Mitigating the impact of the epidemic on the households and families of older people in rural Uganda : lessons for social protection

Using quantitative and qualitative data drawn from a population cohort of 20,000 people in South-West Uganda we explored factors (household size/dependency ratios/ residence patterns/ socio-economic status) which may have exacerbated or eased the impact of the AIDS epidemic for older people (70 years and older). We found that higher socio-economic status, large family size and reciprocal relationships with kin/neighbours were particularly important in shielding older people from the negative effects of AIDS-related deaths among their children; elders without these fared badly. Our findings provide pointers for the targeting of social protection for older people in such resource-constrained settings.


Introduction
Madina is over 80 years old.She is bent nearly double and has trouble walking.She is very thin and she wheezes constantly because of what she says is asthma . . 1 Madina lives in a small mud hut, with a grass roof.Her grandson lives on the same compound in another, similar, house with his new wife and their baby.During our visit Madina tells us that she wants to visit her cousin who lives 25 kilometres from her village.The local taxi driver knows where her cousin lives and will take her there for 5000/-(£2).We give her the money she needs and discuss the idea of making a telephone call from the telephone stall in the trading centre to make the arrangement with the cousin, but at this point her grandson arrives.We tell him what Madina wants.His response is to smile and tell her `you don't need to go for that visit, you are all right here, your aches and pains are just because you are old', his words sound well-rehearsed.This is a conversation they have had many times before.Madina holds the money in her hand and looks at us, defeated, because she knows her grandson is not going to help her and if we try and do anything more he will be angry with her.She leans back against the wall of her hut and looks into the distance.
Madina's daughter, the boy's mother, died of AIDS-related illnesses in 2007.The daughter had lived in the same village as Madina and her mother had often told us that she was the one she relied on, the one she had expected to be there for her in her old age; sadly, that was not to be.
The impact of the loss of adult children, as a result of AIDS-related illness, on older people's own care and general well-being as they age throws into relief the vulnerability of many older people (defined here as those over 70 years of age) in resource-constrained settings (Williams and Tumwekwase 2001, Dayton and Ainsworth, 2004and Seeley et al. 2009).There is increasing concern over dwindling support for older people because `material family support' is not available as family members move away for work or to settle in towns and cities (Aboderin 2004a, Lloyd-Sherlock 2004); the effects of the AIDS epidemic add to this loss of potential carers.In this paper we examine the factors that have exacerbated or eased the impact of AIDS for older women and men in rural Uganda as they grow frail and ask what lessons this may hold for social protection policy and practice.We use quantitative methods to generate hypotheses about the living conditions of older people, and then investigate more deeply these hypotheses using our findings from ethnographic research.

Literature review and theoretical framework
Available evidence suggests that the family continues to be the main source of support for older people in resourceconstrained settings (Heslop and Gorman 2002, Aboderin 2004b, Schröder-Butterfill 2004), a pattern that has persisted despite changes increasingly brought about by so-called westernised life-styles in many places.Other factors are also important: older people's access to paid work, basic services and social networks are all important for their wellbeing (Barrientos 2003 et al.).
Income from other sources which older people may bring to a household may also have an important impact on the way they are treated.In places such as South Africa where a means tested social pension is available older people make an important financial contribution to their households (Barrientos et al. 2003).In other places, where pensions are not available older people often make important in-kind contributions to the productive and reproduc-tive activities of the households in which they live (Gorman andHeslop 2002, Schröder-Butterfill 2004).Edmonds et al. (2003) comment that the literature on the impact of pensions in the developed world stresses how that income enables older people to maintain their independence, whereas in the developing country setting a pension or some other form of income is used to 'alter living arrangements'; grandchildren or others may come to stay because there is a source of support.Camarano (2004) found a similar situation in Brazil, where there was an increase in household size for those where an older person was in receipt of a pension.However, Burman (1996) reports on a negative outcome of receiving a pension, she found that some old age pensioners in South Africa were being forcibly relieved of their pensions by family members.These findings corroborate Lloyd-Sherlock's (2006: 992) observation that the impact of social pensions and cash transfers on older people may result in 'complex potential effects' and is highly variable, so it difficult to make generalisations about their effect.
