Cambodia is one of the poorest countries in the world, ranking 131 out of 177 countries on the Human Development Index, with a per capita gross domestic product of US$2727 . An estimated 1.6% of the adult population in Cambodia is living with HIV, a decline from 3% in 1997 . Although the economic and social impacts of AIDS on macroeconomic growth are likely to be modest in concentrated epidemics , an HIV-related illness or death can have devastating effects on economic and social wellbeing at the household and individual levels and can increase household vulnerability to poverty, as confirmed by previous studies in the region [4,5]. A case–comparison study from India released in 2006 looking at more than 8000 households found that medical expenses of HIV-affected households were four times higher than those of non-HIV households. The same study also showed that children in HIV-affected households were more likely to drop out of school .
This paper explores the effects of HIV and AIDS on household economics and the social wellbeing of children in HIV-affected families throughout three provinces in Cambodia. The study was designed to augment the current literature on the impacts of HIV and AIDS in concentrated epidemics, while providing country-specific information to the Royal Government of Cambodia, donor agencies and programmme planners, in order to guide the mitigation response. The primary objectives of the study were to quantify the magnitude of the economic effects of HIV and AIDS on households in Cambodia, identify the survival and coping strategies used in response to HIV-related illness, and explore the effects of HIV and AIDS on the social wellbeing of children living in HIV-affected households. A secondary objective was to provide a baseline measure for a panel assessment that would examine the effects of mitigation interventions on social and economic wellbeing over time, if funding became available.
The study was designed in collaboration with the Royal Government of Cambodia's Ministry of Social Affairs, Veterans and Youth. We used a case–comparison design to assess differences in social and economic wellbeing between HIV-affected households (case households) and a comparison group of ‘nearest neighbours’ (households in closest proximity to case households that met eligibility criteria).
Measuring economic effects of HIV
One of the greatest empirical challenges in measuring the socioeconomic effects of HIV is in distinguishing risk from impact, as HIV and poverty are likely to influence each other, with HIV exacerbating poverty and poverty increasing the risk of HIV infection. We attempted empirically to disentangle this relationship through our study design. By using a nearest-neighbour comparison group, we assumed that the households were similar in socioeconomic status. We also tried to verify this relationship in the results by categorizing socioeconomic variables by the rate at which they are likely to change. For example, basic housing characteristics such as materials, source of water supply and sanitation are unlikely to change quickly over time and are expected to be good indicators of socioeconomic background, and are also used to indicate socioeconomic status in the country's Demographic and Health Survey . In contrast, the possession of household assets such as bicycles, radios and televisions is more likely to change relatively quickly over time when households face an economic shock. It is helpful to think of the household characteristics as ‘fixed assets’ and the household assets as ‘variable assets’. Similarities on ‘fixed’ assets would thus lend support to the claim that households originated from a similar socioeconomic background, even if they differed on ‘variable’ assets. In addition, the inclusion of items related to retrospective changes in economic indicators provided some insight on the impact of HIV-related illnesses on socioeconomic status over time.
Selection of households
Five hundred pairs of households were selected to participate in the study, totalling 1000 households interviewed. Each pair included a case household, defined as households in which a parent is living with HIV and a comparison household, defined as households in closest proximity to case households that met eligibility criteria. All households included at least one child between the ages of 6 and 17 years, inclusive. A purposive sample of case households was selected from networks of people living with HIV and AIDS. The country's low HIV prevalence meant a random sample was not feasible. Individuals in the HIV-affected population were preselected from a list of more than 8000 network members, using a random numbers table, and contacted by network representatives. If eligible, they were asked to participate either by phone or at network meetings. After giving informed consent, case households were added to the list of participating households and their address was provided to the research team.
