Infochange India

HIV AIDS

Tue 2Sep2014

You are here: Home | HIV/AIDS | Treatment | Economic implications of inertia on HIV/AIDS and benefits of action

Economic implications of inertia on HIV/AIDS and benefits of action

Large amounts of money are spent in fighting HIV/AIDS. Ajay Mahal justifies the expenditure by citing the large adverse impacts on economic indicators and other socially desirable goals of society if the epidemic is not curbed

The HIV epidemic affects people in their most productive ages with adverse impacts on life expectancy, the productivity of the labour force and household incomes. It has not always been possible to measure the economic impact of AIDS empirically with a reasonable degree of precision. Moreover, while there is some evidence of negative individual, household and firm level impact, the empirical evidence on the impacts at the sector and national levels is still weak. While purely humanitarian considerations may be relevant in supporting investments in HIV/AIDS intervention, they may not always appear to be so for finance ministers and planners in developing countries. To justify spending more on policies to address HIV/AIDS in a regime of tight resource constraints, it is sometimes important to justify investments in AIDS prevention and treatment as being more critical relative to other investments. To the extent that HIV/AIDS has large adverse impacts on economic indicators and other socially desirable goals of society, policy action may be desirable, preferably early in the epidemic, rather than later. 

Introduction  

This paper has two main objectives. The first is to assess the economic impacts of the HIV/AIDS epidemic. The second is to highlight the potential returns to policy action to address HIV/AIDS. 

The paper’s focus on the economic impacts of HIV/AIDS epidemic is guided by at least three related considerations. The first is simply that adverse economic impacts are potentially a key element of the price economies must pay if policy-makers do not pay adequate attention to the health of its population, of which HIV can be an important determinant. There is a large, albeit relatively recent, theoretical and micro-econometric literature that highlights the adverse impacts of ill health and poor nutrition on individual labour force participation, earnings, asset holdings, and the like [Strauss and Thomas 1998]. In its emphasis, this recent literature on the economic implications of ill health at the micro-level is a departure from the traditional treatment of health as an outcome of economic circumstances. Thus, a well-known review of the literature notes, 

    There is a general consensus that income has a strong effect on the demand for health. Until recently, there was less agreement on the reverse relation: the effect of health on income, or more generally, labour outcomes. Even fairly recent reviews…conclude that there is little reliable evidence that health has an important impact on labour productivity… [Strauss and Thomas 1998: 813]. 

The relative neglect of health as a causal factor in influencing economic outcomes is even more marked in the literature if the concern is with economic impacts at a more aggregate level than the individual – at the level of sectors, or national economies. Up until the early 1990s, the empirical and empirical economic growth literature focused exclusively on the role of capital and labour, the latter often augmented by schooling, and technological change, but hardly ever on health as a key element of human capital [Barro 1991, Barro and Sala-I-Martin 1995, Solow 1956, 1957]. Even when a relationship has been found between indicators of health and income per capita, it has either been discounted, or thought to be an indication of the impact of economic development on health. Thus, Barro and Sala-I-Martin (1995:432) conclude “It is likely that life expectancy has such a strong, positive relation with growth because it proxies for features other than good health that reflect desirable performance of a society”. (italics mine). The standard perspective of this earlier literature appears to have been that of Preston (1976) who noted the key role of economic development in improving life expectancy. Nonetheless, there is now a significant literature focusing on the aggregate impacts of health on growth of real GDP per capita and real GDP, or their levels [Bhargava et al 2001; Bloom, Canning, and Sevilla 2001; Bloom and others 1996; Gallup and Sachs 2000; Hamoudi and Sachs 1999]. Gallup and Sachs, for instance, demonstrated that countries with high levels of malaria had much lower levels of per capita income, even after controlling for potentially confounding factors. This new literature also finds a strong positive effect of life expectancy at birth on real GDP per capita [Bhargava et al 2001; Bloom, Canning and Sevilla 2001]. However, Bloom and others (1996) examined, using cross-country data, the impact of malaria and tuberculosis prevalence on real per capita income and found no statistically discernable effect. The new literature on the aggregate economic impacts of health has also emphasised ‘indirect’ ways in which health can influence economic outcomes – Bloom and Williamson (1998) demonstrated the long-term effect of health improvements associated with the process of demographic transition. Their work focuses on the declines in child and infant mortality followed by declines in fertility characteristic of the demographic transition – taken together, these lead to an initial bulge in the age-distribution of the population in the very young ages. Initially, this process leads to an increase in dependency ratios, but this situation is reversed when members of the age-group belonging to the ‘bulge’ reach the working ages. In this latter phase, they have the potential of contributing to increased production and savings, a possibility that Bloom and Williamson refer to as the ‘demographic dividend’. Indeed, they point out that this dividend accounted for nearly one-third of the East Asian ‘miracle’. 

The second reason is the size of the HIV/AIDS epidemic as a health policy challenge. Twenty-five million people have now died of AIDS worldwide and the number of deaths is certain to rise from its current level of 3 million per year. At its current levels, AIDS is the fourth largest cause of mortality worldwide, ranking just below heart disease, cerebro-vascular disease, and acute lower respiratory tract infections [World Health Organisation 2000]. In Africa, it accounted for nearly one-fifth of all deaths in the most recent year for which data were available, making it the leading cause of mortality on that continent by a large margin. Nearly 40 million people are currently living with HIV/AIDS and 5 million were infected with HIV in 2001 alone. Infection rates may be stabilising in sub-Saharan Africa, home to 70 per cent of those infected with the virus, principally because relatively few high-risk individuals remain uninfected. In other areas, however, the epidemic is still growing. Russia saw nearly a 50 per cent increase in HIV infections in 2001, and the number of cases in eastern Europe and central Asia has risen by more than onethird in the last year [UNAIDS 2000a, 2001]. With UNAIDS also voicing concerns over complacency in the West and in, it seems likely that, if anything we may be underestimating the future impact of this devastating epidemic [Haney 2001].1 Indeed, the number of people living with HIV today is 50 per cent above the year 2001 predictions made by the United Nations in 1991 [UNAIDS 2000b]. 

Finally, apart from its large scale, HIV/AIDS is potentially characterised by several elements that suggest its influence on individuals, sectors and national economies. The HIV epidemic affects people in their most productive ages with adverse impacts on life expectancy, the productivity of the labour force, and household incomes. Mahal (2002, Table 1) summarises data provided by UNAIDS which shows that more than 90 per cent of the world’s HIV-positive cases belong to the age-group 15- 49 years. Some researchers also suggest that the economically more productive groups could be potentially at greater risk of infection, especially in the early phases of the epidemic [Over 1992]. The large size of the epidemic and its impact on the more productive members of the labour force suggests a large negative effect on growth in real income per capita in the context of standard models of economic growth. Moreover, taken in conjunction with the work of Bloom and Williamson (1998) discussed above, they imply a third way in which economic growth can be adversely affected – a ‘reverse demographic gift’ because of the deaths and morbidity among people in prime working ages. There are other channels through which HIV/AIDS could have negative consequences for economic growth as well. These could include a decline in savings rates that result from increased medical treatment costs associated with HIV/AIDS [Cuddington 1993a,b; Cuddington and Hancock 1994; Over 1992]. Savings rates could also decline if people expect to live for a fewer number of years owing to HIV/AIDS, and so feel less need for savings to meet their old age consumption needs.2 In similar vein, high rates of AIDS-related deaths among the more educated agegroups could directly act to reduce the stock of human capital, as well as indirectly, because people will have less of an incentive to acquire costly educational capital, if they do not expect to live long enough to enjoy substantial gains from acquiring it. Future stock of educational capital could also be affected if children whose parents die prematurely due to AIDS face economic bottlenecks in efforts to continue their education.3 In addition to the influences on aggregate economic performance, it is obvious that HIV/AIDS has the potential to adversely influence individuals, households, and sectors within national economies. Much as in the micro-literature on the impacts of illness, HIV/AIDS can influence household incomes, labour force participation, child schooling, nutritional intake and the like. The role of stigma associated with HIV/AIDS would, if anything, only serve to worsen already unfavourable outcomes. Moreover, the nature of the HIV epidemic and its large size suggest that it would negatively influence sectors that are labour intensive, such as agriculture and mining; sectors that have to do with ill people, such as health care and health insurance; sectors that may be particularly sensitive to the risk of ill health – such as tourism; and sectors, such as transportation, whose mobile workers are often a major conduit of HIV transmission [Barnett and Blaikie 1992, Barnett and Whiteside 2000, Bharat 1999, Bloom and Glied 1993; Bloom and Mahal 1996, Bloom, Mahal and River Path Associates 2002, Bloom et al 2002, Guinness and Alban 2000]. 