There are other factors which affect the support older people receive.The gender of the older person may influence not only their role but also their residence options.In urban Mexico Varley and Blasco (2001) report that older men were less likely to be welcomed into the homes of their children in later life than older women.This was because the older women were perceived to be useful in the home and in taking care of children; roles it was believed men could not perform.This contrasts with the situation described by Sagner (2001:177) for the Xhosa in South Africa in the nineteenth century where: ... it was mainly the women who found themselves destitute in old age.Since women held, irrespective of age, a marginal status within the political economy and cosmology, they had far fewer opportunities to translate their culturally prescribed prominent old age status into action than men.While times may have changed, and older women's role as carers of the sick and children is now often greatly valued (HelpAge International 2003, Schatz andOgunmefun 2007) in such patrilineal societies, where children (particularly sons) inherit from their fathers, the fate of an older woman may depend upon her having her own children to provide care.She may be dependent on the availability, goodwill and resources of her own children or natal kin, since the children of co-wives and other relatives of her husband may not be ready to give support when she is widowed (Kirwen 1979, Potash 1986), particularly if they have their own elderly mothers to support.
If co-residence with close kin is important for older people's wellbeing, as the literature implies, it would be important to know how many older people live alone who may be vulnerable to neglect.Zimmer and Dayton (2005) in their review of Demographic and Health Surveys of 24 Sub-Saharan countries looked at the household composition of households containing older people (aged 60 and over) and found that 59 percent lived with a child and 46 per cent with a grandchild.Only five percent of older adults lived alone (2005: 297).John Iliffe (1987) in his seminal work on the poor in Africa makes a number of references to the idea that poverty in Africa can be equated with not having kin to provide assistance, adding support to the concern that those older people who live alone are among the significant number of older people in the developing world living in poverty (Barrientos et al. 2003: 567).However, Schröder-Butterfill (2004: 498) cautions against taking living arrangements as indicators of the welfare and support of older people, suggesting that it is necessary to look at `exchanges within and beyond households' to understand the support that older people may receive, as well as what they may give.This observation is in keeping with our previous observations on the impact of HIV on a rural population in general (Seeley et al. 2008) and is a factor in the mitigation of the impact of the epidemic on older people that we explore in this paper.
The findings presented in this paper are from research in a rural sub-county in Masaka District, Uganda.The people living in the area are largely subsistence farmers who produce small amounts of cash crops such as bananas, beans and coffee.The majority of the population are ethnically Baganda (75 per cent), but there is a large representation of immigrants from Rwanda (15 per cent).Four per cent of the population are immigrants from Tanzania.The community is predominantly Christian (Roman Catholic 58 per cent and Protestant 12%), with a substantial minority of Muslims (28 per cent).Just over 50 per cent of the population is under 15 and three per cent are over 70 years of age.The large majority of households occupy less than five acres of land but few are landless.Zimmer and Dayton (2005) remark on the importance for care and support of an older adult living with an adult child 'in sub-Saharan Africa'; a generalisation that does not, however, appear to hold true for the Baganda.Descent among the Baganda is patrilineal, with all children belonging to their father's clan (Roscoe 1965: 128).However, the system of descent does not mean that a father and his children, once grown, lived together.As Nahemow (1979: 172) has observed, 'While they have a patrilineal kinship system, [the Baganda] are nuclear in their households and generations are often residentially segregated by considerable distances.This was the traditional pattern and is still true today.'She attributes this pattern to the availability of fertile land over which settlements could spread.Most ownership involves customary land tenure agreements (kibanja) made with a small number of large land owners which has allowed young people to gain access to land independent of parental holdings.The advent of the AIDS epidemic 25 years ago may, therefore, have increased existing variability rather than fundamentally altered co-residence patterns among kinship groups.