Comparison households were selected on the day of the study by identifying the residence nearest the case household. While in urban areas, the ‘nearest’ neighbour may live quite close to the case household, in some rural areas the ‘nearest’ neighbour lived some distance away. The distance between the pairs of households was not of particular importance but the methodology for selecting the household was important. Interviewers were trained to identify the nearest household by eye. Two interviewers were required to reach consensus regarding the house(s) of closest proximity. If households were judged to be of equidistance apart, the interviewer selected a number of coins covered in paper (equal to the number of houses equidistant from the case household) and assigned each household a number. The interviewer then randomly selected one of the coins from a bag, and approached the corresponding house. In the event that the first house selected did not meet eligibility criteria (did not have children between 6 and 17 years), the interviewer drew another coin and approached the next household. The response rate was very high, only 2% of eligible case households and fewer than 2% of eligible comparison households declined to participate.
The HIV status of the adult in the comparison household was not obtained, although questions about cause of death of family members do provide insight into the comparison group's HIV status. We expected the prevalence in the comparison group to be similar to the prevalence in the rest of Cambodia, approximately 1.6%. (The analysis found that 1.4% of the comparison households reported having lost a family member to HIV-related illnesses.) These seven out of 500 HIV-affected households in the comparison group were not excluded from the analysis and it is expected that any selection bias would only underestimate the differences in economic and social impacts between case and comparison households.
Sample size, selection and location
When possible, the researchers interviewed one parent, one child (aged 6–12 years) and one adolescent (aged 13–17 years) in every household. The stratification of age groups ensured a large enough sample of both groups of children, allowing us to understand better the age-specific effects of HIV. If, however, only a child from one of the two age groups was present, only one child was interviewed. When more than one child in each age group was living in the household, one name was selected in a blinded manner from a paper bag to identify the child or adolescent to be interviewed. Across the 1000 households, the sample included 1000 adults, 856 children and 587 adolescents. Fifty per cent of the interviews were conducted in urban areas of Phnom Penh, 40% in rural areas of Battambang and 10% in rural areas of Takeo province.
Only one parent per household was selected to participate in the study; therefore, in cases of households in which more than one parent was HIV-positive, the study included only the individual selected by the organization. If more than one adult (with a child between the ages of 6 and 17 years) in the comparison household was available, the adult was selected randomly by drawing a number from a bag. Although some parent interviewees may not be ‘heads of households’ (some may be spouses or relatives of heads of households), all adult interviewees were parents of children living in the households and for this reason we believed they were in a good position to answer questions about the household, themselves and their children. Interviews were conducted between October 2003 and January 2004.
Survey instrument design
Survey instruments specific to each of the three groups of interviewees were designed in English, translated into Khmer, and back-translated into English. The adult survey included modules on socioeconomic status, survival and coping strategies and exposure to interventions. The child and adolescent surveys included modules on school enrollment, nutrition, health, employment status, household responsibilities, nutrition and health status. Stigma and discrimination and quality of life indices were also included (results available from the authors). The items in the survey instruments were obtained or adapted from the Demographic and Health Survey in Cambodia, World Bank Living Standards Surveys, a survey developed by the USAID Community Reach project, and the KINDL Quality of Life Index . A local research team received extensive training and conducted a pilot test before collecting data. Revisions to the questionnaires were made after the pilot test.
Interviewer training, ethical review and confidentiality
A local internal review board and a child psychologist in Cambodia approved the research protocol and survey instruments. Interviewers were instructed to obtain informed consent before administering the questionnaires. To maintain the confidentiality of HIV-positive respondents, questions related to HIV were asked only to the adult in the case population. We were aware that some family members may not have disclosed their status and we retained this confidentiality. Unique identifiers were entered into the data entry system so that no match could be made between individuals and results.
Data were analysed using SPSS 11.5 statistical software (SPSS Inc., Chicago, Illinois, USA). The overall strategy for the analysis was: (1) an assessment of the comparability of the case and comparison households on the matching variables; (2) a bivariate analysis (including paired t-tests, Wilcoxon signed rank tests, odds ratios and Chi-square tests) to assess the differences between households in the variables associated with HIV expenditure, income effects and child wellbeing; and (3) multiple regression analysis of the relationship between the variables associated with HIV expenditure and income effects and household income (the dependent variable) while controlling for potential confounding variables that influence this relationship.