Counteracting some of these negative influences is the possibility that community responses and support mechanisms could have immediate and long-term consequences by sharing the economic and psychosocial burden of HIV/AIDS. Moreover, in a purely technical sense, it is not obvious that declines in the rate of growth of real income would emerge in settings with substantial levels of unemployment. In the final calculus, of course, these are issues that are best settled empirically. Section II of the paper reviews the literature on the subject and assesses the potential economic impacts of the HIV/AIDS epidemic. 

To the extent that HIV/AIDS has large adverse impacts on economic indicators and other socially desirable goals of society, policy action may be desirable, preferably early in the epidemic, rather than later. The question of appropriate policy is also relevant here, and there is a well known objection to policy action that must typically be addressed satisfactorily in this regard. This is the possibility that individual (or private) actions act to neutralise government policy, so that the net outcome of the policy intervention ends up being rather small. By way of illustration, consider a policy that supports government subsidies of HIV testing of blood for transfusion. Suppose that in the absence of any government policy, individuals using blood would actually have paid for testing of blood, because they assess safe blood to be of value. In this circumstance, because HIV-testing would have occurred in either scenario (with, or without, government policy action), the only impact of the government policy of subsidising HIV tests could be a transfer of public resources to individuals using transfused blood, with no influence on HIV. Unless there is a clear assessment that such a transfer is socially of benefit to society, the policy is not a desirable one.4 Sometimes the effects of a policy can even be in a direction opposite to the intended effects, as for example, policies that seek to identify and imprison HIV-positive individuals [Mahal 2002]. In this case, individuals who might otherwise have visited formal health facilities and thereby come in contact with HIV counselling and methods to reduce the risk of infection to others, may choose not to do so. This would enhance HIV transmission, instead of the intended policy effect of reducing it [Misra, Mahal and Shah 2000, Philipson and Posner 1993, Kremer 1998]. 

The question of what an appropriate policy response to HIV/ AIDS might be has been more fully addressed in a companion Economic and Political Weekly March 6, 2004 1051 piece [Mahal 2002]. That still leaves unanswered, at least partly, two policy questions. The first has to do with the question of what the returns to investment in an ‘appropriate’ response to HIV/AIDS are. The second has to do with the timing of HIV/ AIDS policy interventions – that is, whether there are social net gains to be had from acting sooner, rather than later. Section III is an initial attempt to address these questions using standard techniques of economic evaluation using selected examples from recent work and is also the concluding section. 

II

Economic impacts of HIV/AIDS

The discussion on the economic impact of the HIV/AIDS epidemic is divided into three sub-sections. The first sub-section focuses on the economic impacts at the level of the individual and the household. The second examines the sector-level impacts of the epidemic. The final sub-section discusses the literature on the economic impacts of the HIV/AIDS epidemic at the national level, and presents new results using cross-country data. 

Household Economic Impacts of HIV/AIDS

As noted in the previous section, there are several ways in which individual households are likely to be economically affected by AIDS. The most visible impact is the expenses associated with treatment of individuals that are borne by the household(s) to whom the individuals belong. Bloom and Mahal (1996) and Bloom and Glied (1993) examined the annual treatment costs of HIV/AIDS in nine Asian countries in the mid-1990s and concluded that in most cases, these costs were more than twice the per capita income in each country. In India, for instance, the ratio of treatment costs to per capita income was 2.2, in Sri Lanka 1.5, Indonesia 2.4 and in China 2.2. The Bloom and Mahal cost estimates did not include the costs of anti-retroviral drugs (ARV), which could only be expected to further push up the economic burden on families and households. The domestic manufacture of many of the combinatorial ARV drugs in India and their export has presumably reduced the estimated economic burden of ARV treatment for individuals in countries within the region, but this will only be a short-term reprieve once patent protections for new HIV/AIDS drugs kick in after a few more years. 

A second key immediate impact is on earnings and incomes of households with members with HIV/AIDS, especially since HIV/AIDS affects individuals in their most productive years, when they are likely to members of the labour force, or in a position to provide in-kind support to the household. The impacts can take the form of lost earnings when individuals are sick, or die prematurely due to AIDS. These earnings losses can be quite large given that they accumulate over several years, even under fairly conservative assumptions about discount rates and working life span [Bloom and Glied 1993:164, Bloom and Mahal 1996:39- 40, Viravaidya, Obremsky and Myers 1993, Yang 1993]. As per estimates reported in Bloom and Mahal (1996), lost lifetime earnings were nearly three and one-half times the annual costs of treating AIDS that were more than double of per capita income. In Sri Lanka, lost lifetime earnings due to an AIDS-death were nearly 11 times the annual treatment costs of AIDS. In Nepal, rough calculations reported in Bloom and Mahal (1996:39) suggest that earnings losses due to an AIDS death were more than four times per capita income. Incomes and earnings can also occur due to the loss of a job from stigma associated with HIV infection, even if the HIV-positive individual is not sick with opportunistic infections associated with HIV/AIDS. For examples, from within the south- and south-west Asian region and elsewhere, refer to Mahal (2002:22-3). In these cases, the discounted value of lost earnings is even greater than in the case of an AIDS death, since the period over which the household does not have access to an individual’s earnings is potentially greater. 

Earnings losses and increased expenditures due to AIDS deaths among adult members of the household are reflected in a number of short- and long-term consequences for households that are not always fully ameliorated by counteracting influences by way of community and extended family support, or by health and life insurance. These could include declines in household savings and asset holdings. Studies in Namibia, Tanzania, Thailand, Uganda and Zambia demonstrate that households with AIDS deaths experienced a greater decline in asset holdings than households with non-AIDS related adult deaths, or those with no deaths at all [Menon et al 1998, Nampanya-Serpell 2000, Pitayanon, Kongsin and Janjareon 1997, Guiness and Alban 2000]. These studies also indicated that the number of dependents increased, within the households that were affected by AIDS, relative to working members. Evidence from a number of African countries also suggests that overall labour input in agricultural activities declined in households that experienced an adult death [Barnett and Blaikie 1992, Guinness and Alban 2000]. The magnitude of the adverse effects varied by the initial economic position of the household – with richer households better able to cope with the adverse economic implications of HIV/AIDS than poorer ones [Basu, Gupta and Krishna 1997, Guinness and Alban 2000, Nampanya-Serpell 2000, Mahal 2002]. Janjareon (1998)’s study of Thailand also suggests that the educational achievement of the head of household had a protective effect with respect to the adverse financial implications of HIV/AIDS. 

In addition to long-term consequences manifested in the form of declining household asset holdings, members of households affected by HIV/AIDS are likely to have lower long-run accumulations of human capital, whether measured in terms of achievements in education, or health. A study of 324 rural and urban households in Zambia by Nampanya-Serpell (2000) suggests that AIDS deaths within households are likely to be associated with declines in nutrition and the loss of educational continuity among children, with the effects likely to be especially concentrated among rural households and the urban poor. Another study from Zambia suggests that 55 per cent of a sample of AIDS-affected households was unable to meet the costs of educating children due to AIDS. Bechu (1998) found household consumption declines of as much as 44 per cent in the year following an AIDS death in Cote d’ Ivoire, although it is less clear, given the length of the study, whether this effect was sustained over time. These studies are based on small-sized samples so that further work may be necessary to generalising their findings. 