The MRC/UVRI has studied the progress of HIV and AIDS in a cohort of about 10,000 individuals living in 15 villages since 1989 and an additional 10,000, living in 10 neighbouring villages since 2000.Cohort participants are entitled to free medical care in the MRC/UVRI clinic.Research in the cohort found HIV prevalence in the adult population (age 13+) of the origi-nal 15 villages to be 8.5 per cent in the annual survey round 1990/1991, 7.0 per cent in 1998/1999 and 7.7 per cent in 2004/2005.In the 10 'new' villages, which include the administrative and trading centre of the sub-county, prevalence rose steadily from 4.4 per cent in the 1999/2000 survey round to 8.6 per cent in the 2004/2005 survey round (Shafer et al. 2008).Overall incidence rates of HIV among males and females over 60 years in the MRC/UVRI cohort from 1990-2007 was 0.9 per 1000 person years (1.5 per 1000 person years in males, and 0.5 per 1000 person years in females).

Data and methods
Annual demographic, medical and serological surveys have been carried out with the cohort since 1989 (Mulder et al. 1994).Information on household assets is collected every four years, including the first year (1989/1990) and in 2005/2006.Members of the cohort also take part in other small studies from time to time, including the ethnographic study on the impact of HIV and AIDS on daily life carried out with 27 household from three of the cohort villages in 1991/1992 and repeated with 24 of those households 15 years later (2005/2006) from which the story of Madina at the beginning of this paper, and the qualitative data presented below are drawn.The results from this paper are drawn from two time points, one at the beginning of the study involving households visited in 1989/1990, and ethnographic research in 1991/92, and the second time point at 2005/ 2006, when demographic, socio-economic and ethnographic data were collected.
We focused our analysis on households with a person aged 70 years and older.We chose this age group because earlier research in this population (Seeley et al. 2009) showed that people 70 and older faced greater difficulties in terms of care and support because of frailty and ill-health than younger older people (aged 50 to 69 years). 2  Marital status was obtained from information collected during the annual surveys on household relationships, as was the presence of children, and adolescents.A household is considered to be laterally extended if it includes siblings or other family member residents of the same age group as the household head who are not the household head's spouse.A household is considered to be vertically extended when it includes more than two generations (for example grandparents and grandchildren).
In 1989/1990 data were collected on ownership of the following household assets: jerry-can, pot for keeping water, place for drying plates, bed, sewing machine, weighing scale, radio, bicycle, motorcycle and motor car.In 2005/ 2006 data were collected on the ownership of a different set of assets: pot for boiled water, breakable cups, bed, radio, bicycle, motorcycle and motor car.At each of the two time points a Socio-economic status (SES) index was calculated as a weighted sum of the assets in a particular household.The weights were computed by carrying out a separate Principal Components Analysis (PCA) for each time point on the set of 0/1 indicator variables denoting absence or presence of each item.The PCA chooses the weights in order to maximize the variability of SES between households; in practice this leads to smaller weights being allocated to relatively common items (such as a pot for keeping water) and larger weights being allocated to relatively rare items (such as a motor car).
Linear regression was used to compare the mean SES index between households with older people and households without older people, and between different groups within households with older people.Poverty in households with and without older people was compared using a chisquared test of the difference in the proportion of poor households in the two groups.A household was regarded as poor if it had an SES index less than or equal to the maximum index of households that owned only one asset.
For the ethnographic study the original 27 households were chosen purposively in 1990 to represent a cross-section of different household types (by sex and age of household head, as well as socio-economic status).The socio-economic rankings (based at that time on a simple sum of assets owned and an assessment of house type and condition) of the 27 selected households were cross-checked through the visits to the households.
A team of local people trained in ethnographic research paid monthly visits for one year to the study households assigned to them to record changes in different aspects of the household, such as composition, employment, health, food consumption and social networks.In 2006 the study was repeated collecting the same detailed information on day to day life, and also changes in socio-economic status.Information was also collected on household members' memories of what has happened in the intervening 15 years.Members of 24 of the 27 original households participated in the restudy. 3Life histories of all adult members were collected.The analysis of the data was done manually using thematic content analysis focusing particularly on information on demographic change, asset sales and acquisitions as well as the situation of older people in the study households.
Overall approval for the study was given by the Ugandan National Council of Science and Technology.Ethical approval was given by the review boards of the Uganda Virus Research Institute and the University of East Anglia.

Results
In 1989/1990, out of a total population of 10,372 people, 310 (3.0%) (144 men and 166 women) were aged 70 years and over living in a total of 274 households defined as households with an older person.Twenty one of these households had incomplete data on assets and were excluded from further analysis.Out of the 253 households analysed, 147 were headed by men and 106 by women.
