Targeting AIDS orphans and child-headed households? A perspective from national surveys in South Africa, 1995–2005

Abstract

In the HIV and AIDS sphere, children remain on the margins with respect to advocacy, prevention, treatment and care. Moreover, concern is generally limited to specific categories of children, most especially children living with HIV, orphaned children and child-headed households. Excluded from view are the very large numbers of children affected by generalized HIV/AIDS epidemics, now in advanced stages, in already impoverished countries in southern Africa. In this paper, we use information from comparable national household surveys in South Africa, in five waves between 1995 and 2005, to examine the impact of HIV and AIDS on children and on the structure of the households in which they find themselves. The question posed is whether it is appropriate to target orphans and child-headed households in this context. The data indicate that orphaning, particularly loss of a mother, tripled during this period, as is to be expected from rising adult mortality. Though they remain a small proportion, child-only households also rose markedly during this time. However, difficult as their situation is, neither orphans nor child-only households appear to be the worst-off children, at least from the point of view of reported sources of financial support and per capita monthly expenditure. Households headed by single adults and young adults are economically vulnerable groups not yet included in efforts to support affected children and families. Poverty is a pitiless backdrop to the AIDS epidemic and needs to be at the heart of strategies to address the needs of all vulnerable children in hard-hit communities.

Keywords:

• children
• AIDS
• poverty
• orphans

Introduction

The HIV/AIDS pandemic is causing untold suffering, including for children. Still ‘the missing face of AIDS’ (UNAIDS & UNICEF, 2005), children are not yet included in global or national advocacy, prevention, treatment or support strategies. As was anticipated by the conception of the epidemic unfolding in waves (Barnett & Whiteside, 2002), the impact on children epitomizes the crises of care attendant on high levels of infection, progressive illness and death. While many organizations, globally and locally, are working to make resources available and to provide services and support to children affected by HIV and AIDS they have, to date, had little large-scale impact (Richter, Foster & Sherr, 2006; Richter, Manegold & Pather, 2004). One reason for this may be the tendency to target specifically orphans and child-headed households in impoverished circumstances, in which much larger numbers of children are hungry, grow poorly, have few opportunities to develop their potential and have little protection from abuse and exploitation (Singhal & Howard, 2003).

There is a growing body of published and unpublished literature comparing the living circumstances, health and nutrition, education and adjustment of orphaned and non-orphaned children and drawing attention to the situation of children heading households, especially in southern Africa (Bishai et al., 2003; Case, Paxson, & Ableidinger, 2004). Some studies document the adverse effects of orphaning; for example: Arnab and Serumaga-Zake (2006) with respect to socioeconomic conditions; Beegle, De Weerdt and Dercon (2005) health and education; de Wagt & Connolly (2005) food insecurity and economic opportunity; and Gregson et al. (2005) education. However, others report few differences between orphans and equally poor and disadvantaged children with living parents; for example: Ainsworth, Beegle and Koda (2005) with respect to schooling; Crampin et al. (2005) mortality and health; and Monasch and Boerma (2004) nutritional status.

The results of the research cited are inconsistent for a number of reasons. For one, the terminology and categorizations used to refer to affected children are problematic. Orphaning in the USAID and other models is a technical term for a child whose mother or father has died (USAID et al., 2002, UNAIDS et al., 2004). Although double orphaning is increasing, as a result of this definitional approach most children referred to as orphans in the scientific, programming and advocacy literature, in fact have a surviving biological parent. Studies therefore often include different groups of children – all orphans, only double orphans, or only maternal or paternal orphans. Such mixed groups are unlikely to show many consistent characteristics across a wide age range, gender and differing social circumstances.

