between orphanhood and primary school entrolments in Sub-Saharan Africa.The methods and the assumptions of the survey of school attendance patterns in Iringa and Dodoma R e g i o n s , T a n z a n i a , w i l l t h e n b e introduced, and the findings on the relationship between orphanhood and school attendance in these two regions will be explored, with particular reference to orphans and children from spatially separated families and the geographical location of the households.It will be demonstrated that both groups of children who do not live with both their parents (i.e.orphans and children from separated families) are similarly disadvantaged in urban and roadside settlements, but hardly so in rural areas.The final sections examine a range of explanations of why this should be the case.

Literature review
There is a substantial literature on the problems of orphans in contemporary Sub-Saharan Africa (e.g.Foster and Williamson, 2000;Guest, 2001;UNAIDS, Unicef and USAID, 2002;Unicef, 1999a).This literature includes the effects of orphanhood and HIV/AIDS on education and schooling, generally arguing that c h i l d r e n i n H I V / A I D S a f f e c t e d households are disproportionately likely to drop out of school temporarily or permanently for one or more of the following reasons: having to care for sick relatives (Kelly, 2000;Robson et al., 2006), being stigmatised and bullied at school (Ndamugoba et al., 2000;Unicef, 1999a), and having an increased domestic and productive work load to compensate for lost family labour (Kelly, 2000).In addition, the household may suffer from diminished resources and hence no longer be able to afford school related expenses (Cohen, 1999;Kelly, 2000;Mukyogo and Williams, 1991;Unicef, 1999a).A few studies have provided empirical support for these widely held arguments about the negative effects of orphanhood on schooling: e.g.Bicego et al. (2003) in five East and West African countries, Lloyd and Blanc (1996) in Kenya and Malawi, Mueller and Abbas (1990) in Uganda, Mukyogo and Williams (1991) in Tanzania, Ndamugoba et al. (2000) in Tanzania and Sengendo and Nambi (1997) in Uganda.
What is less well-known, and in many respects seems counter-intuitive, is that there are also many studies of the relationship between orphanhood and education that have found either a mix of negative and positive associations for different age and sex groups (e.g.Ainsworth and Filmer, 2002 for Tanzania;Lloyd and Blanc, 1996 for seven sub-Saharan African countries; Urassa et al., 1997 for Tanzania), no association at all (e.g.Ainsworth and Filmer, 2002 in Chad and South Africa;Boler, 2007 for South Africa : Katabaro, 1999 for Tanzania), or even a positive relationship, i.e. orphans being more likely to attend school (e.g.Ainsworth and Filmer, 2002, for Nigeria and Tanzania;Lloyd and Blanc, 1996, for Namibia and Tanzania).
T h e s e apparently anomalous findings have been generally explained by the continuing but differential strength and ability of the extended family to absorb children whose parents have died.Urassa et al. (1997) remind us that adult mortality levels were high in Africa before the HIV/AIDS crisis, and child-fostering was common practice (whether the parents were alive or not).The greatly increased likelihood of there being o r p h a n s i n H I V / A I D S -a f f e c t e d populations has certainly expanded the numbers and proportions of orphans and fostering (Madhavan, 2004), but have they generated new forms of coping and caring for children or necessarily affected established forms and processes of vulnerability and marginalisation of children?