In the second period (2005/2006), there were 541 (2.7%) people (247 men and 294 women) aged 70 years and over from a population of 19,999 people. 4These older persons were living in 480 households although three did not have complete data on household assets and were excluded from further analysis.Out of the 477 households analysed, 250 were headed by men and 227 by women.
In 1989/1990 the mean SES index for all the 1,757 households in the gen-eral population was 0.965 (SD = 0.015) and 449(25.6%) of them were poor.However, households with an older person had a mean SES index of 0.881 (SD = 0.589) which was significantly lower compared to 0.979 (SD=0.616)for households without older people (t-test=2.37,p=0.018).The proportion of poor households was 29.6% (75/ 253) among households with an older person which was not significantly different from 24.9% (374/1,504) among households without older people (Chisquared test=2.598,p=0.107).
In 2005/2006 the mean SES index for the general population of 3,406 households was 1.369 (SD = 0.569) and 674 (19.8%) were classified as poor.Households with an older person had a mean SES index of 1.227 (SD = 0.596) which was again significantly lower than 1.392 (SD=0.562)for households without older people (t-test=5.899,p<0.001).In this period the proportion of poor households was significantly higher among households with an older person than among households without older people, 30.2% (144/477) versus 18.1% (530/ 2929) (Chi-squared test=37.98,p<0.001) (see Table 1).2) show that the average difference in SES between households including an older persons with more than five people and those with no more than five people is 0.247, which is slightly more than the average difference between households with one asset and households with two assets in the general population.
Also, households including older people headed by a married person had a higher mean SES index, 1.063 (SD=0.592)versus 0.739 (SD=0.549)for households including an older person headed by an unmarried person (t-test=4.504p<0.001).Table 2 shows that the average difference between these households is 0.204, which is slightly lower than the average difference between households with one asset and households with two assets in the general population.Other factors such as sex of household head, number of older people in a household, number of children below 10 years, number of children below 18 years, relation of older person to household head and lateral and vertical extension were not independently associated with SES in households including an older person in 1989/1990.3 show that the average difference in SES between households with an older person with more than five people and those with no more than five people is 0.204, which is slightly less than one and a half times the average difference between households with one asset and households with two assets in the general population.Households with an older person including adolescents under 18 years had a higher mean SES compared to those without, that is, 1.392 (SD=0.525)versus 0.738 (SD=0.524)(t-test=11.828,p<0.001).Table 3 shows that the average SES of households with an older person with adolescents under 18 years is higher by 0.421, that is, about two and a half times the difference in SES between households with one asset and households with two assets.Also, households with an older person headed by married people had a higher mean SES index, 1.518 (SD=0.526)versus 1.025 (SD=0.557)for households with an older person headed by unmarried people (t-test=9.744p<0.001).The findings in Table 3 show that the average difference between these households is 0.271, which is slightly over one and a half times the average difference between households with one asset and households with two assets in the general population.Similarly, laterally extended households with an older person had a higher mean SES than non extended households with an older person, 1.489 (SD=0.519)versus 1.173 (SD=0.596)t-test=4.459,p<0.001.The multiple regression results in Table 3 show that the difference in average SES between laterally extended and non extended households with an older person is 0.131, that is slightly less than the average difference between households with one asset and households with two assets in the general population.
Our findings from the quantitative data provide information about associations between some household characteristics and socio-economic status which paint a picture of the juxtaposition of lower SES for households with older people that are smaller and have fewer younger members than better off households .However, given Schröder-Butterfill's (2004) words of caution about the importance of support external to the older person's household in Indonesia, where she did her research, and our own previous research on household size and poverty (Seeley et al. 2008) for the study population as a whole, can just size and associated relative prosperity be taken as indicators for the wellbeing of older people?We now turn to the qualitative data to explore this question.
Among the 27 households in 1989/ 1990, 14 had a total of 16 members aged over 60, who, had they all lived, would have been aged over 70 during the restudy in 2006.Nine of these people had died by 2006.One person, aged 80, moved into a study household in 2002.We began this paper by describing the living arrangements of one of these older people, Madina, we now describe the living and care arrangements of four others, selected to show the different situations in which older people, among even our small number of households, were living.