This work is premised on the assumption that knowing which children are vulnerable, why and how is important because it determines our comprehension of the problems children and families experience and points to the solutions to be pursued. As described above, the formulation of the problem and the resultant focus of international and local attention with respect to children affected by HIV and AIDS, has targeted primarily two categories of children – orphans and child-headed households. Images of orphaned children living alone with only their siblings or with aged grandmothers are powerful indeed, and the children in such circumstances certainly require support, but there are several reasons why restricting research and intervention efforts to these groups of children may be inappropriate. For one, it has long been known that membership of unidimensional categories does not predict children's developmental course, even with respect to seemingly robust biological and/or socioeconomic factors (Sameroff & Chandler, 1975; Werner & Smith, 1989). Secondly, visible groups of children, such as street children or orphans, are frequently the tip of the iceberg of large numbers of children whose circumstances are equally, or more, precarious. The analysis presented in this paper is designed to examine the economic vulnerability of household structures, such as child-headed households and households with only elderly persons and children, which are often the focus of attention and responses. These are compared to household structures, which may be overlooked when the challenge is framed by an orphan lens, such as young adults living with children, single parent homes and large households.

This comparison of household types through comparable national level household surveys allows us to question the targeting of orphans and child-headed households, given the generally poor circumstances of children. We document the rates of orphaning and child-headed households and empirically examine which children are vulnerable, with a focus on economic vulnerability. Although there are other sources of vulnerability, economic vulnerability is central as many of the responses are designed to address the material needs of children.

Methods

The data presented in this paper are drawn from five comparable household surveys conducted by Statistics South Africa in South Africa between 1995 and 2005. They include three so-called October Household Surveys (OHS) – because they were conducted in the month of October, 1995, 1997 and 1999 and two General Household Surveys (GHS), which replaced the OHS surveys, in 2002 and 2005.

The methodology used in the OHS surveys, which were conducted each year, varied a little. Because trends are important for this analysis, only the three most comparable years are examined (Casale, 2003). The GHS is the replacement of the OHS, thus it was appropriate to continue the comparison with them by selecting years to render a similar time interval between surveys.1

The OHS was launched in 1993 by the central statistics department as an annual sample survey to render accurate and reliable national socio-economic data. Information was collected at both the household and individual level on a range of social, economic and demographic variables including household composition, education levels, personal income, access to services and the like. There is almost complete overlap in the questions asked across the surveys and the same sampling approach was maintained. There are some discrepancies between the OHS and GHS surveys but, in order to avoid unnecessary diversions, only the most similarly sampled data and phrased questions are included in this analysis.

Table 1 details the data sets used and the approximate sample size for each year considered in the trend analysis. Data examining the current situation are all from 2005.

Table 1. Surveys used, source and sample size.

Year	Survey	Source	Sample size	Notation
1995	October Household Survey	Statistics South Africa	30,000 Households	OHS1995
1997	October Household Survey	Statistics South Africa	30,000 Households	OHS1997
1999	October Household Survey	Statistics South Africa	30,000 Households	OHS1999
2002	General Household Survey	Statistics South Africa	30,000 Households	GHS 2002
2005	General Household Survey	Statistics South Africa	32,000 Households	GHS 2005

The results presented are essentially descriptive in nature. In all cases the weighted data were used, varying between household and person weights where appropriate. All references to significance are based on the use the 5% level as a cut off, using Chi-squared tests. Tables report significant differences in distributions relative to the largest category. The majority of tests referred to in the text involve significance in the probability of specific responses.

The analysis covers, first, the trends in orphaning in South Africa over the past ten years. It then examines the distribution of children across different household types and how this has changed over the decade. Following the trend analysis, the results concentrate on the most recent data with respect to orphaning in 2005. Thereafter, the focus is on differences in economic vulnerability faced by different household types.

Following on from the discussion of orphan trends, we examine the distribution of children across household types. Households, in South Africa as elsewhere, can have a variety of forms (Amoateng & Richter, 2003), including child-headed households and so-called skip-generation households, which contain only children and older persons (+60 years), with no working-age adults present.

Many households reported in surveys to be child-headed actually contain adults. This may result from data problems (Ziehl, 2002), male children being listed as heads when the only adults are women (Desmond, Richter, Makiwane, & Amoateng, 2003), adults being inconsistently present in the household and therefore inadvertently omitted from listings, or for other reasons. Given these problems, we consider only households in which no adults are reported, which we call child-only households.