There are two key weaknesses in the orphanhood literature discussed a b o v e : o v e r -g e n e r a l i s i n g a b o u t 'vulnerability' of children, and overgeneralising about 'orphans'.In the first place, studies have tended to compare the well-being of orphans only with that of non-orphans (e.g.Bicego et al., 2003;Katabaro, 1999;Lloyd and Blanc, 1996;Makame et al., 2002).This is ethically problematic, since the underlying assumption that orphans are the most vulnerable children disregards other c h i l d r e n w h o m a y b e e q u a l l y disadvantaged.However, studies have now started to redress this weakness by also identifying vulnerable non-orphans, and placing them together under the category of 'orphans and vulnerable children' (OVC) (e.g.Ainsworth and Filmer, 2002;Foster, 2002;Unicef, 1999b;Smart, 2002;Urassa et al., 1997).For instance, Urassa et al. (1997) draw attention to the large number of poor children who are not orphans: in their analysis of the 1994-6 Kisesa Community Study (Mwanza Region of Tanzania), they find that amongst the 3253 children who lived in households classified as very poor, only 10 per cent were orphans.Moreover, in an analysis of the relationship between orphanhood and education, mainly through Demographic and Health Survey (DHS) data of twenty-eight countries from around the world, Ainsworth and Filmer (2002) find that poverty is a stronger determinant of a failure to go to school than orphanhood.In many countries orphans do not have a worse attendance record, and even when they do, 'in the majority of cases the size of the orphan enrolment gap is dwarfed by the gap in enrolment between children at the bottom and the top of the income distribution' (Ainsworth and Filmer, 2002: p.3).
In addition to poor children, a second group of children to which attention has been drawn as deserving equal concern to orphans are children whose parents are alive but who are living with only one or neither of them (Unicef, 1999b, andUrassa et al., 1997 -both for Tanzania).Unicef (1999b: 22) calls these ' s o c i a l o r p h a n s , ' b e l i e v i n g t h a t (considering that most of these children reside with their mothers) '[t]he absence of a male breadwinner affects children equally, whether the man has died or abandoned his family.'Urassa et al. (1997), who call them 'foster children' and compare their school attendance with other children, have some evidence to support this view: at most ages for both boys and girls, they find enrolment rates to be similar between orphans, foster children and children living with both parents; however, among boys 13-17 years, orphans and foster children have significantly lower enrolment and higher dropout rates than boys living with both parents.Lastly, a number of authors (e.g.Bicego et al., 2003;Foster and Williamson, 2000) feel that HIV/AIDS has a knock-on effect, and that those non-orphans whose parents' lives are disrupted by HIV/AIDS need to be considered too.For instance children may experience a reduction in their quality of life because a substantial amount of their parents' time or resources is channelled to supporting a sick relative who lives in the household or, even more critically, elsewhere (Foster and Williamson, 2000;Robson et al., 2006).
A second common weakness in the literature is that orphans are treated as a homogeneous, disadvantaged group.However, children of different age, gender, social status, ethnic group, place of residence, etc., will experience orphanhood in different ways.Furthermore, simple case finding and identification of orphanhood status has been shown to be problematic, with substantial misreporting of foster children as natural children, for example in Zimbabwe (Robertson et al., 2008).The value of 'orphan' as a meaningful analytical category is therefore open to substantial debate.
T o i d e n t i f y b o t h t h e m o s t vulnerable children and all sub-groups of children who may experience orphanhood in different ways is a difficult task, and goes well beyond the scope of the present study (but see Smart, 2002).However, the analysis below will explore one further (assumed) vulnerable group of children: those children who live with only one or neither of their parents even though both parents are alive.Our hypothesis that, taken as a whole, these are be as vulnerable as orphans (because they normally live in households of the same composition as orphans) has been confirmed by tentative findings by Unicef (1999b) (Ainsworth and Filmer, 2002).Their large numbers are a result of large-scale internal mobility and migration, of high levels of relationship break-up (Koda, 1995), and of a widespread and well-used tradition of child fostering (Urassa et al., 1997;Omari, 1995).
Here, neither Unicef's (1999b) term 'social orphans', nor the term used by Urassa et al. (1997), 'foster children', will be used to describe these children.The former term seems to imply an invariably disadvantaged life and the latter (though different in meaning and frequency in different parts of Africa) normally includes all children, including orphans, who are being looked after by people who are not their biological parents.Instead, the term 'children from spatially separated families' will be used, which is designed to express the spatial separation of the child from at least one of his or her biological parents, with no necessary implication of the causes of that separation, whether positive (i.e. a deliberate decision to operate a spatially dispersed household and migration of a child's mother of father, or fostering arrangements for sound economic reasons) or negative (i.e. a social break-up of the household due to marriage/partnership disruption).