In 1991 Grace, then aged 56, was household head living with her 85 year old mother, Marita, and her 54 year old sister Eunice.In addition, two children of Eunice (a son aged 17 and a daughter aged 12) and four of her grandchildren (grandsons aged 13, 13, and three and granddaughter of seven years) and a six year old granddaughter of the household head's brother were living with them.Grace had been married twice, both marriages had failed and she had moved back to live with her mother, Marita.She had had no children.Eunice had seven surviving children.By 2006, Grace and Eunice's mother had died but their mother's sister, who was in need of care, had moved in with them. 5In addition, three nieces and two nephews (children of a brother who had died) and two of Eunice's grandsons lived with them.They lived on land that had belonged to their mother in a house of bricks with an iron-sheet roof, which Grace had had built in 2000 (because their old mud and grassthatched house leaked).She had paid for the bricks herself while Eunice's grown-up children paid for the roof and the labour costs.Grace was not categorised as being 'poor' in either 1991 or 2006.Nanono was in her 80s by 2006.In neither 1991 nor 2006 was her socioeconomic status considered to be 'poor'.While she had some of her children resident in her the village in which she had lived in 1991/1992 (her deceased husband's home) it had been decided by her children that she should be moved out of the study area to live near another son and daughter.Unfortunately the son died of AIDS-related illnesses soon after she moved.Two teenage grandchildren, the children of her daughter resident in that area, were sent by that daughter to live with her in the house her son had given her.A seven year old grandson also lived with her.The two teenagers moved away during the course of 2006, having had misunderstandings with their grandmother leaving her with the seven year old.Nanono spent a lot of time alone and became increasingly unwell.She died in late 2007.
In 1991 Namwandu lived with her daughter Betty in 1991.Betty was considered household head.The household was not classified as 'poor' in 1991 but was considered 'poor' by 2006.Namwandu was aged about 80 and Betty about 40 in 1991.They lived with Betty's three daughters (the eldest was called Beatrice) and two sons.Three daughters of Beatrice were also in the household.The house was made of mud and wattle with an iron sheet roof.
It was an old house (it belonged to Namwandu who had been allowed to stay on in it after she was widowed by her husband's kin) and was not in a good state of repair.They had a very small plot of land for cultivation.The MRC/UVRI baseline survey gave the size of their land to be 14 acres, but this was the size of Namwandu's husband's land.After he died the relatives had divided up the land and left both Namwandu and Betty each with a small portion of banana plantation and coffee on poor land that was not very productive.They had borrowed some land from one of Betty's brothers to enable them to grow seasonal crops.Namwandu died in 2000.In 2006 Betty was living with a twenty year old son and her daughter Beatrice (aged 35).Beatrice's seven year old son and two year old granddaughter also lived with them.They lived in the same house and were still borrowing land from Betty's brother to grow crops on because their plantation remained unproductive.They no longer grew coffee.Both Betty and Beatrice are HIV-positive and Betty has lost one son, when still a young child, to AIDS-related illness.
Kiwanuka was aged about 80 by 2006.His socio-economic status was not considered 'poor' in 1991 or 2006, but this seems to have been largely on the basis of his assets (he had a bicycle) because his living conditions were very poor.In 1991 he had lived with his son Victor aged 18.Next door to him lived his wife of that time and the three children she had had with him (and some of her children from an earlier marriage).By 2006 Kiwanuka was living alone in a small hut close to his own land.His own house had collapsed in 1994 and a neighbour had lent him the place in which he lived (which had been left empty when the occupant died).Like his old house it was made of mud and wattle with a grass roof.It was in bad condition.He did not have a kitchen, so he cooked outside.His wife next door died in 1997 and her son from a previous marriage moved into her house with his family.The children she had with Kiwanuka came to live with him when she died but they were all teenagers so they soon left home.Victor, his son from a previous relationship, moved to live with his mother in another district because he was very sick.By 2006, the two boys Kiwanuka had had with his last wife were fishermen living at Lake Victoria and two daughters were also working near the Lake.In 2007 his eldest daughter who had lost her husband because of AIDSrelated illnesses moved to live in the same village (from the shores of Lake Victoria) so that she could access antiretroviral therapy from MRC/UVRI.Kiwanuka still lived alone, but his daughter now lived close by.She helped her father construct a new two roomed brick house and cultivated his land for him.