Background characteristics of the case and comparison households
There was no significant difference in background socioeconomic indicators such as education level and other household characteristics that are expected to stay relatively constant over time (see Table 1). The number of children was also similar in both groups of households. The two groups differed significantly on marital status, however: a significantly higher number of respondents in the case household were widowed/widowers (59.6 versus 14.4%). Correspondingly, the number of adults and adult equivalents in the house was significantly lower among the case households. Respondents in the case households were also significantly younger than the comparison households.
The case households reported receiving an average of 4.3 different support services: 94% received donated medicines or financial assistance for basic medical support; 13% received donated antiretroviral drugs; 42% received education support; 81% received food assistance; 97% received emotional counselling; 15% received small loans; and 3% received vocational training. Eighteen per cent of comparison households received support, on average 1.3 services per household.
Economic effects of HIV and AIDS
The economic effects of HIV and AIDS on households can be categorized as expenditure-related effects (e.g. increased health spending), income-related effects (e.g. loss of income from HIV-positive individual and caregivers), and mitigation effects (e.g. coping strategies that a household may use; savings, external support, and the provision of antiretroviral therapy).
Household expenditure-related effects
The top of Table 2 illustrates the expenditure-related effects of HIV and AIDS. Although total expenditures did not differ significantly, the proportion directed to medical care was 17.3% among case households and 8.4% among comparison households. For case households, the proportion of expenditures used for medical care was 20.8% among households in which a death had occurred in the 12 months preceding the interview, whereas in case households in which no death had occurred, the proportion of expenditures used for medical care was 16.5%.
HIV-affected families dealt with funerals more frequently. In the year preceding the survey, 20% of the case households experienced a death, compared with 5.6% of the comparison households.
Household income-related effects
The second half of Table 2 shows income-related effects of HIV and AIDS. Reported monthly household income was 48% lower among case households. Although more case households were headed by single caregivers, income in the case households was not significantly different between married and unmarried households.
The HIV-affected families reported significantly lower employment rates and among, those working, higher absenteeism rates than their neighbours. Similarly, significantly more case households reported a decrease in income (77%) relative to the comparison households (54%).
The results of a multiple linear regression in Table 3 illustrate the factors affecting income. Identifying as HIV positive, having fewer adult equivalents in the household, living in a rural location, having low levels of education, and an increased age of the parent participating in the interview, were all factors associated with a significant reduction in household income. Marital status and death of a family member, however, were not significant predictors of household income.
Table 4 illustrates that the HIV-affected households were more likely to ration their expenditures on healthcare, food, children and leisure activities, sell off household assets, reduce savings, and receive donations from family members or take out a loan. This rationing, selling and borrowing probably contributed to the difference in Table 2 showing that case households had fewer assets than comparison households at the time of the interview.
In the case households, 55% of the children were single orphans (55% paternal orphans; 5% maternal orphans), 2% were double orphans and 38% had both parents still living, but at least one parent living with HIV. In the comparison households, 14% were single orphans (12% paternal; 2% maternal), 2% were double orphans and 84% had both parents alive. The HIV status of the children was not revealed, but the children in the case population were not significantly more likely to be chronically ill than their comparison peers (see Table 5).
Both children and adolescents in HIV-affected households reported more frequent hunger than their peers in comparison households (P < 0.001) and had fewer meals in a day (P < 0.001). There was no significant difference between case and comparison ‘ever attendance’ in school or current enrollment (see Table 5). Nor was the age of attending level one significantly different across groups (7.61 years in the case population versus 7.50 years in the comparison population, P = 0.813).
In both age groups, children in case households were more likely to work for money, and among the older age group, more likely to take on an increased level of responsibility, such as cooking, looking after younger children and caring for sick family members. Children and adolescents in HIV-affected households were also shown to have a lower measure of quality of life, the details of which were presented in a previous paper .
The similarities in socioeconomic and demographic variables, such as education and household characteristics, suggest the nearest-neighbour approach provided a reliable comparison group with respect to past socioeconomic background. Similarities on ‘fixed’ socioeconomic variables imply that the pairs of households originated from a similar socioeconomic background, and therefore the comparison households were a reasonable proxy for what the case population would look like in the absence of HIV. Although the households differed on other demographic variables (marital status, and correspondingly, number of adults), these differences are related to the impact of HIV and should not suggest that the populations were inherently different. The results also show that the case households were younger than the comparison households, with a higher concentration of people in their most productive years. We cannot confirm the reasons for this difference but speculate that it may be caused by the fact that HIV prevalence is higher among people in younger, more productive age groups in Cambodia.