A note on the availability of formal health and life insurance is useful. In general, such insurance is not accessible to people with HIV/AIDS due to a variety of excludability clauses and due to the low coverage of private health insurance in developing countries [Bloom et al 1997]. Thus, safety nets offered by the public sector (such as public insurance) are the sole options open to households [Bloom and Glied 1993]. The latter too, is hampered by the poor fiscal situation in many of the developing countries and the predominance of informal sector employment that lies outside the ambit of the formal public sector. One of the few ways in which households can reduce the impact of AIDS is by using public sector health facilities that are often available at subsidised rates to the poor. Unfortunately, in countries with a substantially advanced HIV/AIDS epidemic, this is resulting in overwhelming the capacity of the public health sector. HIVrelated bed occupancy rates in public hospitals in several countries in sub-Saharan Africa range from 25-70 per cent, with obvious implications for the health budget [Guinness and Alban 2000]. Moreover, it appears that not all households can access services at subsidised rates. A National AIDS Accounts study in Rwanda indicates that less than 30 per cent of the AIDSaffected households were able to meet their health care treatment expenditure needs from their own resources [Guinness and Alban 2000:7]. 

Sector Impacts

In addition to impacts at the level of the individual and the household, HIV/AIDS can have implications at the level of sectors and national economies. This section presents a discussion of some of the findings regarding the impacts of HIV in five sectors – health (including health insurance), tourism, agriculture, private industry and transport. 

The relationship between HIV and health is obvious. But have its impacts been reflected at the sector-level, say, by way of an increased burden on health services, increased share of health budgets, and on the health insurance sector? There is some evidence, particularly from sub-Saharan Africa, that this indeed is the case. In their survey of the literature, Guinness and Alban (2000:5) cite studies from Burkina Faso, the Democratic Republic of Congo, Uganda and Tanzania indicating that bed occupancy attributable to HIV/AIDS exceeded 50 per cent in selected hospitals. These are countries with HIV-prevalence rates of 5-10 per cent at around the time the study was conducted. For countries with HIV prevalence rates in excess of 10 per cent (Cote d’Ivoire, South Africa, Swaziland, Zambia and Zimbabwe), available data suggests that bed occupancy due to AIDS ranged from 25-70 per cent in urban hospitals [Guinness and Alban 2000:10]. Estimates reported in Kone et al (1998:258) show that 21 per cent of inpatient bed-days in Cote d’Ivoire in 1996 were accounted for by AIDS patients. Guinness and Alban also summarise studies that indicate significant shares of the health budget being accounted for by HIV/AIDS – 20 per cent of the Malawi health budget in 1996 and 13 per cent of the Swaziland ministry of health budget in 1994. In Cote d’Ivoire, about 5.7 per cent of public health spending in 1995 was AIDS-related, the corresponding figure for Tanzania being 3.1 per cent [Shepard 1998: 247]. Public health spending is not the only casualty. A national AIDS accounting exercise recently conducted for Rwanda suggests that nearly 10 per cent of all health spending, public or private, was accounted for by HIV/AIDS, and more than 90 per cent of all spending on treatment and prevention of HIV/AIDS took the form of out-of-pocket spending by households [Barnett et al 2001]. Estimates reported in Shepard (1998) for Tanzania and Cote d’Ivoire suggest that AIDS-related health spending accounted for 13 per cent and 7 per cent, of all health spending, public or private, in 1996. Projections undertaken by some of the research reported in Guinness and Alban indicates even greater shares in the health budget being allocated to HIV/AIDS in later years, and greater bed occupancy in public hospitals on account of HIV/AIDS. Information available for Thailand, the one country in Asia with a significant record of public spending on HIV/AIDS, indicates that more than 5 per cent of all public sector health spending in the mid-1990s was on HIV/AIDS. 

Additional estimates of the effects of HIV/AIDS on health spending patterns is available from a study by Arndt and Lewis (2000) who used a 14-sector computable general equilibrium model to assess the economic impact of HIV in South Africa. In their framework, and unlike previous work, the health sector was explicitly accounted for along with assumptions on household and government spending on health due to HIV/AIDS. Simulations reported in their paper suggest that the health sector in South Africa would not suffer as much as the other sectors Economic and Political Weekly March 6, 2004 1053 on account of HIV/AIDS over the period 2001-2010, with the GDP in 2010 under projected AIDS scenarios being nearly 6.5 per cent higher than if it would otherwise be if the health sector was excluded from the GDP computations [Arndt and Lewis 2000:12, Arndt and Lewis 2001]. In fact, compared to the no- AIDS scenario, the overall GDP under AIDS would be nearly 17 per cent lower in 2010, whereas the corresponding figure for value added in the health sector would be 10 per cent higher as per their simulation results [Arndt and Lewis 2001:16]. 

Public expenditures on HIV/AIDS for countries in the South and south-west Asian region are much smaller, as in Sri Lanka and India [Bloom et al 1997, National AIDS Control Organisation of India 2001]. Scenario analyses undertaken in Bloom et al (1997) also indicate that the effect of HIV/AIDS on bed occupancy, while small in magnitude, could still be severe given that excess capacity (supply of inpatient days relative to demand) appears to be extremely limited at present. This study also pointed to a second way in which an expanded HIV/AIDS epidemic could pose problems for the health sector – by constraining supply of medical personnel, many of whom revealed that they would need added payments to compensate for increased risk of HIV infection in health care settings with high HIV prevalence rates. 

Relative to studies of the impact of HIV/AIDS on health services and public health spending, there are very few analyses of the impact on the private health insurance sector. A major reason for this presumably lies in the exclusion of HIV-positive individuals from the pool of insurable individuals, as indicated above. However, exclusion clauses may not always be effective in reducing the costs to insurance companies, especially if treatment costs for opportunistic infections are passed on to thirdparty payers without disclosure of an individual’s HIV status. Indeed, one Zimbabwean insurance company estimated that 45 per cent of its health insurance claims in 1995-96 were AIDSrelated [Bloom, Mahal and River Path Associates 2002]. Insurance company reaction to the HIV/AIDS is another way to discern potential impacts of HIV/AIDS, even if they cannot be directly measured. Thailand’s American International Assurance (AIA) works with non-governmental organisations to promote HIVprevention among factory owners. The company gives financial incentives and discounts to companies with strong workplace and community prevention programmes. 

HIV/AIDS can potentially affect the tourism sector in high prevalence countries by reducing the demand for visits by foreigners who do not want to face the risk of HIV infection, a concern expressed by several of the respondents in a survey of tourism experts in Sri Lanka [Bloom et al 1997]. In general, however, it is not apparent why the standard visitor will face an increased risk of infection unless he or she is likely to undertake activities at high risk for HIV infection, such as unprotected sex, sharing of injecting equipment and the like. To the extent that the proportion of such visitors in the overall flow of tourists is unlikely to be very large, it is difficult to imagine large effects of HIV on tourist flows. This is also backed up the little empirical evidence that exists. A cross-country analysis of 31 countries for which data on tourist flows, HIV/AIDS and other determinants of tourist flows were available, suggests no association between HIV/AIDS and tourism inflows into a country [Bloom et al 1997]. 

Given its disproportionately high impact on young adults, it is possible to imagine that HIV/AIDS will have a significant impact on the agricultural activities, which tend to be labour intensive. A study for Rwanda estimated that the loss of a female adult member of an agricultural household could lead to a nearly 50 per cent decline in its farm labour inputs [Gillespie 1989], and similar results have been documented elsewhere in sub- Saharan Africa [Guinness and Alban 2000, and references cited therein]. One consequence of the AIDS epidemic also appears to have been a shift to less labour-intensive cash crops, declines in the area cultivated, and less animal husbandry [Barnett and Blaikie 1992, Guinness and Alban 2000:7-8]. In Zimbabwe, household survey results suggest that AIDS-affected households experienced significant declines in production on average – 61 per cent in maize production, a 47 per cent decline in cotton production, and a 37 per cent decline in groundnut production [Kwaramba 1997]. 

HIV/AIDS could also impose costs of replacing labour, increased insurance premiums, and funeral expenses on agricultural firms, thereby affecting their profitability, and chances for economic survival. But do these results translate into firm, or even sector level effects? There is some evidence that these effects are being felt at the level of individual firms. One early-1990s study of a sugar estate in Zambia suggests that HIV/AIDS-related illnesses accounted for about 2 per cent of lost labour time and 2 per cent of its costs of production, projected to increase to 3.1 per cent by the mid-1990s [Halswimmer 1994]. Another study of a sugar estate in Swaziland suggests that 30 per cent of the deaths in its labour force over a three-year period from 1996-98 could be attributed to AIDS [Bollinger and Stover 1999]. Analyses of agricultural firms in Kenya suggest a decline in labour force productivity and profitability due to AIDS in recent years [Rugalema 1999]. Production losses due to HIV/AIDS related deaths and morbidity were estimated to be 3.4 per cent of gross profit in one Malawi tea and coffee estate [Jones (1997) cited in Guinness and Alban 2000]. 