School attendance and its inverse, school dropout, is the primary index of vulnerability by which these different groups of school-age children are identified.Dropout studies are familiar in the education literature, where voluntary dropping out of school is identified as a key symptom of educational failure, of a failure to engage the children and their parents in the long-term process of learning and educational attainment.In Tanzania, as elsewhere in Africa, the government seeks to maximise retention and continuation of those who are enrolled at the beginning of each cycle of the schools system, in primary and secondary schools.Dropping out signifies marginalisation from the schools system, but is a far from random occurrence.It is known to be systematically related to the range of individual vulnerabilities that are explored in this study (Filmer and Pritchett, 1999;Lloyd and Mench, 2008).
National HIV/AIDS prevalence in T a n z a n i a , a s i d e n t i f i e d b y t h e population-based HIV/AIDS Indicator Survey of 2003/4, was 7 per cent: 12.0 per cent of the urban female population (15-49 years of age) and 9.6 per cent of the male urban population, 10.9% of the overall urban population, but 5.3 per cent of the overall rural population (5.8 percent of rural females and 4.8 per cent or rural males) (TACAIDS, NBS, Macro, 2005).High adult mortality due to high HIV/AIDS infection has led to increasing numbers of orphans.That same DHS national survey estimated that 10.8 percent of those aged under 18 were biological orphans.In 2001 42.3 per cent of orphans were estimated to be AIDS orphans; 5.7 per cent of Tanzanian children under the age of 15 were estimated to be maternal orphans, 8 per cent paternal orphans, and 1.8 per cent double orphans (UNAIDS, Unicef and WHO, 2002).
Primary education in Tanzania is notionally compulsory, and children are expected to attend school from age 7 to age 13.Enrolment rates have undergone great fluctuations: they were low during the colonial period (e.g. in 1947 less than 10 per cent of the school age population was enrolled); rose to an impressive 93 per cent in 1980 as a result of Nyerere's strong drive for universal primary education (UPE) in the 1970s; fell steeply in the 1980s to a net enrolment rate of 47.3 per cent in 1988, but have recovered following the adoption of the 'Dakar Framework for Action, Education for All: Meeting Our Collective Commitments' which envisages primary education for all by 2015 (UNESCO, 2001).The net primary school enrolment rate for 2005 was estimated by the World Bank at 91 per cent, and Tanzania seeks to meet the Millennium Development Goal of universal primary enrolment (UPE) by 2015.

Data and methods
Using survey data from two regions in Tanzania, this paper aims to explore, first, whether and to what extent orphanhood really does adversely affect primary school attendance in Tanzania; and second, some social and geographical explanations for this lack of association, to include not only traditional explanations associated with the nature of family support networks, but also wider geographical and economic relations.
The main characteristic by which children are identified and compared in this study is geography.Children were sampled first from a region with, by Tanzanian standards, a particularly high antenatal HIV/AIDS prevalence rate (Iringa -HIV prevalence of 13.4 per cent in 2003/4), and a region with a particularly low prevalence rate (Dodoma -HIV prevalence rate of 4.9 per cent in 2003/4) (DHS, 2005), with the assumption that levels of HIV/AIDS prevalence would affect the coping capacity of kinship networks to support children (Baylies, 2002).
Secondly, within each region, children were sampled from urban areas, roadside settlements (on major roads at 30 to 60 km from the nearest town) and rural areas (within 8 to 30 km off major roads).T h i s g e o g r a p h i c a l d i f f e r e n t i a t i o n acknowledges that the very different ways in which people earn their livelihoods in those three types of areas will be likely to influence children's activities, childcare and school attendance patterns.