Among these households only Kiwanuka lives alone.Grace, a childless woman, has benefited not only from the house and land she inherited from her mother but also from the lateral extension of her household: the presence of her slightly younger divorced sister whose adult children now provide help to both their mother and their aunt, as well as their great aunt who is now cared for by her nieces.Madina and Nanono were not so fortunate.Both lived with grandchildren but both complained of loneliness and felt neglected.Both had lost cherished children, who would have been their carers, to AIDS.In terms of living with supportive kin Betty seems better off, but she had had to manage the illnesses associated with HIV in her own life and the loss of her husband and son to AIDS-related illnesses.The poor productivity of her small piece of land had compounded her problems, since she could not support her family with that land she and her daughter continue to struggle to provide support to their family.

Discussion
Our findings from the quantitative analysis of the cohort survey data for both 1989/1990 and 2005/2006 suggested that smaller households with an older person were poorer.However, as White argued in 2002, qualitative data are valuable in providing a more nuanced view of the relationship between household size and poverty.It is apparent from the qualitative findings that small households with small kin networks made up of poor people, such as Betty's and Madina's, suffer because of the lack of labour and support available to them (Harper et al. 2003: 541).However, as Kiwanuka's story illustrates, even the presence of just one person willing to provide resources can make a dramatic difference to the wellbeing of an older person living alone.Kiwanuka's daughter's return, a significant source of support, had changed his life.She worked for him, fed and rehoused him, and ensured he was cared for, yet he still lived alone.
Mobilising social networks provides the potential for emotional or financial support to older people.It is therefore ironic that high dependency ratios and increased household size are identified by Lawson et al. (2006) as important factors in the persistence of poverty in Uganda.Deininger et al. (2005: 324) argue that the addition of foster children to households, again in Uganda, `resulted in significant reductions of per capita consumption, income, and household investment which were more pronounced for the poor'.Our findings from both the quantitative and qualitative data would suggest otherwise.On the basis of the evidence from our study we would argue that such factors may be detrimental for some, but for others the deprivation may be transitory and attributed to the domestic cycle with high dependency ratios translating in later years into a potential labour force and a source of support, both financially and socially as children grow up. 6 There seems little reason to suggest that small family size in the study area, and in other parts of Uganda, is an attractive proposition at the moment when parents look to secure their old age.A large kin network may not be a route to riches in old age, but it may well stop an elderly person being left without support.Nahemow (1979) found that among her sample of 115 elderly Baganda, 18 lived alone but most lived near a relative, most often adult offspring.She comments that `while dissatisfaction [with support and care] is prevalent among elderly Baganda, loneliness is not.[…] their separation is a matter of personal preference and can be viewed as adaptive and consistent with societal norms ' (1979: 182).The information from our research would suggest that that may be less true than before the AIDS epidemic shrank the safety net of available kin for elderly to rely on for economic support, however, many do still have kin nearby.