The effects of HIV/AIDS on national or regional agricultural production levels, however, have not been as well documented in available literature on the subject. A major reason could simply be the substitution of any lost production by increased agricultural production among households not affected by AIDS, a process facilitated by land transfers/sales from AIDS-affected families to such households. In their survey of the literature on the economic impacts of AIDS in Africa, for instance, Guiness and Alban state, “The adverse impacts of the epidemic on small holder agriculture are often subtle enough to be invisible at the macrolevel...” [Guinness and Alban 2000:2]. The only sector level estimates available are from CGE model-based simulations undertaken by Arndt and Lewis (2001:16) for South Africa who report that value added in the agricultural sector in that country would be 17 per cent lower in 2010 under a projected AIDS scenario compared to a situation of no AIDS. 

HIV/AIDS has the potential of influencing private firms’ operating in non-agricultural sectors along the lines suggested in the previous paragraph – costs of worker replacement, absenteeism, insurance expenses, and health care expenditures [Bloom, Mahal and River Path Associates 2002]. There is the possibility of legal action related to discrimination against HIV-infected employees, potential loss of the customer base and loss of morale in the workforce as workers lose many of their compatriots to AIDS, or if HIV-positive workers are stigmatised and cannot work [Bloom, Mahal and River Path Associates 2002, National AIDS Fund 2000]. The evidence on the economic impact of HIV/ AIDS on the private sector thus far is mixed. Using data from a survey of nearly one thousand firms in sub-Saharan Africa, 1054 Economic and Political Weekly March 6, 2004 Biggs and Shah (1997) concluded that the impact of AIDS on staff turnover was minimal. They did find, however, that replacing professional staff, often thought to be at high risk for HIV infection based on early studies in Africa, to be a significant problem, with firms taking 24 weeks to replace a deceased professional, compared to 2-3 weeks for less skilled staff. Indeed, there is an example of a multinational in South Africa hiring three workers for each skilled position to ensure that replacements are on hand when trained workers die [Bloom, Mahal and River Path Associates 2002:7]. 

There are cases of individual private firms or a small set of firms facing an increased economic burden on account of HIV/AIDS, and some are discussed in a recent survey of the literature in Bloom, Mahal and River Path Associates (2002). Zambia’s largest cement factory saw a 15-fold increase in funeral related absenteeism between 1992 and 1995. As a result, the company has restricted employee absenteeism for funerals to only those cases where the AIDS-death was a spouse, parent and child. In Benin, a 14-firm case study found that 50 per cent of the HIVpositive cases held positions considered ‘important’ by these firms. These firms noticed increased absenteeism and their policy of holding salaries constant while reducing workloads was leading to increased costs (and reduced profits) at the time of the study. In Zambia, at Barclays Bank, the death rate among employees rose from 0.4 per cent in 1987 to 2.2 per cent in 1992 – the company lost an average of 36 of its 1,600 employees to HIV/ AIDS. Ex gratia payments to families increased substantially over this period and more than 70 per cent of the deaths occurred among employees aged less than 40 years. In Zimbabwe, one study estimated that at a large firm with 11,500 employees, there were nearly 3,400 HIV-positive workers, with the costs of AIDS in 1996 amounting to roughly 20 per cent of the company’s profits, mainly on account of health care benefits. Much of this literature is recent and further work is obviously necessary to arrive at concrete conclusions. 

There are scenario type analyses as well. One study for the US in the early 1990s constructed scenarios of the economic implications of hiring a single HIV-infected person in four settings (high-cost city, large firm), (high-cost city, small firm), (low-cost city, large firm), (low cost city, small firm). The analysis considered the following costs likely to be borne by firms that hire HIV-positive people – health insurance, medical care of employees of HIV, life insurance benefits, sick leave and costs of disability payments – before deducting any monetary benefits in terms of pension savings resulting from early death of the employee, and adding any hiring and training costs of new employees. After discounting (since the death from AIDS of an HIV-infected employee occurs in the future), the study found the estimated cost of hiring an HIV-infected person to be US $ 31,800, US $ 20,600, US $ 4,400 and US $ 2,300 under the four scenarios [Bloom and Glied 1991]. 

HIV/AIDS can affect firms by adversely affecting the customer base, since the group hardest hit by AIDS – young adults of working age – is also the major source of demand for goods and services. Indeed, caring for people living with AIDS is expensive, so while certain sectors such as health might see increased demand, most others ought to experience spending redirected away from them. Such effects are not readily detected by individual firms because of the dissipation of spending implications across local and international economies. Effects on the customer base are more likely to be transparent if there are dominant firms, or firms organised into business associations. Thus, the JD Group (South Africa’s leading furniture retailer), which performed its own research on the potential impact of the epidemic on its markets and used a forecast of HIV-prevalence among its customers, found that changes in demography would reduce its customer base by 18 per cent by the year 2015. 

Another sector that has received attention in the context of the HIV/AIDS epidemic is the transport sector. Several analyses have focused on the role of the trucking industry as a facilitating factor in HIV transmission [Bloom and Mahal 1996 and references cited therein, Giraud 1993]. There are also a few analyses of the impact of HIV/AIDS on the transport sector, relating to railways and the trucking industry. Giraud (1993) developed a methodology to assess and predict the impact of HIV among long-haul truck drivers on Thailand’s trucking industry over the period from 1991 to 2000. Giraud concluded that HIV/AIDS related costs to the trucking industry would increase from an estimated US $ 40,000 to nearly US $ 14.5 million by the year 2000. Another more recent study, of the Uganda Railway Corporation, concluded that HIV/ AIDS had substantially increased the labour turnover rate for the Corporation and that nearly 10 per cent of its employees had died of AIDS in recent years [Bollinger, Stover and Kibirige 1999]. Bollinger et al (1999) report an absenteeism rate of nearly 15 per cent among employees of the National Railways of Zimbabwe, mainly for AIDS-related reasons, and estimate that one of the major bus companies of Zimbabwe was losing roughly 7 per cent of its profits to expenses related to AIDS. Another set of results is available from the CGE analysis of Arndt and Lewis (2001:16) who report that the transport sector in South Africa would have 20 per cent lower value added in 2001 under a projected scenario of the AIDS epidemic, relative to a no-AIDS scenario. Although few in number, these studies taken together suggest that the transport sector could possibly be a major casualty of HIV/AIDS. There are no studies of the economic impact of AIDS on the transport sector in any of the countries in the South and south-west Asian region. Available data do indicate, however, of behaviour at high risk of HIV infection being common among truck drivers in the region [Mahal 2002]. 

Impact on National Economies

Two types of impacts are worth noting – on national outputs (or outputs per capita) and on the distribution of national income. The two taken together, have implications for the proportion of national population living below the poverty line, as well. This sub-section assesses primarily the impact of HIV/AIDS on economic growth, given its emphasis in the bulk of the literature on the economic impacts of AIDS. The implications of the epidemic for poverty have been discussed in depth in Bloom et al (2004), and only a summary is presented here. 

At the global level, there is a statistically significant link between low income per capita and HIV prevalence rates – the poor the country, the greater its HIV prevalence. Moreover, absolute poverty rates across countries, defined as the proportion of population living on or below US$1 per day (1993 PPP) are positively correlated with HIV prevalence. There is a positive and statistically significant correlation between HIV prevalence and economic inequality as well [Bloom et al 2001, Over 1998]. Nonetheless, going beyond these simple associations to the causal impact of HIV on poverty and inequality has not been demonstrated thus far, at least with national-level data. 