The study was conducted from November 2001 to February 2002 in rural and urban districts of Iringa and Dodoma R e g i o n s , s e l e c t e d t o b e b r o a d l y representative of the opposite ends of the range of Tanzanian regional prevalence rates.The large differences in prevalence rates between the two regions were consistent with our data: in Iringa, 39 per cent of our sample children of ages 7-17 year were orphans, defined as children whose mother or father or both biological parents are dead, in contrast to 'only' 18 per cent in Dodoma.However, there in these initial summaries these is already evidence which weakens the presumption that high HIV/AIDS prevalence will erode community coping mechanisms and hence make it more difficult for children to go to school.The net enrolment rate in 2001 for the high HIV/AIDS region, Iringa (74 per cent enrolment) was much higher than that for the low HIV/AIDS region, Dodoma (56 per cent enrolment) (Ministry of Education and Culture, 2002).This is consistent with Ainsworth and Filmer's (2002) finding that amongst six Sub-Saharan African countries the countries with higher HIV prevalence had higher enrolment rates (with the exception of Malawi).In Tanzania, at household level, as well as nationally, HIV/AIDS prevalence is positively (but weakly) associated with economic status: the richest quintile have highest prevalence and the poorest lowest prevalence (TACAIDS, NBS, Macro, 2005).Secondly, a preliminary exploration of the data revealed that even though different factors influenced school attendance between the two regions, differences were much more pronounced within each region: i.e. between rural areas, roadside settlements and urban areas.Thus the analysis p r e s e n t e d b e l o w c o n c e n t r a t e s o n differences between the types of areas rather than between the regions.The local geography of settlement is a more critical differentiating variable for enrolment status than is region.
In consultation with the local District Education Offices, six primary schools were selected in each region (two schools in a rural, a roadside and an urban area).To make possible an exploration of schooling of children from a range of socioeconomic backgrounds, in urban areas data were collected from a particularly well performing school and a particularly badly performing school.Urban schools were selected with the help of recent school mapping reports which contain indicators ranging from the quality of the buildings, over the availability of teaching materials, to pass rates of students (Dodoma Municipal Council, 2000;Iringa Municipal Council, 2000).
A l t h o u g h a c o m b i n a t i o n o f q u a n t i t a t i v e a n d qualitative data was collected, in this analysis only the quantitative evidence is used.For the collection of the quantitative data, to enable a comparison between groups of children with different attendance statuses, a purposive sample was drawn, initially with the intention of identifying equal proportions of regular attenders, irregular attenders, dropouts and never attenders.Definitions for these attendance groups can be found in Table 1.The sample was drawn from the same cohorts, Standards I to VI at schools in Iringa (in November at the end of the school year) and from Standards II to VII at schools in Dodoma (in January/February at the beginning of the next school year).Regular and irregular attenders were identified through attendance records, and dropouts and never attenders through snowball sampling.
Three or four questionnaires were completed for each household; first the selected children (hereafter called index children) were interviewed at school about their education experience; then they were accompanied home, where an adult from the child's household completed a general questionnaire about the socio-economic characteristics of the household and a specific questionnaire about the education of the index child; lastly a second child from the household (referred to as 'sibling', even though it could be any eligible child from the household) was interviewed about his or her education experience.The sibling was chosen preferably but not necessarily to have a different attendance status to the index child to enable us to explore reasons for intra-household differences in school attendance.If there was no second eligible child in the h o u s e h o l d , n o s e c o n d c h i l d w a s interviewed.Each respondent was interviewed alone, without the presence of the other interviewees.
In practice, since it was very difficult to find equal numbers of index children for the four groups, the sample had to be adjusted.Dropouts and never attenders were more difficult to find because many dropouts had migrated after dropping out, and families of never-attenders were often reluctant to allow their children to be interviewed.The main adjustments made were to reduce target numbers of nonattenders and, in order to boost numbers of non-attenders, to count non-attenders who were siblings as index children (because the same data was collected on both).
The numbers of index children and siblings interviewed from each attendance group are shown in columns 1 and 2 of Table 1.Column 3 shows the numbers of children in each attendance category used for the analysis.In this column, in order to come as close as possible to equal proportions of regulars, irregulars and non-attenders, the numbers of problematic attenders are maximised.Also, to avoid bias which would arise because of doublecounting of households, only one child per household is included.In the analysis presented here we are not examining intrahousehold differences.It is important to note that the proportions recorded in each category merely represent the sampling structure.They are not estimates of the proportions of children in the sample school catchments in each of these four attendance categories, not are they representative of the whole population.The purpose of the analysis is to identify the socio-demographic characteristics of each of these four attendance groups, not their relative size.