Skip-generation households refer to those in which children are in the care of a person, usually a grandparent over the age of 60, with no younger adults in the household. There are, of course, instances where the grandparent is younger than 60, but these are not included as the focus is on households that do not contain any adults of working age. The retirement age for women in South Africa is 60 years of age and this was selected as the cut off.

There are other household types that might have lowered capacity to care for children. Firstly, there are households that contain only children and young adults – those aged between 18 and 25, called young adult households. Secondly, there are households in which there is only one adult caring for one or more children, referred to below as single adult households.

Households that do not fall into one of these four categories are referred to as ‘other’ households. These are the most common form of household structure – containing at least one child and more than one adult, where at least one of the adults is of working age. There are also households that contain no children, which have increased progressively over the decade – from 33.35 to 41.19% of households – but these are omitted from the analyses. The household categories examined can be summarised as follows:

•	Child-only: All members below 18 years of age.
•	Skip-generation: Contains at least one child with all members either below 18 or over 59.
•	Young adult: Contains at least one child and at least one adult below the age of 25 and no members over the age of 24.
•	Single adult: Contains at least one child and only one adult, where the adult is over the age of 24.
•	‘Other’: All other households containing children not captured in one of the above.

Variation across the above categories was examined, using 2005 data, by a range of indicators of vulnerability including: household size, expenditure, reported child hunger and main source of income. In order to account for household size, we consider per capita expenditure. There are ways of controlling for household size, all of which require a number of assumptions about the relative weight to be attached to members of different ages. The results presented below are based on the simplest method, which uses the mid-point of the expenditure range reported and assumes common weights for all members. This tends to make households with more children appear to be worse of than they are – if, as is generally assumed, children need less than adults. Other methods, with different weightings for children of different ages were tested, but the ordering of household structures by income remains the same, even if the distribution across categories differs. Expenditure results are reported in South Africa Rands (R1 equals approximately US$ 7).

Results

Orphan numbers

Examining first the trend in orphan numbers, Table 2 shows children according to how their parent's status was reported. The paternal orphaning figures are almost certainly over-estimates as a result of the absenteeism affect, whereby fathers who are not involved are reported as dead (Udjo, 1998), but the implications for children's care remain.

Table 2. Reported status of children's biological parents.

	October Household Survey 1995 (%)	October Household Survey 1997 (%)	General Household Survey 2002 (%)	General Household Survey 2005 (%)
Both alive	85.09	87.82	84.57	82.5
Mother dead Father alive	1.49	1.49	2.88	3.01
Father dead Mother alive	11.84	9.58	10.81	11.37
Both dead	1.59	1.12	1.75	3.12
Total	100	100	100	100
Notes: 1. The OHS 1999 was excluded from this analysis because it only asks about the mother's status and in a different section of the questionnaire from the other surveys.
2. A conservative approach to orphaning was adopted, in that parents whose status was reported as unknown are included in the alive category. The numbers involved, however, are very small, particularly in the later surveys.
Source: Own calculations based on Stats SA data.

The data indicate that over the past 10 years the level of reported maternal orphaning has doubled (from 1.49 to 3.01% of children), while paternal orphaning has remained relatively constant. As a result of the increase in maternal orphaning, the percentage of children who are reported to have lost both parents – double orphans – has also doubled.

Orphaning is increasing; but while parental loss might increase the risk to which children are exposed, it does not define the risk. What is also important for children is the circumstances in which they live.

Where children live

Table 3 shows the percentage of children in South African households, according to the six household categories outlined above, from 1995 to 2005. In order to create mutually exclusive household categories, single adult households in which the adult is younger than 25 are included in the young adult household category.

Table 3. Percentage of children living in different household types (1995–2005).

