Bias, Not Error: Assessments of the Economic Impact of HIV/AIDS Using Evidence from Micro Studies in Sub-Saharan Africa

Abstract

Economists struggle to understand the macroeconomic impact of HIV/AIDS. To this end, they have constructed macro models that utilize simplified pictures of the working of the economy and then factor in channels by which HIV/AIDS will have an effect. These models have considerable influence on HIV/AIDS policy; however, they do have their critics. Criticisms in the literature have focused on the simplifications in the construction of the economy that seem most misleading. Using micro studies of sub-Saharan Africa as examples, this contribution argues that there are other important simplifications used by models that need to be reconsidered. Rather than a series of unconnected errors in the modeling process, the approaches show pervasive gender bias, which means that many of the impacts of greater female mortality and morbidity in sub-Saharan Africa are ignored. Gender-aware modeling is crucial to improving assessment of the aggregate impact of the pandemic both in sub-Saharan Africa and elsewhere.

Keywords:

• Gender economics
• growth
• HIV/AIDS
• macroeconomics
• reproduction
• JEL Codes: E27, I10, O47

Notes

Countries in southern Africa appear to have the highest prevalence rates, with more than 30 percent of pregnant women testing positive in Botswana, Lesotho, Swaziland, and Namibia (Joint United Nations Programme on HIV/AIDS [UNAIDS] and World Health Organization [WHO] 2004). By the same estimates South Africa also has a very high prevalence rate, and due to its large population, it is home to the largest number of HIV-affected individuals.

This may result because the economy contains the same amount of machinery and land while the number of workers has fallen. In theory, then, production techniques would change to pair each worker with a greater amount of complementary resources and so achieve a higher productivity.

The UN also notes that small percentage point changes lead to large impacts over time, such that an economy whose growth rate is lowered by 1 percentage point per year will in fifteen years be 15 percent smaller than it would have been in the absence of AIDS (2003: 84).

Charles Kenny and David Williams (2001) have questioned the utility of common approaches to growth modeling by pointing to the contradictory results when such models are applied in practice. They argue that the abstract, ahistorical approach to growth modeling cannot capture what is in fact a complex, historically centered process and consequently argue that growth models are not, in fact, useful representations of the real world.

Given the crucial nature of the assumptions about the impact on the labor supply, Sender, Cramer, and Oya's (2005) warning about the lack of information on prevalence by sectors and skill levels becomes worthy of note.

Therese Jefferson and John E. King (2001) note that the terminology and definitions surrounding “reproductive activity” are problematic; however, I eschew the alternative term “caring activities” here for two reasons. First, as Urdang (2006: 166) notes, the term “caring activities” carries a certain mystique that obfuscates the debate. Second, caring activities can be wrongly envisaged as being related only to child or adult personal care, particularly in a discussion of the impact of HIV/AIDS, when in fact reproductive activities include a wide range of care and household maintenance tasks.

The method and rationale of construction of satellite gross household production accounts is discussed in Young (2000) and Debbie Budlender (2004). Among those countries with high HIV/AIDS prevalence rates, only South Africa has constructed such accounts for non-SNA production. Debbie Budlender and Ann Lisbet Brathaug (2002) estimate that for South Africa, non-SNA production activities would be equal in value to between 11 percent and 55 percent of GDP. Using a more limited set of time-use information, Marzia Fontana (2002) estimates that non-SNA production might be valued at 21 percent of GDP in Zambia.

It should also be noted that the treatment by macro models of activities within the SNA production is also problematic. Thus, later in this paper, I discuss how models tend to focus on what has been termed the “formal economy” and often contain drastic simplifications of the activities carried out in other productive sectors.

Because the SNA economic production boundary includes paid care work carried out by employed carers, it should therefore be included in GDP measures, provided data is collected appropriately.

Indeed, some studies of households affected by HIV/AIDS, such as the study by Booysen and Bachmann (2002), select respondents through NGO or government-care organizations, which may lead to a bias against the identification of those relying on private carers.

The models by Arndt (2003) and MacFarlan and Sgherri (2001) make passing reference to the productivity lost by workers due to time taken from work to care for sick relatives. However, this is not discussed in any detail, and no information is given about how this relates to other factors leading to a fall in productivity.

Although the studies by Bell, Devarajan, and Gersbach (2003) and Arndt (2003) do factor in child withdrawal from school, they do not explicitly link this to the need to carry out reproductive activities, focusing instead on the need for children to engage in productive activities.

A similar impact on education prior to a death in the household was also found by Takashi Yamano and T.S. Jayne (2002) in their study of Kenya.

There is some disagreement in the literature about whether girls are likely to be more disadvantaged in terms of educational access than boys following orphanhood. Ainsworth, Beegle, and Koda (2005: 417) report that two large cross-country studies, surveying a total of thirty-eight countries worldwide, find no evidence that girls are disproportionately affected.

Interestingly, Ainsworth and Semali (2000: 27–8) found that maternal mortality did not affect child illness, while paternal mortality did. They speculate that one reason could be that reported illness rates were used in the survey, and if fathers were less well-informed about child illnesses than mothers, this could lead to a bias in the results.

It should be remembered that female labor force data is poor in many countries, and thus the data presented in this paragraph may have been compiled in the light of many gaps. Budlender (2004: 19–20) reminds us that, while there have been specific attempts to improve the data on women's productive activities in countries such as South Africa and India, in many others, labor-force data continues to undercount women's work in the areas of informal paid work, subsistence production, and self-employment. Such employment may be particularly important for poorer women, so it is likely the labor-force data often poorly represents the employment experiences of such women (see John Sender and Deborah Johnston [1996] on South Africa).

The International Conference of Labour Statisticians has adopted an international statistical definition of the informal sector to refer to employment and production that takes place in small and/or unregistered enterprises or employment or self-employment that does not confer worker benefits or protection (UNIFEM 2005). This encompasses a very large cross section of activities, such as unprotected wage work in both urban and rural areas, self-employment, and own-account agricultural work. In response, a number of authors have argued that the formal/informal dichotomy is in fact not useful for analytical purposes (Jean-Pierre Lachaud 1994).

Gender segmentation of economic activities is not the only factor determining the manner in which women farmers are affected by HIV/AIDS. Gendered disadvantage, as enshrined in either formal or customary land practices, may lead women to lose their land rights after widowhood.

Nilüfer Çağatay, Diane Elson, and Caren Grown (1995) show that models of economic restructuring give little attention to the reproductive sector, and in effect, they assume an unlimited supply of female labor. Consequently, such models are not able to predict the interaction between economic restructuring, productive activity, and women's overall work burdens.

See Young (2000) for a review of the various methodologies. Different methodologies provide different estimates of the size of the reproductive economy. A common approach, which usually yields the highest estimates is to apply a specialist wage rate to input activities. However, Marilyn Waring reminds us that we need to consider carefully the measures of value obtained in this way, as the market-equivalent functions of reproductive work are “sex-segregated or sex-stereotyped jobs,” so the wages paid to such workers will reflect gendered disadvantage (1989: 280).