Results
An initial exploration of the data leads to findings similar to those of many other studies conducted in Tanzania: an aggregate comparison of the school attendance of non-orphans and orphans does not yield significant differences.However, differences do emerge when children's school attendance is examined by their place of residence, and whether they live with both parents, whether their family is spatially separated, or whether they are orphans.Table 2 supports the claims put forward in the introduction that orphans are not a homogeneous group.Nor are they the only vulnerable group.The table shows that whether orphans are more or less likely to go to school than children living with parents or in spatially separated households depends strongly on their place of residence.In rural areas orphans are only very slightly overrepresented amongst irregular attenders and never attenders compared with regular attenders, but in both roadside settlements and in urban areas they are strongly over-represented amongst dropouts and never attenders.In roadside settlements orphans are significantly more likely than the other groups never to have attended school, and in urban areas they are significantly more likely than children who live with both parents to have dropped out of school.
As anticipated, a group with most similar attendance records to orphans are children from spatially separated families.They also show no particular attendance problems in rural areas, but in roadside settlements are significantly more likely than children who live with both parents to be irregular attenders and never attenders, and to be school dropouts in urban areas.In urban areas, among dropouts, combining these two sets of vulnerable children in one group increases the significance of the level of difference with children living with both their parents.Numbers are small, but clearly indicative.Considering that these patterns only emerged when children from spatially separated families were examined as a separate group, Table 2 also supports the claim voiced in the introduction that not taking other vulnerable groups into account can obscure any disadvantage of orphans.That orphans and children from spatially separated families struggle more with their school attendance in roadside settlements and urban areas than they do in rural areas is initially surprising, since in the country as a whole (as in most other countries) urban school enrolment rates are higher (and increasingly so) than rural ones (Al-Samarrai and Reilly, 2000).

Discussion
The next two sections will explore some of the reasons behind these patterns of enrolment and dropout that Table 2 describes: first by examining who these children live with and who supports their education, and secondly by searching for the roots of rural and urban/roadside differences.

A) With whom do orphans and children from spatially separated households live, and who supports their schooling?
As stated above, an argument which has repeatedly been brought forward to explain why orphans may not be more vulnerable than other children is the strength of the extended family.This line of reasoning, in addition to the familiar claim that rural areas are the stronghold of extended families and that extended families structures are being eroded in urban areas (e.g.Koda, 1995;Lloyd and Blanc, 1996;Foster and Williamson, 2000), could be invoked to explain the above findings.However, this explanation does not appear to hold true for this sample.Table 3 shows that the largest proportion, around 40 per cent of orphans and also of children from spatially separated families, live in households headed by their mothers rather than by other members of their extended families.Table 3 also shows that around 30 to 40 per cent of these mothers are the only adults above the age of 18 years in the household -suggesting that they manage the household on their own (probably with help from their children).

Mo r e o v e r , o t h e r e v i d e n c e i n t h e
questionnaires identified that most of the sample receive little outside financial help beyond their main carers for their children's education.The average number of contributors to school fees for a child's education is just above one.
The second largest important group of heads of household who care for orphans and children from spatially separated families are indeed members of the extended family, namely grandmothers and aunts.Together they look after 20 to 30 per cent of orphans and around 17 per cent of children from spatially separated households.However, rather than the image that the word 'extended family' suggests (namely one of large supportive families, with the members of different generations looking after each other), just like the mothers, most of the grandmothers tend to be the only adults in their h o u s e h o l d s ( T a b l e 3 ) a n d b o t h grandmothers and aunts tend to receive no help for the schooling of the children from other people.