	October Household Survey 1995 (%)	October Household Survey 1997 (%)	October Household Survey 1999 (%)	General Household Survey 2002 (%)	General Household Survey 2003 (%)
No child in household	–	–	–	–	–
No adult – only children	0.11	0.34	0.45	0.67	0.66
Skip-generation	1.69	2.44	2.23	2.3	2.29
Young adult (18–25) with children	1.22	1.86	1.71	1.88	2.27
Single adult with children	7.31	9.28	9.39	9.71	11.27
Other	89.68	86.09	86.22	85.44	83.52
Total	100	100	100	100	100
Note: Source: Own calculations based on Stats SA data.

The majority of children live in so-called ‘other’ households – containing child/ren and more than one adult, of which at least one is older than 25 but younger than 60 – nearly 90% of all children in 1995, declining to 84% in 2005. The percentage of children in child-only households increased markedly from 0.1 to 0.66%. In 2005 there were in the region of 64,000 households in South Africa without an adult present, comprising 120,000 children. Most (70%) child-only households consist of only one child and the majority (90%) are headed by individuals of 15 years or older; the majority (64%) are headed by males and the greatest proportion (78%) of children living alone are male – comprising nearly a quarter of all child-only households.

The percentage of children living in households with only young adult households also increased noticeably over the period, as did the percentage of children in single adult households and skip-generation households.

The increase in seemingly vulnerable households (child-only, skip-generation, young adult and single adult households) is associated with the increase in orphaning. The data provide strong evidence that children reported to be biological orphans are more likely than other children to find themselves in these household types. Table 4 shows the 2005 household types in which children live, given the reported biological status of their parents.

Table 4. Percentage of children living in different household types by reported status of biological parents (2005 General Household Survey).

	Both alive (%)	Mother dead father alive (%)	Father dead mother alive (%)	Both dead (%)
Significance		*	*	*
No adult – only children	0.6	0.9	0.9	2
Skip-generation	1.9	6.4	3	6.8
Young adult (18–25) with children	1.9	4.4	2.8	7
Single adult with children	10	10	19	9.7
‘Other’	85.6	78.3	74.3	74.5
Total	100	100	100	100
Notes: 1. Significance of difference in distribution across household types relative to category ‘both alive’; *denotes significant at 5% level.
Source: Own calculations based on Stats SA data.

While the majority of children who were reported to have lost both parents were in an ‘other’ household (a household with more than one adult, at least one of whom is older than 25 and younger than 60), 25% of them were in a child-only, skip-generation, young or single adult household. However, 15% of children reported to have two living parents also found themselves in child-only, skip-generation, young or single adult households; but the situation of orphaned and non-orphaned children is different in a crucial respect. Over half of children with both parents reported as alive, but living in a young adult household, were living with at least their mother, and over 80% of those in single adult households were living with their mother.

Although about a quarter of double orphans find themselves in vulnerable households, the large majority are living in ‘other’ households with more than one adult and in which at least one adult is of working age and over 24. Skip-generation households, as a specific category, make up a very small percentage of households and comprise less than a third of the vulnerable households in which double orphans are found. Over 20% of ‘other’ households are headed by a person over 60 years old, the difference being that these households also have younger adults present.

Indicators of vulnerability across household types

We examined a small number of indicators of vulnerability, across household types, using data from the most recent survey, the GHS 2005. The depth of the analysis is limited by the nature of the survey data, but some important issues can be raised and some questions shaped. It should be kept in mind that the data collected from child-only households are obviously reported by a child, with implications for validity and reliability.

Household size

Table 5 shows the total number of people in the household and then the number of children in the household, by household type.

Table 5. Household size and number of children by household type (2005 General Household Survey).

People n	No adult – only children	Skip-generation	Young adult with children	Single adult with children	Other
Number of people (%)
Sig.	*	*	*	*
1	54	–	–	–	–
2	28	29	35	31	–
3	9	35	29	32	16
4	7	22	15	18	24
5 +	2	14	21	19	60
Total	100	100	100	100	100
Number of children (%)
Sig.	*	*	*	NS
1	54	43	54	31	30
2	28	30	23	32	31
3	9	16	9	18	18
4	7	5	10	12	11
5 +	2	6	4	7	9
Total	100	100	100	100	100
Notes: Significance of distribution relative to category ‘other’ households, *denotes 5% significance, NS = not significant.
Source: Own calculations based on Stats SA data.