It is evident from reported household expenditures and income levels that the households in both the case and comparison are among the poor in Cambodia. Both groups hover around the national poverty line; 35% of households live below it . The survey results also show that reported expenditure was greater than income in both case and comparison groups, suggesting that this population is forced to live beyond its economic means. This situation is especially true for those households affected by HIV faced with catastrophic expenditures such as medical care and funerals. It is important to note a limitation in interpreting these results: it is well established that income is underreported in surveys, and total expenditure is a much more reliable measure of household economics than self-reported income.
Despite similarities in overall expenditure, the composition of expenditures was significantly different, as a large percentage of case households' expenditures went to medical care and funerals. As outlined in the Cambodia Participatory Poverty Assessment, medical care is often the crisis that drives households into poverty . This is especially true among the case households. This is because of the higher incidence of illness, but results show that even among a subgroup of households who experienced a death in the past 12 months, medical expenditures were higher among the case population, indicating that an HIV-related death incurs higher costs on medical expenditures than deaths from other illnesses or accidents. The results on medical expenditures call into question how well medical assistance offered through non-governmental organizations mitigates the impact on household economics. Almost all case households received free medical assistance, but a substantial portion of out-of-pocket expenditures still went to medical care. Moreover, this medical assistance excluded antiretroviral treatment, the costs of which were absorbed by donors. Medical expenditures were also high in the comparison households, which was not surprising given that healthcare in Cambodia consists of a poorly paid public sector in which unofficial fees are often requested, medicines are not regularly stocked, follow-up of patients is sporadic, and staff motivation levels are low (personal communication with non-governmental organization and home-based care organizations in Cambodia, October 2004). The problems with the healthcare system are, however, particularly acute for people living with HIV . A further drain on the system will occur if donors stop providing free antiretroviral treatment, the costs of which are much higher in Cambodia than in the region and almost four times more than in neighbouring Thailand . These findings reveal the need to accelerate capacity-building within the healthcare system as a strategy for addressing the country's AIDS epidemic.
Findings on income-related impacts indicated that significantly lower earning power in case households was caused by higher unemployment rates, higher absenteeism rates and changes in wage income. These results are consistent with other studies from the region [5,6,14,15] as well as eastern Africa  and southern Africa , which show the drop in income and increase in expenditure when a family member becomes ill. Although the case households were more likely to have lost a spouse, marital status was not a significant factor associated with household income, nor was the death of a family member. Case households, older parent interviewees, households in rural areas, and households with less educated parents and fewer household equivalents were, however, associated with lower income levels. The fact that the loss of a family member was not associated with lower income, and having fewer adult equivalent household members was associated with lower income, suggests that living in larger extended family networks may help mitigate the income effects resulting from the death of a family member.
This study confirms that dissaving is associated with HIV-related illnesses and investment in future generations is lacking. Households responded to crises by decreasing expenditures on other household members, spending less on food, medical care and leisure activities, borrowing from household and community members and reverting to more irreversible strategies such as disposing of productive assets. Within the sample, however, these assets are probably too small to play a major role in managing loss, as was found in an earlier study from India . Numerous other studies reveal that these survival strategies come at the expense of longer-term investments in the household [6,19,20].
This study sheds further light on the impacts of HIV-related illnesses on children. Many of the affected children in this survey had lost at least one parent and were under the care of a parent who was sick or dying at the time of the interview. The effect of HIV and AIDS occurs in a number of overlapping and interdependent domains, but this study shows that children in HIV-affected households are significantly disadvantaged in a number of ways compared with their neighbouring peers. They are more likely than their peers to eat fewer meals and experience hunger more frequently, even though 81% of the case households received food assistance.