The only other two noteworthy categories of heads who look after these children are fathers and uncles.If mothers die, fathers are disproportionately likely to take on their children.However, because there are many fewer maternal than paternal orphans in Tanzania, in our survey in total only 10 per cent of orphans lived in households headed by their fathers.It is likely that more children lose their fathers than their mothers because of large age differences between husbands and wives and overall higher male adult mortality rates (Ainsworth and Filmer, 2002;Lloyd and Blanc, 1996).Uncles come into the picture particularly in urban areas when both parents have died.Similar to fathers, overall they are head of household to around 10 per cent of orphans.Children from spatially separated families live in father and uncle-headed households only in 10 and 4 per cent of cases respectively.Not only are there not many men who head households with orphans and children from spatially separated households, but also men in this position are much more likely to be helped in paying school fees by the mothers of the children, or by their new wives, step-mothers of the sampled children, than women in the same circumstances are helped by men.In rural areas 27.8 per cent of fathers who live with their child without the corresponding mother get a contribution from the latter, whereas only 9.8 per cent of mothers in the same position receive help from the men.
The data offer no evidence that the extended family is more supportive in rural than in urban and roadside areas.The proportion of children looked after in households headed by members of the extended family is, at around 35 per cent, similar in both areas.Also, the mean number of people (generally just one!) who contribute to a child's education are similar; they even tend to be slightly higher in urban and roadside areas.Beyond n a t u r a l p a r e n t s , i n r u r a l a r e a s grandmothers are most important in taking on vulnerable children; in urban areas it is aunts and uncles.
To sum up, the credit for looking after orphans and children from spatially separated families should go to femaleheaded households, particularly mothers, but also grandmothers and aunts (who together looked after 60 to 70 per cent of these children in our survey) rather than a generally 'strong extended family'.1.In contrast to the UNAIDS definition of paternal, maternal and double orphans cited in footnote 2, here categories of orphans do not overlap, i.e. a child in the category of having lost a father has not lost a mother, and a child in the category of having lost a mother has not lost a father.2. 'Head of household: 'child's generation'' refers to households headed by the child him-or herself, a sibling, or a brother-or sister-in-law of the child.

B) Children's schooling in rural, urban and roadside female-headed households
The fact that the majority of orphans and children from spatially separated households are living in femaleheaded households is rooted in women's traditional role as nurturers, reinforced by Tanzanian law which generally grants women custody of children under the age of 7 (Tanzania Gender Networking Programme (TGNP) et al., 1997).It is also rooted in women having a greater incentive to look after their children because of their tending to be more dependent than men on their offspring in later years.This is particularly so in polygamous households, where older men may be looked after by younger wives (Lloyd and Blanc, 1996).However, the 2003/4 National HIV/AIDS Indicator Survey showed that the proportion of polygamous households was falling rapidly, and was only 10 per cent of all households (TACAIDS, NBS and Macro, 2005).
But why is it that in the present study female heads of household have been shown to enable children to attend school (i.e. they fulfil their traditional role as child carers) in rural areas, but much less so in urban and roadside settlements?Most female-headed households have formerly been male-headed.With 21.4 per cent of all household heads, widows nationally constitute the largest proportion of female heads of household, followed by divorced and separated women (13.7 per cent of all household heads); single women (9.3 per cent) constitute the smallest proportion of female heads (TGNP et al. 1997).These now female heads of household are unlikely to have had much control over finances when they were still living with a male partner, since there is substantial evidence both from rural and urban areas that when women and men live together it is mostly men who control the household budgets (Creighton and Omari, 1995;TGNP et al., 1997).
A combination of women tending to take responsibility for caring for their children and their spending power being restricted by their husbands suggests that when they lose their husbands there is potential for them to care better for their children than before.This has been found by a number of authors in studies from around the world (Meekers and Meekers, 1997;Creighton and Omari, 1995).The strength of female-headed households being directly reflected in children's school attendance in Africa has been found by Ainsworth and Filmer (2002) and Lloyd and Gage-Brandon (1994) in their respective studies in Tanzania and Ghana, where they observed that children in female-headed households were more likely to enrol in school than children in male-headed households.Similarly Lloyd and Blanc (1996) found that in seven Sub-Saharan African countries (including Tanzania), after controlling for all other variables, including the socio-economic status of the household, children in femaleheaded households were consistently more likely to be enrolled in school and to have completed grade four than children living in households headed by men.Whereas Lloyd and Blanc (1996) find that female heads of household disproportionately send their children to school, despite their socio-economic disadvantage, in the present study the ability of female heads of household to send their children to school appears to be closely linked to their socioeconomic status.