The majority (over 60%) of households contain only one or two children. What is most striking though, is that the majority of households with no adults consist of only one child; 82% consist of one or two children. The general image portrayed of child-headed households is of groups of children living alone, but only 18% of households without adults contain more than two children. This means that, nationally, there are approximately 11,500 households containing more than two children living without adults. The most common of these households is a boy over 15 living alone or with only one other child.

Single adult households are smaller in total size compared to ‘other’ households but have a similar distribution of children. This implies that care and support in these households may be a greater burden on the one adult present.

Expenditure

An analysis of monthly expenditure by household type indicates severe generalized poverty; between 13 and 54% of children live in households that spend less than R400 per month (equivalent to about US$ 57). Skip-generation households are somewhat protected by the universal old age pension of R780 per month from the extremes of poverty experienced by other vulnerable households.

Total household expenditure can be misleading because it does not take into consideration household size. Table 6 presents the percentage of household types with estimated per capita monthly expenditure patterns, using the mid-point of expenditure categories as a starting point.

Table 6. Estimated per capita expenditure in Rands, percentage distribution by household structure (2005 General Household Survey).

Monthly expenditure	No adult – only children (%)	Skip-generation (%)	Young adult with children (%)	Single adult with children (%)	‘Other’ (%)
Sig.	*	*	*	*
R0–199	24	39	57	48	48
R200–299	41	22	15	14	15
R300–499	10	25	16	14	10
R500–999	19	9	10	12	19
R1000 +	6	5	2	12	8
Total	100	100	100	100	100
Notes: Significance of distribution relative to category ‘other’ households, *denotes 5% significance.
R7 = ±US$ 1, January 2007.
Per capita estimates ignore the differences in consumption associated with children relative to adults and ignore economies of scale. Adjustments were made for both of these factors, but the pattern across households remained unchanged. Using the mid-point of expenditure estimates can be problematic as it causes households to clump together. Distribution within the category can be very important when combined with household size; for this reason the lowest category was made large to avoid distortions associated with the combination. The survey also asks for estimated expenditure across a number of items. Combined, these provide an estimate of total expenditure not in pre-defined categories. These estimates were also used, but again did not affect the general pattern between households a great deal.
Source: Own calculations based on Stats SA data.

The proportion of households falling into the bottom category is significantly different between all types with the exception that the difference between single adult households and ‘other’ households is not statistically significant.

Although there are problems using expenditure data of this type in this way, it does highlight the importance of accounting for household size. The analysis suggests that half of all children in households with working age adults (young and single adult households and ‘other’ households) live in very difficult circumstances, with per capita expenditure of less than R200 per month. While child-only households have low overall expenditure, their small size results in relatively less disadvantaged per capita expenditure than other households.2

Possibly a more reliable way to investigate the relative need of different household types is to examine reported frequencies of child hunger. The survey contained a question about how often in the past 12 months a child went hungry because there was not enough food. The results are presented in Table 7.

Table 7. Reported child hunger percentage distribution by household type (2005 General Household Survey).

Reports of hunger in last 12 months	No adult – only children (%)	Skip-generation (%)	Young adult with children (%)	Single adult with children (%)	‘Other’ (%)
Sig.	*	*	*	*
Never	74	81	67	69	77
Seldom	3	6	4	6	5
Sometimes	14	10	23	19	13
Often	2	0.9	4	3	3
Always	3	1	2	2	2
Unspecified	4	1.1	–	1	–
Total	100	100	100	100	100
Notes: Significance of distribution relative to category ‘other’ households, *denotes 5% significance.
2. Respondents may have been embarrassed to say that children went hungry resulting in under-reporting. This would be a particular problem if certain respondents were more likely to be embarrassed than others. It may be the case that child-only and young adult households were less embarrassed to say that they were hungry.
Source: Own calculations based on Stats SA data.