Although the children in the study's HIV-affected households were no less likely to attend school, these results should be interpreted with caution. Qualitative data collected as part of this study suggest that children and adolescents affected by HIV and AIDS may experience a higher level of disruption in education, which the data on enrollment may not fully reflect. In Cambodia, enrollment in school is free, so all children could technically enroll but not attend regularly. In the study from India by Pradhan et al. , children in HIV-affected households not only had a lower rate of enrollment than those from unaffected households, but the dropout rates were higher and school attendance was lower for those who had not dropped out. Girls were found to be more affected, being more likely to be withdrawn from school . Future studies on educational outcomes should look beyond ‘enrollment’ to understand the education dynamics.
The high percentage of children and adolescents engaged in household tasks is not surprising, as housework seems to be standard among children in Cambodia. The data showed, however, that adolescents living in HIV-affected households were more likely to cook, care for small children and look after sick household members compared with the comparison adolescents. Moreover, significantly more children and adolescents in the case households reported that they were working for money, compared with their peers. These results are consistent with other studies showing that responsibilities and work for children, both within and outside the household, increase dramatically with the illness or death of one or both parents [9,21].
Some important limitations of our study should be noted. First, although the nearest-neighbour approach was successful in identifying pairs of households that were similar on socioeconomic indicators, the pairs were not truly ‘matched’, they differed on age, and many women interviewees in the case households may have been heads of households, whereas women interviewees in the comparison households may have been the wives of heads of households. They also could differ on other characteristics not measured. (The groups also differed on marital status and household size but this was a factor related to the impact of HIV/AIDS, rather than the socioeconomic background.) Second, much of the data used to identify changes in socioeconomic status were based on self-reported current and retrospective information. A more reliable method of determining temporal causation would be to follow populations over time. Finally, the purposive sample of case households selected through networks of people living with HIV may not be reflective of the general population of people living with HIV. That this group is poor, open about their status, and seeking services may make them inherently different from the rest of the population living with HIV.
The findings help stakeholders understand the links between HIV and poverty. Similarities on long-term socioeconomic indicators suggest that the groups were once of the same socioeconomic status but that as a result of HIV/AIDS the case population is poorer because of the expenditure and income-related effects presented above. The study does not quantify the long-term social and economic costs on a society that is forced to reduce its investment in the nutrition and wellbeing of adolescents and children. The data imply that economically and socially, children and adolescents in HIV-affected households are worse off than their peers. As illnesses progress, the burden on children, households, communities, and support services will probably increase dramatically.
It is clear that the services to HIV-affected households interviewed in this study stopped short of mitigating the economic and social impact. The extent to which programmes have mitigated the impact could, however, be confirmed in a follow-up survey with the same population.
These survey results have been particularly useful to policy makers and programme managers in Cambodia, and were first disseminated with findings from a study by Care International, to help develop a strategic planning process for orphans and vulnerable children and facilitate discussions about best practices. The study has thus already contributed to evidence-informed planning of policies and programmes. Although impact assessments are useful, however, it is time to question the value of conducting additional impact assessments in Asia at this stage of the epidemic when a wealth of evidence now confirms consistent results. Would it not be more useful to allocate resources to implementing, and evaluating, mitigation strategies? There is a dearth in information about how to mitigate the impact of HIV/AIDS through comprehensive strategies that improve support networks, reduce stigma and discrimination, improve gender equality, and prepare children for the death of one or both parents through succession planning. Such evidence will be instrumental in carving the way for policy makers and programme managers as they design multifaceted responses that truly mitigate the devastating impacts of HIV/AIDS that have already been so well documented.
The authors express their gratitude to the following individuals and organizations: the guardians and children who generously gave their time to participate in this study; the Ministry of Social Affairs, Veterans and Youth Rehabilitation in Cambodia, for partnering with the POLICY Project in the design and dissemination of the study findings, and the Center for Advanced Study (Phnom Penh, Cambodia) for their participation in the data collection process.
Sponsorship: Funding for this study was provided through the POLICY Project by the United States Agency for International Development (USAID) under contract HRN-C-00-00-00006-00. The POLICY Project was implemented by the Futures Group (now Constella Futures) in collaboration with the Centre for Development and Population Activities (CEDPA) and Research Triangle Institute (RTI).