At the micro-level, there is some evidence to support the assertion that the poor and the less educated are at greater risk for HIV infection [Bloom et al 2001]. A study in rural Uganda that followed a cohort of 20 thousand adults over three and a half years found, however, that HIV associated mortality was highest among the better educated. Over (1992) also presented evidence of greater HIV prevalence among the economically better off groups, using small sample studies in sub-Saharan African countries. Despite this contrary evidence, it does appear that the trend is towards the poor being increasingly affected. Another study from Uganda, for example, shows that the better educated are likely to be hit hardest during the early stages of the epidemic, but that the infection rates are now falling most quickly among those with better education (Bloom et al 1998). Moreover, poverty forces people to make suboptimal choices that put them at risk for HIV infection. A series of small-scale studies from sub-Saharan African countries, Haiti, Sri Lanka and Brazil all show how poor women can be forced into sex work, into providing sexual favours in return for money, and to be less able to insist on condom use [Bloom and Mahal 1996 and literature cited therein; Bloom et al 1997, Bloom et al 2001]. 

Growth Impact of HIV/AIDS

Early work on the impact of AIDS on growth of real income (or real income per capita) inferred, rather than directly demonstrate, the aggregate economic impact of HIV/AIDS from the combination of large projected numbers of prime-age HIVpositive individuals and the relatively high costs of treating people with AIDS. For instance, the World Bank (1993:20) concluded “the heavy macroeconomic impact of AIDS comes partly from the high costs of treatment, which divert resources from productive investments...(and that) ... AIDS ... poses a threat to economic growth.” Again, the United Nations Development Programme (UNDP) has stated, “the extent of illness and death caused by the epidemic could deplete critical sectors of the labour force, ... and adversely affect every sector of the economy. The consequences of the spread of the virus could be inexorable and awesome” [UNDP 1992:1]. Michael Merson, the former head of the World Health Organisation (WHO) Global Programme on AIDS, stated “the deaths of millions of able-bodied adults will ... rob society of their education, skills and experience. The resulting productivity losses will ... threaten the very process of development” [Merson 1992:2]. 

More recent work on the aggregate economic impact of AIDS has essentially taken a more rigorous methodological route and falls into mainly two groups. The first group derives its conclusions from well-established economic models, customised in various ways to account for key aspects of the AIDS epidemic. It includes analyses that use computable general equilibrium (CGE) models, as well as those using a neoclassical growth model. Kambou, Devarajan and Over (1992) simulated the economic impact of the AIDS epidemic using an eleven-sector computable general equilibrium model of Cameroon. In their analysis they assumed that the AIDS epidemic would claim the lives of 30,000 workers (or 0.8 per cent of the labour force) each year from 1987 to 1990, with deaths occurring disproportionately among the more skilled segments of the workforce. Thus 6.0 per cent of the skilled urban work force was assumed to die of AIDS each year, compared to 0.4 per cent of the unskilled rural labour force. In their simulations, the AIDS epidemic lowered the rate of growth of real GDP by nearly 2 percentage points per year, with the rate of growth of real income per capita not being significantly affected. 

More recently, Arndt and Lewis (2000) used a 14-sector general equilibrium model to assess the future economic impact of the AIDS epidemic in South Africa. Their model was intended to be more comprehensive than the Kambou, Devarajan and Over approach, and included a health sector, allowed for impacts on savings on account of medical expenditures undertaken by the government and households, labour force impacts, household and government allocations to health sector spending, exogenously given assumptions on trends in sector productivity. Moreover, in their model the impact of the AIDS epidemic was assumed to fall disproportionately on low-skill segments of the labour force, in line with available evidence from South Africa [Arndt and Lewis 2000:9]. Dynamic elements, such as savings and capital formation, were incorporated by including outputs from the one-period model as inputs into the model for the second period. The main conclusion of Arndt and Lewis was that over the period 2000-2010 the annual rate of growth of real GDP in South Africa under their projected AIDS-scenario would be substantially lower in comparison to a no-AIDS scenario, with the difference ranging from 1 percentage point to 2.6 percentage points, depending on the year. The net effect would be a real GDP in 2010 that would be 17 per cent lower in size, compared to a no-AIDS case. They found that per capita real GDP would also suffer on account of HIV/AIDS although not as much as real GDP, being about 8 per cent lower in 2010 compared to a no-AIDS scenario. 

Cuddington (1993a,b) and Cuddington and Hancock (1994a,b) conducted simulation analyses in the context of neoclassical growth model to explore the effect of AIDS on growth of real income per capita. In these models, the AIDS epidemic affects economic performance through two main channels. First, AIDSrelated morbidity and mortality decreases the size of the labour force, as also its average experience, the latter a key element of productivity. Second, AIDS-related medical expenditures lower public and private savings, leading to reduced investments in physical capital. This set of studies focused on the epidemic’s impact in Tanzania and Malawi and indicated that AIDS would depress the annual rate of growth of real GDP per capita by an average of 0.25 percentage points over the period 1991-2010, using World Bank projections of the AIDS epidemic in these countries that were available at the time. Using a similar modelling approach, MacFarlan and Sgherri (2001) recently examined the macroeconomic impact of AIDS in Botswana for the period 1996-2021. Their main findings are that overall GDP under projected AIDS scenarios would be substantially smaller in 2021 relative to a no-AIDS scenario – with the magnitude under various projection scenarios smaller by 17-30 per cent relative to a situation with no AIDS. However, HIV/AIDS also reduces population substantially in their model, so that the direction of the impact of HIV/AIDS on the relative rate of growth of real income per capita is less clear. It could increase at a rate faster than in a no-AIDS scenario, or slower, depending on the specific scenario considered. 

Applying a related framework to data for sub-Saharan Africa, Over (1992) assumed that AIDS cases would be disproportionately concentrated among the more educated classes and also that 50 per cent of AIDS medical treatment costs would be financed by reduced savings (which translated into reduced investment 1056 Economic and Political Weekly March 6, 2004 and a slower rate of expansion of economic capacity). He concluded that the AIDS epidemic would depress growth rates of real GDP per capita in Africa by roughly 0.15 percentage points per year (0.33 percentage points under a worst case scenario), a sizeable amount in the context of sub-Saharan Africa’s 1980s growth experience – a 1.2 per cent average annual decline in real income per capita from 1980 to 1991. 

In contrast to the studies above that rely on simulations conducted under various assumptions of the HIV/AIDS, an alternative approach is to econometrically estimate the link between HIV/AIDS and national economic performance. Bloom and Mahal (1997) used standard empirical equations of the form found in Barro (1991) and Mankiw, Romer and Weil (1992) to measure the nature and strength of statistical associations between the prevalence of AIDS and the rate of growth of real GDP per capita, using cross-country data for 51 countries. The main rationale for using an empirical approach is its potential use in taking account of standard influences of AIDS as reflected in simulation models of the type discussed above, as well as others (such as community responses to AIDS, life cycle savings behaviour by individuals and the like) not readily captured by the latter. Indeed one obvious benefit is in avoiding the pitfalls of simulation models that rely on assumptions that often lack an empirical justification. The econometric approach adopted by Bloom and Mahal took into account the possibility of simultaneity bias resulting from the effect of economic growth on HIV transmission, as well as possible non-linearities in the relationship between HIV prevalence and economic growth. Their main finding was that the AIDS epidemic had a statistically insignificant effect on the growth of real income per capita, with no evidence of reverse causality during the period 1980 to 1992. 

There are factors that can potentially confound the results found in Bloom and Mahal’s analysis. The first is the possibility that their study was undertaken at a time when HIV-prevalence rates were still too low to have a detectable economic effect at the national level [Bonnel 2000:3, McDonald and Roberts 2001:6]. To be sure, Bloom and Mahal (1997) also presented results for the impact of HIV/AIDS over the period 1987-92, when HIV might have been expected to have a greater effect on economies, relative to earlier years, but the prevalence rates at the time were obviously much lower than at present and, the number of years too small to isolate any long-term effects. Bonnel (2000) examined the association between rate of growth of real income per capita during the period 1990-97 and a quadratic term in HIV prevalence (after controlling for factors that could potentially confound the relationship) and found it to be negative and statistically significant. He also concluded that the HIV/AIDS epidemic depressed the rate of growth of real income per capita in Africa during the period 1990-97 by nearly 0.7 percentage points per year. There are several methodological issues that Bonnel’s work does not appear to fully address fully. These include the issue of the robustness of his findings – the relatively short time period of the study, and whether the coefficients on the HIV variables are robust under different specifications of the growth equation, given especially that the explanatory variables in the regression are rather sparse [Bonnel 2000:21]. 