Just as was shown above that orphans and children from spatially separated families have problems with their school attendance in urban and roadside settlements but hardly in rural areas, female-headed households in urban and roadside settlements are found to be more vulnerable socio-economically than female-headed households in rural areas (Table 4).In rural areas there are relatively small and not significant differences between female and male-headed households in the proportions with a low quality of dwelling (as recorded by the interviewers) and with serious economic problems (as stated by an adult interviewee).In contrast, in urban and roadside areas female-headed households are significantly more likely to have reported serious economic problems than male-headed households, and figures for the quality of dwelling (even if not significant) point in the same direction.
Considering that an estimated 78 per cent of the Tanzanian population continues to reside in rural areas (Population Reference Bureau, 2005), the finding that in rural areas female-headed households do not lag far behind maleheaded households is in turn supported by national data from the 2000/1 Household Budget Survey (HBS): the proportions of male-and female-headed households experiencing poverty are very similar: 35.8 and 35.3 per cent respectively (National Bureau of Statistics (NBS, 2002).
There are two broad reasons why in Tanzania female heads of household may be worse off socio-economically in urban and roadside settlements than in rural areas, and therefore struggle more to send their children to school.First, the history of urban women being engaged in income generating activities is shorter and more problematic than it is for rural women, and consequently relatively few women are involved in paid work and work under more difficult conditions than in rural areas.In contrast, in rural areas, Tanzanian women have a long history of being the principal workers, not only in their home villages, but during the colonial period also as plantation workers (Bryceson, 1995).While the 2000/1 HBS survey shows that roughly equal proportions (75 per cent of men and women) in rural areas are engaged in farm work (NBS, 2002), a study b y t h e M i n i s t r y o f C o m m u n i t y Development in 1986 (cited in International Labour Office (ILO), 1996) revealed that women worked an average of 2600 hours in farming a year, in contrast to only 1800 hours worked by men.
A second reason for female-headed households to struggle more outside of rural areas is that to sustain an urban household requires much higher cash expenditures than to keep up a rural h o u s e h o l d .W h e r e a s m o s t r u r a l households own land, a dwelling, livestock and poultry and grow a large proportion of the food they consume, most urban households do neither own productive assets nor a dwelling, and a large p r o p o r t i o n o f t h e i r c o n s u m p t i o n expenditure is on food.In urban households education also takes up a higher proportion of the consumption expenditure than in rural areas.In the present survey, the mean total costs of schooling per child for the year 2001 were with 16,046/-Tanzanian Shillings (TSh), significantly higher in urban areas than in rural areas and roadside settlements (6369/-TSh).This is consistent with the finding of the national Household and Budget Survey of 2000/1, 42.2 per cent of households in Dar es Salaam gave high education costs as a reason for children's non-attendance in contrast to only 9.6 per cent of rural households (NBS, 2002).Unfortunately, there is no national data for roadside settlements.However the indepth interviews held as part of the present study indicated that people's livelihoods in roadside settlements resembled those in urban areas more than those in rural areas: many households did not own farms, many lived in rented accommodation, and, being situated next to major roads, there was a greater exposure and expenditure on consumer goods.
It is easier to earn the substantial income required to sustain an urban household when many household members contribute.A number of authors have found that since the economic crisis of the 1980s, the need for more cash has meant that in urban areas involving as many household members as possible in economic activity has become a crucial strategy (Creighton and Omari, 1995).