Close to 20% of all households reported a child being hungry because of lack of food ‘sometimes’ during the previous 12 months. On the index of reported hunger, child-only households do not seem to be doing as badly as young adult and single adult households and have much the same rates of reported hunger as ‘other’ households. The likelihood that a child-only household would report hunger sometimes, often or always was significantly less than the likelihood for young-adult households. It was also less than for single adult households, but this difference was not significant.

Sources of support

If some child-only households seem to be doing better that many households with adults, at least in terms of consumption and reported hunger, where do they get their support? The GHS asks what the household's main source of income is, the results of which are shown in Table 8.

Table 8. Main source of income percentage distribution by household type (2005 General Household Survey).

Income source	No adult – only children (%)	Skip-generation (%)	Young adult with children (%)	Single adult with children (%)	‘Other’ (%)
Sig.	*	*	*	*
Salaries/wages	10	4	29	37	58
Remittances	73	5	55	27	8
Pensions/grants	6	87	12	25	27
Sales of farm products	0	0	0	1	1
Other non-farm income	2	4	1	5	5
No-income	8	0	2	3	1
Unspecified	1	–	1	2	–
Total	100	100	100	100	100
Notes: Significance of distribution relative to category ‘other’ households, *denotes 5% significance.
Source: Own calculations based on Stats SA data.

The results indicate that child-only households rely mainly on remittances as their main source of income and reported receiving these significantly more often than all other household types, an important source also for the majority of young adult households. However, 9% of child-only households reported no or unspecified income, which may indicate that they are engaging in the most basic survival activities, such as begging. Pensions are, unsurprisingly, the main source of income in skip-generation households. Social grants, in general, including pensions are also the major source of income in roughly a quarter of single adult households and ‘other’ households.

These results suggest some level of adult involvement in child-only households. Further information in this regard is provided by examining the number of social grants received in each household type. The data shows that close to 20% of child-only households report receiving one or more grants. This indicates some level of adult engagement because, at the time of the survey, children could not access grants without the involvement of an adult. Furthermore, when asked if an adult had gone hungry, child-only households responded to this question in 40% of cases, again indicating some connection with one or more adults.

Close to half of all young, single adult and ‘other’ households receive no social support, despite very high levels of poverty. Most of these households are poor and those with children under the age of 15 should be able to access at least the child support grant. Just over 35% of young adult households contain a child under the age of 15 and report household expenditure less than R1199 a month3 but receive no grants; 25% of single adult households and 12% of ‘other’ households are in a similar position. These households are not only financially vulnerable but also seem to lack the capacity to access available forms of social assistance for which they are eligible.

Discussion

The programming literature on children and AIDS tends to focus on orphans, child-headed and skip-generation households, partly because funders and others are roused by the poignancy of children in such vulnerable conditions (Henderson, 2006; Meintjies & Giese, 2006). However, several unintended adverse consequences result from a narrowly targeted approach in the context of a generalized epidemic and widespread poverty, including stigmatization of orphans and exclusion of other vulnerable children. National survey data provides a counterbalancing source of information,

Like many of its neighbours, South Africa has an advanced HIV and AIDS epidemic (Dorrington, Bourne, Bradshaw, Laubscher, & Timaeus, 2001), with rising orphan rates (Anderson & Phillips, 2006). Tracking orphan numbers as a measure of the increase in risk to children and the growing burden of care as a result of HIV/AIDS is appropriate. But using this to target policy and programme responses is, however, often not. For a start, it is important to differentiate between social and biological orphaning. When a child looses both their parents they are a biological orphan, but not necessarily a social orphan, as in a child without family. Biological orphans most often remain in the care of supportive families (Foster et al., 1995). On the other hand, there are children whose parent/s are alive, but absent or generally uninvolved and who may not have other family or only have very little family to turn too for support. As argued by Evans, ‘A broader concept of ‘social orphans’ is needed, that is children whose parents (or families4) are unable to provide for them, rather than just children orphaned by AIDS’ (2005, p. 126).