The empirical approaches of Bloom and Mahal (1997) and Bonnel (2000) have faced methodological objections, given their reliance on ‘single-period’ cross-country regression methods of the type first used by Barro (1991). Of these, three are likely to be particularly crucial. It has been argued, for instance, that lagged GDP per capita, used commonly as an exogenously determined explanatory variable in these analyses is not exogenous. Second, it is also possible that there are country specific factors that can potentially affect the relationship between HIV and economic growth, but that these may not be readily observed – examples could include cultural practices, policy measures that are not readily quantified, and the like. Taking the latter objection into account requires using panel data estimation methods (which can also simultaneously address the first concern), but that requires data for several periods in time, which are not readily available. Third, it has been suggested that the specification used for crosscountry regressions of the type used in Bloom and Mahal (1997) and Bonnel (2000) are essentially ad hoc, without being solidly grounded in economic theory [McDonald and Roberts 2001]; although as to this last point, Bloom and Williamson (1998), provide theoretical grounding to the models used in these papers.5 Another way around these concerns is to estimate ‘aggregate’ production functions that link labour inputs, capital inputs and technology to output (GDP), in which the effects of the AIDS epidemic are felt through these various inputs. Bloom, Canning and Sevilla (2001) carried out such an exercise linking health to output, although they did not include HIV/AIDS in their analysis. Undertaking this exercise would bring the simulation models that rely on production functions, and empirical growth approaches much closer in spirit, than has hitherto been the case. Another way would be to undertake adopt the Mankiw, Romer and Weil approach which also uses a Solow-Swan growth framework as in Bloom and Williamson (1998) but arrives at an alternative formulation of the empirical specification, by focusing more directly on the savings term as an explanatory variable, and on variations in technological change across countries [McDonald and Roberts 2001:8]. 

McDonald and Roberts (2001) sought to address some of the above concerns, by using panel data methods to estimate the impact of HIV/AIDS, and using a modified version of the Mankiw, Romer and Weil (1992) empirical elaboration of the neoclassical growth model. Their main modelling contribution was in linking HIV/AIDS to economic growth via its impact on life expectancy, the latter serving as an indicator of health capital in an empirical equation of the link between growth of real income per capita and its determinants. They report the finding of a statistically significant effect of the HIV/AIDS epidemic on life expectancy, and via life expectancy on growth of real income per capita. Their empirical findings are subject to several caveats, however. First, it is unclear how they were able to obtain national HIV prevalence data for a sample of more than 100 countries for more than one point in time, since HIV data were not available for a large sample of countries for the period prior to 1996.6 Second, their analysis does not appear to have adequately taken into account influences of HIV/AIDS on economic growth that do not work through life expectancy, such as declines in public savings on account on treatment costs due to HIV/AIDS, impacts on productivity and the like. Third, the coefficients in their econometric analyses [McDonald and Roberts 2000:16-27] appear to be remarkably unstable over different specifications. Moreover, the absolute magnitude of the AIDS variable on life expectancy that their estimate is much greater in the OECD country sub-sample than for the full set of countries, or for countries of sub-Saharan Africa, a result that does not accord well with intuition. On balance, however, their empirical approach of emphasising the role of HIV/AIDS in influencing per capita income via life expectancy serves to highlight the role of one major pathway through which the AIDS epidemic will have an effect on national economic performance. Recent empirical works by Bhargava et al (2001) and Bloom, Canning and Sevilla (2001) using panel data techniques also highlight the link between life expectancy and economic growth, and could potentially be modified to serve as a means to understand the links between HIV and economic growth. 

Impact on Growth of Real Income Per Capita: Some New Evidence

In this section we present new evidence on the links between HIV/AIDS and growth of real income per capita. This is a useful exercise for four reasons. First, compared to the situation a decade ago, data is available for a much greater set of countries, a fact highlighted in the work of Bonnel (2000) and McDonald and Roberts (2001). At the time of Bloom and Mahal’s work, data for only about 51 countries were available. Now, however, UNAIDS provides estimates of HIV prevalence in more than 200 countries. Second, the data are of improved quality compared to a decade ago. In particular, sentinel surveillance data for women visiting antenatal clinics in many countries offers a glimpse into HIV-prevalence rates in a group reasonably representative of trends in the general population.7 Third, the HIV/ AIDS epidemic is now entering into its third decade and hence its effects on national economies have a greater chance of being visible. 

A fourth factor has to do with the use of AIDS case estimates, instead of HIV, in our analysis. Use of AIDS data is desirable, because many of the adverse consequences of the HIV/AIDS epidemic for aggregate economic performance have directly to do with effects on labour force via premature death or morbidity, on treatment costs, and the fact that many of the individual responses to HIV are likely to kick in the AIDS stage when they are more likely to be aware of their HIV status. Most developing countries have poor record-keeping systems, so in all likelihood recorded AIDS cases will be biased downwards. Thus modelbased approaches to estimating AIDS cases have been used for developing countries. 

The standard approach has been to use data on HIV prevalence (taken to be representative of the whole population), combine it with a description of the rate at which HIV cases progress to AIDS and to death (normally approximated by a Weibull distribution, Bloom and Mahal (1997)), along with a further assumption about the start date of the epidemic. Unfortunately, an infinite number of time profiles of HIV prevalence that can achieve the HIV prevalence at a point in time exist, even with these requirements. Thus the typical approach to derive the time profile of HIV/AIDS cases has been to make an assumption that HIV incidence follows a gamma function of one (or, two) parameter variety, add to that a further statement about the peak year of incidence, and then to choose the value of the gamma distribution parameter itself [Chin and Lwanga 1991]. Bloom and Mahal (1997) introduced the methodological innovation in a maximum likelihood framework whereby the gamma distribution parameter was chosen simultaneously as part of the econometric specification linking AIDS to economic growth. However, owing to data from sentinel surveillance sites being available for several recent years and developing countries, it is possible to directly derive the time profile of HIV incidence using curve-fitting techniques and software provided by UNAIDS for these countries. The methodology is more fully discussed in UNAIDs (2002). 

For the purpose of examining the impact of AIDS on economic growth, we re-estimated two sets of empirical equations – (a) modified version of the empirical specification used in Bloom and Mahal (1997) with new data, for the period 1980 to 1998, for 66 countries; and (b) a modified version of the equation used by Bloom and Williamson (1998) for 57 countries. The set of countries chosen was smaller than the countries for which UNAIDS provides HIV prevalence data, in order to include only those developing countries for which a large number of sentinel surveillance data were available for some years, for reasonably sized samples. Our sample of countries also included developed nations from Europe and North America, as well as Australia, Japan and New Zealand, where reported AIDS cases can be expected to be a reasonably accurate indicator of the true AIDS cases. These were combined with data on range of geographic demographic and socio-economic variables as additional explanatory factors – real GDP per capita in 1980, government expenditures on education and defence as a proportion of GDP, mean years of schooling, the ratio of exports and imports to GDP, rate of growth of population, the rate of growth of working age population (15- 64 years), whether the country was landlocked, quality of institutions, whether located in tropical regions, life expectancy at birth in 1980 and the rate of growth of lagged per capita income. The sources of this data included the World Development Indicators database [World Bank 2000], the Penn World Tables [Summers and Heston 1991], Barro-Lee database on education indicators, Human Development Reports for various years (UNDP, various), and Gallup and Sachs (2000). 

Our empirical approach was to estimate the following equation, the same as equation (1) in Bloom and Mahal (1997:112), and equation (5) in Bloom and Williamson (1998:431) after including a term for AIDS. (1) Yi = á + âAIDSi + Xið + åi (i=1,2, ..., N). 

Here Yi is the rate of growth of real income per capita (alternatively, the rate of growth of real GDP), AIDSi is the average annual increase in the cumulative adult prevalence of AIDS (the average annual increase in the number of AIDS cases over the estimation period, taken as a proportion of the population aged 15-64 years in 1998), Xi is a vector of variables that influence economic growth and åi are independently and identically distributed error terms, each with zero mean; á, â and ð are parameters to be estimated. A major goal of our analysis is to obtain a consistent estimate of the coefficient of the AIDS variable, â. 