Madina, whose story began this paper, had anticipated that her grandchildren might move away for work and marriage but she had not expected her daughter, who lived close by in her village, to die so young.Madina' situation mirrors those in Nahemow's study where she found older people to be socially isolated for three reasons: `1) spouse not living 2) residential separation from kin 3) poor health' (ibid).Even though Madina's grandson lived next door, his lack of care for her wellbeing made her feel isolated.This is a finding corroborated in the work of Mba (2007) for the elderly in Ghana and Williams (2003) in a village close to our study site.Kiwanuka had been in that situation before his daughter returned.Her presence addressed his isolation; he no longer tells us that he needs a new wife to take care of him!One of the factors in Dayton and Ainsworth (2004) and Williams and Tumwekwase (2001) assessment of the impact of HIV and AIDS on older people stands out as most important for the people in our study aged over 70 `as dependent old people who are deprived of any support in their old age that their deceased adult children might have provided'.It is particularly so for those, like Madina and Nanono for whom the lack of caring kin exacerbates the losses they have suffered from AIDS and Betty, who has struggled with the care of dependents as well as her own ill health and poverty that interventions, such as a non-contributory cash transfer (often termed `social pension'), could make a significant difference (Devereux 2007).However, as Burman's (1996) work reminds us, there is no guarantee that cash will reach the intended beneficiary, particularly perhaps in circumstances where the older person is seen to be a burden, and just a source of cash rather than being a valued member of the family in her or his own right.Nevertheless, the pension may help guard against neglect.The experience of the introduction of a pension for people over 70 in Lesotho provides evidence that such a scheme can work in resource-constrained settings (Croome and Nyanguru 2007).We note with concern the increasing number of poor households with an older person in relation to the general population in 2005/2006 compared to 1989/1990.In the absence of such cash transfers existing socio-economic status and family size are important in mitigating the impact of the epidemic for older people; the `old age security motive' (Nugent 1985, among many others) for having children persists while State support for older people in Masaka District is lacking. 7

Conclusion
Formal social protection for all those over 60 or 70 years of age, such as oldage pensions, outside the small minority of salary earners who currently benefit from national insurance schemes, is rare in Uganda.While we welcome the introduction of a pilot scheme in six districts in Uganda to provide a monthly stipend to poor families for basic items; 8 we hope that some of this support benefits older, vulnerable, people.Targeted food aid or medical care from local community-based organisations specifically for those most in need, like Madina, may be a better way to provide protection against abuse or neglect, since there is evidence that such resources `may at least guarantee a basic level of care and status within households, and reduce the risk of abuse' (Barrientos and Lloyd-Sherlock 2002: 14).Extended family cohesion remains largely strong, however, and generally provides adequate protection for most elderly in the study area today but many older people, particularly those without caring relatives near at hand, or assets with which to support their households, do live in poverty.Where a child or grandchild resident in the 1991/1992 household had set up an independent household, their new household was included in the restudy if their parent/grandparent's household no longer existed.4. The increase in population size in 2005/2006 is accounted for by the expansion of the cohort in 2000. 5.This aunt had fostered Eunice as a child, a common practice in the study area to provide help to kin without their own children as well as socialise the children themselves.Eunice saw her role as carer now as being something in return for the care she had received as a child.

References
6.These findings echo the words of caution of Lanjouw and Ravallion (1995) and White (2002) over generalisations about the relationship between poverty and household size.7.In October 2008 the Government of Uganda began a pilot scheme to provide cash transfers to poor families in 12 Districts.Poor families receive 18,000/-to 24,000/-a month (£6 to £8) for food, health, education and housing.The scheme is supported through donor funding.Masaka is not one of the pilot districts.8. 'Government to pay the poor monthly allowances' Saturday Monitor 14 th March 2009, page 3.Under this scheme, financed by donor aid, poor households will receive between Shs 18,000/-(£6) to Shs 24,000/-(£8) per month for food, health, education and housing in a pilot scheme in up to twelve districts.

Table 1 :
Mean SES index and proportion of poor households in households including a person 70 years and households without older people

Table 2 :
Factors associated with SES in 253 households including a person 70 years and over in 1989/1990

Table 3 :
Factors associated with SES in 477 households including a person 70 years and over in 2005/2006 [1] Adjusted for household size, whether a person under 18 years lived in the household, marital status of the household head and lateral extension of the household.
). Modernisation and ageing theory revisited: current explanations of recent developing world and historical Western shifts in material family support for older people.Ageing and Society 24: 29-50.Barrientos, A., Gorman, M., & Heslop, A. (2003).Old Age Poverty in Developing Countries and Dependence in Later Life.World Development 31(3): 555-570.Barrientos, A., & Lloyd-Sherlock, P. (2002).Non-Contributory Pensions and Social Protection.Paper for the series `Issues in Social Protection' International Labour Organisation.Available at: http://www.eldis.org/ageing used 50 years and as a general definition of an older person (see http://www.who.int/healthinfo/survey/ageingdefnolder/en/index.htmlaccessed 13th June 2009).3.One household refused to take part in the restudy and for two others none of the former household members was available to take part.