According to the data presented here, the vast majority of orphaned children are in the care of their mothers, if their father is deceased, and most live with extended family if their mother is deceased. But, by 2005, about 2% were living in child-only households. Child-only households increased six fold during the decade 1995–2005, rising from a low base of 0.11%. However, the needs of these households may be contrary to media stereotypes; child-only households tend, in the main, to consist of only one child, most frequently a boy older than 15.

We identified two other potentially vulnerable families from a structural point of view, in addition to child-only and skip-generation households: (1) households headed by young adults, and (2) households headed by a single working age adult. All indicators demonstrate the extent of household and child vulnerability on a national scale in South Africa, with deep poverty, and occasional hunger, amongst a sixth of children and a high degree of dependence on secondary income sources such as remittances and social grants.

Child-only and skip-generation households are indeed vulnerable, and obviously merit the attention of the state and donors. But many other households seem even more economically vulnerable, especially households headed by a single adult or by a young adult. More seemingly normal families, with at least one adult of working age, are poorer than either child-only or skip-generation households and more likely than either to report occasional child hunger. While reference is made here to measures of economic vulnerability, it should be noted that child-only and skip-generation households may well be more vulnerable in other regards. What should also be kept in mind is that both orphans and non-orphans live in all of these household types.

From this data a number of key points emerge. Arguably the most important is that the AIDS epidemic is impacting on children and families in the context of widespread, and even extreme, poverty. Many children are at risk for a range of reasons; most because their families are destitute and some because they have lost their mother or father or both. Others are at risk because they live in household structures with seemingly little security and high levels of dependence on only one adult or young adults. It would appear that, increasingly, orphaned children are reported in such vulnerable households.

In conclusion, the loss of one or both parents is likely to increase the risks faced by children, but it is not the sole determinant of hardship, including extreme poverty and hunger. A large number of biologically orphaned children are living in better circumstances than many children with living parents. Orphaning resulting from HIV/AIDS needs to be tracked as a measure of the scale of increased risk for children associated with the epidemic. This tracking should not be translated into targeting or criteria for assistance. The loss of a parent is a difficult thing for a child to go through, but in terms of the long-term risk to the child's survival, development, education and adjustment, what is important is where the child finds him- or herself before (for example, see Gray et al., 2006) and after the parent's death. That is, who takes on the care of the child and in what circumstances? There are other indicators of risk that could better guide responses, including stability, income predictability and food security. The most vulnerable orphans will also be covered by such an approach. At the point of delivery, the only marker of need is need itself. Targeting orphans or particular household structures ensures that vulnerable children in these groups are covered but may mean that non- vulnerable children are also included – while children in other groups or household types are missed, although they may be as much, or more, in need.

In terms of material support, limiting interventions to small groups of orphaned children and child-only and skip-generation households is inappropriate in contexts where many children with living parents are living in such difficult curcumstances. From the perspective of economic support, what is needed is large-scale support to communities, the reduction of poverty and improved access to services. Targeting is the optimal approach when only a few individuals or groups need help and it is necessary to avoid leakage to those who are not so needy. This is not the situation in many southern African countries. In South Africa, for example, more than a few – some 9–11 million5 children – are being supported on less than a $1 a day.

Notes

1. It would have been ideal to use years 2001, 2003 and 2005 but the survey started in comparable form only in 2002.

2. This conclusion needs to be considered with caution given the problems with using category data and not knowing where in the category households fall. Here, it has been assumed that, on average, households fall in the middle of the expenditure category they reported, but it may be that for child-only households they were on average closer to the bottom of the range than ‘other’ households, which would bias the results.

3. The eligible income level to receive a Child Support Grant is between R800 and R1100 per month depending on the area in which they live.

4. Added by authors.

5. The estimates are based on expenditure data and conservatively assume an equal spread of household expenditure across all household members. The 9 million is the most conservative estimate because it assumes that households are spending the highest possible amount in the income bracket that they selected. The 11 million is likely to be more accurate because it is based on the sum of reported expenditure, which gives a point rather than a bracket estimate.