Descriptive statistics for the variables used in our analysis are reported in Tables 1 and 2. We report our main results in Tables 3-6, two each based on the specifications used by Bloom and Mahal (1997) and Bloom and Williamson (1998). Tables 1A and 1B provide the descriptive statistics for the two sets of specifications. Tables 3 and 5 report specifications that estimate the effect of different explanatory variables on the rates of growth of real income per capita and real GDP, respectively, under the Bloom and Mahal model. Tables 4 and 6 present parallel results under the Bloom and Williamson (1998) model. GDP growth was included as a dependent variable in our analysis because even though HIV/AIDS may not influence per capita income (because of its effects on both total output and total population), it has a greater chance of being noticed in effects on overall GDP. In each case, moreover, we consider the growth rate in the dependent variable for the period 1980-98 and for 1990-98. The use of the period 1980-98 in the specifications was obvious, given that it permitted the longest possible time period, given our data, to assess the impact of HIV/AIDS on growth, whether of per capita income or GDP. We also considered the period 1990-98, given that the epidemic was much more severe during this period and therefore, the effects of HIV/AIDS likely more significant. All specifications were estimated both by ordinary least square methods, as well as by instrumental variable techniques, given the possibility that population and HIV/AIDS might be influenced by rate of growth of income [Bloom and Mahal 1997]. The rate of growth of population (and the working age of population) was adjusted to remove the effect of AIDS on population, so that the coefficient on the AIDS variable captures any effects on the rate of growth of real GDP and real income per capita that operate through these variables. The list of instruments is provided in the footnotes to the Tables 1 and 2, and is essentially similar to those used by Bloom and Mahal and Bloom and Williamson in their work. Our estimation method does not rely on panel data methods and thus is open to the methodological objections noted previously. This we propose to rectify in future work, as more HIV prevalence data becomes available. 

Our main findings are as follows. Specification tests undertaken as part of our analyses do not support the hypothesis of reverse causality. This confirms the earlier findings of both Bloom and Mahal (1997) and Bonnel (2000) of there being no statistical evidence of a reverse causality in such models. Tables 3 and 4 report the results of specifications with the rate of growth of real income per capita under the BM (Bloom and Mahal) and BW (Bloom and Williamson) specifications, respectively, for the period 1980-98 and 1990-98. For the period 1980-98, the coefficient of the AIDS variable was statistically indistinguishable from zero under both the BM and BW specifications. However, for the period 1990-98, the coefficient of the AIDS variable was sufficiently large so that the null hypothesis of AIDS not having an effect on the growth of real income per capita was rejected. Unfortunately, it is also the case that the estimated specification for the 1990-98 period provided a much poorer fit to the data under both the BM and BW specifications, relative to 1980-98 period, with many of the coefficients becoming statistically insignificant, and typically with substantially higher standard errors. Thus we have greater confidence in the results based on the specification for the period 1980-98 than for 1990-98. Even though the coefficient of the AIDS variable is statistically significant in BM specification (for the period 1980-98), it is worth noting that it is larger in absolute magnitude than in Bloom and Mahal (1997), which is consistent with the idea that as the scope of the AIDS epidemic increases, its economic effects are more likely to become visible. 

Tables 5 and 6 present analogous findings for BM and BW specifications that have the rate of growth of real GDP as the dependent variable. The results follow the pattern reported above for the rate of growth of real income per capita, except that the coefficients are greater (in absolute value) and are negative in all cases. This accords well with intuition in that one would expect the impact of the AIDS epidemic to be of considerably greater severity for GDP, than for GDP per capita, because of its impacts on the growth rate of population. 

Even though the coefficient of the AIDS variable is statistically insignificant in all specifications that have growth rates during 1980-98 of GDP and per capita GDP as the dependent variable, it is instructive to compare the effects of AIDS based on our estimated coefficients and projections based on simulation models in recent years. Two such exercises, as noted above, have been undertaken for Botswana and South Africa. According to the simulation results of Arndt and Lewis (2000), the AIDS epidemic in South Africa is expected to result in a real GDP in 2010 that would be 17 per cent lower, and its annual rate of growth being between 1.0 per cent to 2.6 per cent lower depending on the year considered, relative to a no-AIDS scenario. These results used ING Baring’s forecasts of HIV/AIDS in South Africa during the period 1997-2010 to arrive at the forecasts of the rate of growth of real GDP [Arndt and Lewis 2000:1]. Using the range of estimated coefficients in Tables 5 and 6, and a forecast of AIDS cases similar to that of Arndt and Channing, suggests declines in the average annual rate of growth of real GDP over the same period ranging from, 1.4 per cent to 3.6 per cent points. That amounts to a GDP in 2010 that is lower than the no-AIDS scenario GDP by 16 to 37 per cent. In the case of the rate of growth real income per capita, the projected declines ranged from 0.7 per cent to 2.8 per cent points over the same period. By contrast, the forecasted decline in the rate of growth of real income per capita in South Africa over the same period was 0.8 percentage points. 

We undertook a similar exercise for Botswana, one of the worst affected countries in sub-Saharan Africa. MacFarlan and Sgherri forecast that between 2000 and 2010, the impact of the HIV/ AIDS epidemic would be to lower the rate of growth of real GDP by 3 to 4 percentage points below its trend rate of growth of 5.5 per cent per year. If we use the AIDS case projections that underlie their simulations together with the coefficients from Tables 5 and 6, we get that the AIDS epidemic has the potential of reducing the annual average rate of growth of real GDP in Botswana by 2.7 per cent to 7.1 per cent points during 2000-2010. Thus, our empirical analyses would point to even more declines in real GDP than would be suggested from the simulation analyses. 

In contrast to the sharp declines in the rate of growth of real GDP and real income per capita that appear imminent in sub-Saharan Africa, the impact on the countries of south and south-west Asia will be small. In particular, given the exceedingly small rate of growth of cumulative AIDS prevalence among countries in the south and south-west Asian region during 1980- 98, the AIDS epidemic has had a negligible impact on economic growth thus far, even India. Indeed, the impact might have been in the opposite direction as suggested by the coefficient of the south Asia dummy variable that was included in the specifications reported in Tables 3-6 (interacted with the AIDS variable).8 How about future economic impacts on countries in the region? That depends on the projected future profile of the AIDS epidemic in these countries. The two cases outlined above – of Botswana and South Africa – offer some insight into the potential future impacts that might result. 

To summarise, while there are economic impacts of the AIDS epidemic it has not always been possible to measure them empirically with a reasonable degree of precision. Moreover, while there is some evidence of negative individual, household and firm level impact, the empirical evidence on the impacts at the sector and national levels is still weak. Even in the case of micro-evidence at the level of individuals, households and firms, researchers need to proceed with caution. The results are often based on data characterised by small non-representative samples collected in extremely hard-hit areas, or on simulations, so that a complete empirical picture of the effects of the virus is not always available. Further work may be necessary to provide conclusive evidence of the size and nature of the effects. 

III

Policy Action on HIV/AIDS

Even if concrete evidence on some aspects of the economic impact of the HIV/AIDS epidemic is not readily available, there are good reasons to think why investment in HIV/AIDS policy, whether prevention or treatment, might be desirable. The first is simply the human development costs of the epidemic – as indicated by the negative effects of stigma and the loss of key adult members of individual households at the micro-level and overall declines in life expectancy at birth in the worst affected countries. There are also measurable economic impacts such as large medical expenditures on treating people with HIV/AIDS, mostly out of pocket, that impose huge financial burdens on the affected individuals and their families. 

While purely humanitarian considerations may be relevant in supporting investments in HIV/AIDS intervention, they may not always appear to be so for finance ministers and planners in developing countries. To justify spending more on policies to address HIV/AIDS in a regime of tight resource constraints, thus, it is sometimes important to justify investments in AIDS prevention and treatment as being more critical relative to investments in other (health and non-health) areas. This section of the paper attempts to measure the benefits of tackling AIDS, as part of an overall development strategy that assess the costs and benefits of alternative courses of policy action. 