Urban female-headed households who in many cases include only one adult (see Table 3) are, as a consequence, at a great disadvantage (Campbell, 1995): not just in terms of it being difficult for one individual to have many income sources, but also b e c a u s e t h e s u r v i v a l s t r a t e g y o f establishing a shamba is a less feasible option, because shambas tend to be out of town, and at least one adult normally has to start living there (Campbell, 1995).
In contrast, as Bryceson (1995: 47-8) puts it, the nature of Tanzania's agriculture, most of which rests on hoe cultivation, means that 'the work process, although facilitated by collective effort, can be readily pursued by single individuals'.Hence rural households with only one female adult are not necessarily at a disadvantage, particularly not those with primary school aged children, who will most certainly help with the farm work.Contrary to expectations, this does not normally affect the children's education.Children tend to fit the farm work around their schoolwork.
Of course, whether a child goes to school is not solely dependent on the gender of the head of household and the head of household's ease in making a living.It will also depend on the cost of schooling, the access to schooling, the quality of schooling, children's and their guardian's perception of the usefulness of schooling, the parents' education, the importance given to other activities the children might have to perform, etc.However, the above analysis does indicate that headship is an important issue for school attendance, and that the ability of female-headed households to cope economically and to send children to s c h o o l d i f f e r s s t r o n g l y b e t w e e n urban/roadside and rural locations in Tanzania.1. * means that the proportion for female headship is significantly higher than that for male headship at the 0.05 level (chi-square test, Pearson's r).

Conclusion
In these survey data there is no sign of the generally assumed universal, negative effect of orphanhood on primary school attendance.
This study gives strength to other recent voices which hold that it is unhelpful to single orphans out as the most vulnerable group of children.Some orphans are well off, and there are other groups of children who can also be vulnerable.Policies targeted exclusively at orphans, and perhaps even solely at HIV/AIDS orphans, run the risk of failing other children in need.This is however not to deny that orphans are a group of children with certain particular needs.It is for instance indisputable that the death of parents (often preceded by prolonged periods of severe illness) is likely to trigger psychological problems (Makame et al., 2002;Mukyogo and Williams, 1991;Sengendo and Nambi, 1997;Unicef, 1999b).However, the large group of children to whom the study draws attention in addition to orphans are children whose parents are both alive, but who live with only one or neither of them: 'children from spatially separated families'.They make up a quarter to a third of Tanzanian children and have in common with orphans that they do not live with both of their parents, which can put them in an equally vulnerable position.
A critical finding arising from the evidence presented here that should be of concern to opinion shapers and policy makers is that it is important not to view orphans as a homogeneous group.One of the many characteristics by which orphans' (and other children's) experiences differ in Tanzania is their place of residence.Geographical location has been shown to be an important factor in whether both orphans and also children from spatially separated families attend primary school.These two groups of vulnerable children have been shown to display poorer levels of school attendance than children living with both their parents in urban and roadside settlements, but hardly so in rural areas.Contrary to popular perception, the majority of orphans and children from spatially separated families are looked after on a daily basis by their lone mothers or single female elders (in particular aunts and grandmothers), who tend not to get any support from other people in educating the children in their care, and not mainly by the much invoked 'extended family'.Hence children's differential s c h o o l a t t e n d a n c e b e t w e e n urban/roadside and rural areas is related to the extent to which female-headed households are able to find sustainable livelihoods in those different areas.
William T.S. Gould, Ulli S. Huber: Orphanhood, Vulnerability and Primary School Attendance: Evidence from a School-Based Survey in Two Regions of Tanzania Table 3 Household headship by parents' survival and living status and children's place of residence (

Table 1 Children's sample
William T.S. Gould, Ulli S. Huber: Orphanhood, Vulnerability and Primary School Attendance: Evidence from a School-Based Survey in Two Regions of Tanzania

Table 2 Primary school attendance by place of residence, parental living arrangement and parental survival status
In the chi -square tests regu lar attenders were compared with irregular attenders, and dropouts and never-attenders with attenders.William T.S. Gould, Ulli S. Huber: Orphanhood, Vulnerability and Primary School Attendance: Evidence from a School-Based Survey in Two Regions of Tanzania