Disclaimer: The authors' views expressed in this publication do not necessarily reflect the views of USAID, the US Government, or Constella Futures.
Conflicts of interest: None.
3. UNAIDS. Report on the global AIDS epidemic
. Geneva: UNAIDS; 2006.
4. UNAIDS, Asian Development Bank. Impact of HIV/AIDS on poverty in Cambodia, India, Thailand and Viet Nam
. Background paper IV. UNAIDS: Bangkok; 2004.
5. Pitayanon S, Kongsin S, Janjaroen WS. The economic impact of HIV/AIDS mortality on households in Thailand. In: Bloom D, Godwin P, editors. The economics of HIV and AIDS: the case of south and south east Asia. New York: Oxford University Press; 1997.
6. Pradhan BK, Sundar R, Singh SK. Socio-economic impact of HIV and AIDS in India.
New Delhi: National AIDS Control Organisation, United National Development Programme and National Council of Applied Economic Research; 2006.
7. National Institute of Statistics. Cambodia demographic and health survey 2000.
Phnom Penh: ORC Macro; 2001.
8. Ravens-Sieberer U. The revised KINDL-R: final results on reliability, validity and responsiveness of a modular HRQOL instrument for children and adolescents. Qual Life Res 2001; 10:199.
9. Alkenbrack S, Chettra T, Forsythe S. The social and economic impact of HIV/AIDS on families with adolescents and children in Cambodia. Findings from a research study in Phnom Penh, Battambang, and Takeo. POLICY Project; 2004. Available at: http://www.policyproject.com/pubs/countryreports/CAM_OVC_En.pdf
. Accessed: June 2007.
10. World Bank. Cambodia: halving poverty by 2015? Poverty assessment 2006.
Washington, DC: World Bank; 2006.
11. Asian Development Bank. Participatory poverty assessment: Cambodia
. Manila: Asian Development Bank; 2001.
12. Dhaliwal M, Ellman T. Improving access to anti-retroviral treatment in Cambodia. Phnom Penh: KHANA, International HIV/AIDS Alliance; 2003.
13. United Nations Development Program. HIV/AIDS and poverty reduction strategies– policy note
. New York: UNDP; 2002.
14. Gertler P, Gruber J. Insuring consumption against illness. The American Economic Review 2002; 92:51–70.
15. Janjaroen W. The impact of AIDS on household composition and consumption in Thailand.
In: Ainsworth, M., Fransen, L., Over, M., editors. Confronting AIDS: evidence from the developing world
. Selected background papers for the World Bank. Brussels: European Commission and Washington, DC: World Bank; 1998.
16. Menon R, Wawer M, Konde-Lule J, Sewankambo N, LiC. The economic impact of adult mortality on households in Rakai District, Uganda.
In: Ainsworth, M., Fransen, L., Over, M., editors. Confronting AIDS: evidence from the developing world
. Selected background papers for the World Bank. Brussels: European Commission and Washington, DC: World Bank; 1998.
17. Booysen F. Poverty dynamics and HIV/AIDS-related morbidity and mortality in South Africa.
Paper presented at The Scientific Meeting on the Empirical Evidence for the Demographic and Socio-Economic impact of AIDS
. Durban, South Africa, 26–28 March 2003.
18. Basu AM, Gupta DV, Krishna G. The household impact of adult morbidity and mortality: some implications of the potential epidemic of AIDS in India. In: Bloom D, Godwin P, editors. The economics of HV and AIDS: the case of south and south east Asia. New York: Oxford University Press; 1997.
19. Naidu V, Harris G. The impact of HIV/AIDS morbidity and mortality on households– a review of household studies. The South African Journal of Economics 2005; 73:533–544.
20. Van Damme W, Van Leemput L, Por I, Hardeman W, Meessen B. Out-of-pocket health expenditure and debt in poor households: evidence from Cambodia. Trop Med Int Health 2004; 9:273–280.
21. Ainsworth M, Filmer D. Poverty, AIDS, and children's schooling: a targeting dilemma
. Policy Research Working Paper 2885. Washington, DC: The World Bank; 2002.
© 2008 Lippincott Williams & Wilkins, Inc.