Returns to Policy Action in HIV/AIDS

The tools of cost-benefit and cost-effectiveness analyses, that compare the benefits of a policy to its opportunity costs, are standard methods used by economists to evaluate alternative policy options. Cost-benefit analysis compares the benefits of a policy action to its costs both evaluated in monetary terms. In Sri Lanka, studies have shown that preventing HIV transmission via the screening of blood used for transfusion, and the use of disposable, instead of reusable injecting equipment in hospital settings can yield benefits that are much greater relative to costs [Bloom et al 1997]. Cost-effectiveness analysis typically compares an outcome indicator such as lives saved and disability adjusted life years averted [for example, World Bank 1993] that is not measured in monetary units, with costs that are measured in monetary units. There are studies demonstrating the potentially high cost-effectiveness ratio of programmes such as needle exchanges, STD prevention, information provision and the like [Kaplan and O’Keefe 1993, Over and Piot 1993]. Cost-effectiveness analyses for health interventions (including HIV/AIDS) are not always useful for policy-makers when the comparison is with policies in sectors other than health, since the former might have outcome indicators in units other than money. Thus, cost-benefit analyses are typically the method of choice since both benefits and costs are reduced to monetary units, provided, of course, policies in other areas are similarly evaluated. 

Private Action

Cost-benefit analysis (or cost-effectiveness analysis for that matter) for HIV/AIDS programmes, as applied in practice, often fails to account for individual action to protect against the risk for HIV infection. Thus, public action to screen blood for HIV prior to transfusion may have little effect on HIV transmission if recipients (or relatives of recipients) of donated blood deem the risk large enough to pay for blood screening on their own. The studies in Sri Lanka cited above failed to consider this possibility and thus likely overestimated the gains from the HIV prevention policy in question. This type of problem is more likely to occur in programmes where the gain from private preventive action is sufficiently large relative to costs, compared to a setting when the gains are predominantly of a public good nature – and the private gains to some HIV prevention activities may indeed, be large. On the other hand, information (about HIV prevention) provision programmes have a substantial public good component and may not be readily provided by suppliers, so public action may indeed be necessary to support the provision of such information. As another example, to the extent an individual who is HIV positive cares only about his own infection status and not about people he subsequently infects with HIV, private preventive action may be less than socially optimal, again providing a justification for public action. Taking account of secondary infections resulting from a single original HIV case yielded quite high benefit-cost ratios in the Sri Lanka studies, suggesting that public action in blood screening and introduction of disposable needles in hospital settings may indeed have been cost-beneficial from a social point of view. 

Rate of Return to HIV Prevention: The Case of Thailand

Cost-benefit calculations sometimes are not immediately helpful in ranking alternative policy priorities if there are several projects with positive net benefits, but resources are limited so not all can be undertaken. In these circumstances, an alternative formulation of cost-benefit analysis that yields a rate of return to the policy is sometimes useful – the internal rate of return (IRR) method. By examining the rates of return on alternative policy programmes, a policy maker can adopt a simple rule to choose projects – first, the policy with the highest IRR, then the one with the second-highest IRR and so on, until the budget is exhausted. Benefits from HIV prevention accrue from both the medical costs averted (by private and public sectors) and the value of lives saved on account of the intervention(s). Research conducted for this paper provides an attempt at assessing the IRR from HIV prevention efforts based on data from Thailand, whose efforts in the 1990s were successful in reducing the number of annual AIDS cases, which had doubled to 26,000 from 1994 to 1997, back to 1994 levels in just two years. 

The time period from 1990, when Thailand’s prevention activities began, to 2020, was chosen for the analysis. Public sector and donor expenditures on HIV/AIDS jumped from US$0.68 million in 1991 to US$82 million by 1997. It is estimated that roughly 15 per cent of these expenditures were on prevention activity. The private corporate sector spent US$80 million on prevention messages in 1991. Data on changes in behaviour suggests that if behaviours had remained unchanged at 1990 levels, there would have been more than 12 million extra deaths due to AIDS in Thailand, cumulatively, by the year 2020 compared to current behavioural patterns. 

Even with conservative estimates of the impact of prevention campaigns on changes in behaviour, the IRR on Thailand’s prevention programmes was calculated to be between 12 per cent and 55 per cent, depending on the scenario posited. If we focus only on benefits in terms of medical expenditures avoided, the annual rates of return range from 12 per cent to 33 per cent over the 30-year period from 1990-2020. It is interesting in this connection to note that this range of estimates brackets the 26.7 per cent estimate of the IRR to HIV prevention based on a scenario analysis carried out for India [Dayton 1998]. If we include averted income losses as additional benefits (taking these to indicate the value of a “saved” life) resulting from the reduced number of AIDS deaths, in addition to the savings in medical expenditures, the rate of return jumps sharply upwards to range from 37 per cent to 55 per cent annually. To the extent that some of the behaviour change that took place in Thailand would have taken place irrespective of any intervention owing to individual preventive action, these rates of return may be upper bounds to the true returns on HIV prevention. On the other hand, if such private preventive “reactions” occur with a lag to an advanced HIV/AIDS epidemic, the revised rate of return that takes account of such behaviour change would not be very different. 

Estimates of the rate of return for some alternative health interventions are worth noting. The rate of return (inclusive of income losses due to disability) of the global guinea-worm eradication programme, for example, is roughly 29 per cent, compared to the 37-55 per cent from HIV prevention in Thailand, using an equivalent methodology. Our estimates for the rate of return on HIV prevention in Thailand (inclusive of income losses) also exceed the range of rates of return from interventions for river blindness eradication in Africa, estimated to be 6-17 per cent. The World Bank considers an annual rate of return of greater than 10 per cent to be acceptable [Bloom et al 2001]. 

Does It Pay to Intervene Early?

Are there greater returns from intervening earlier, rather than later, in the HIV/AIDS epidemic? Few analyses of this type have been conducted this far, although one can visualise the challenge of choosing the optimal timing of policy as a technical problem with three key components: (a) the reduction in costs (in present discounted value terms) of waiting one more “time period” instead of implementing the policy immediately; (b) the added benefits of implementing the policy immediately in terms of the lower number of new HIV infections that will occur in the current period; and (c) the difference in the number of HIV infections (if any) from implementing the policy now versus one period later, in all future periods, other than the current period. 

To see this decision problem clearly, suppose that a country has two new HIV infections in time period 1, and that the number of new infections doubles each year in the absence of any intervention. That is, there are two new HIV infections in year one, four in year two, eight in year three, 16 in year four, 32 in year five, and so on. Now suppose there is an intervention that costs C in each year starting from the date it is first implemented, and which reduces the number of new HIV infections by one-half each year. Then, if the intervention is introduced in year 1, the time profile of new HIV cases is 1,1,1,1… If the intervention is introduced in the second year, time profile is 2,2,2… If the money value of an averted HIV case is V, then the problem of waiting one more period before implementing the policy becomes one of choosing V/r, where r is the added discount rate and V/r is the added discounted benefit of implementing the policy immediately, versus waiting one more period, and C. If V/r exceeds C, then the policy ought to be undertaken immediately. Otherwise, it might be worthwhile to wait one more period. 

The above example highlights two issues, the first of which has relevance for many of economic evaluation methods being currently used to assess HIV prevention programmes. In particular, it highlights the fact that assessments that do not include the full set of secondary infections caused by an initial case of averted HIV infection in the calculus will provide misguided results. In the example above, if one includes only the initial averted case for all time periods, once the policy introduced in time period 1, then the net benefit to the policy would be V/r – C/r. On the other hand, if one were to consider the introduction of prevention programme in period 2 in isolation, the net benefit (at time period 1) would be [(2V/r) – C/r]/(1+r), since the policy would avert 2 cases in each year not including secondary infections, starting with time period 2. Thus, the second policy will appear to be better on grounds of a higher net present value if calculated in this manner, provided the rate of discount is less than 100 per cent. The obvious problem with this calculation is that the first policy, in fact, avoids even more cases than the second on an annual basis, because it also has an impact on the base of infections one starts out with. The correct calculation for comparing the two approaches is that reported in previous paragraph, which demonstrates this fact unambiguously. 

A second related point has to do with the circumstances when acting early is likely to be cost-beneficial. In the example above, we paid no attention to prevention activity of the individuals themselves. In general, whenever individual behaviour is not too important as a factor in neutralising policy, early intervention may be useful. A blood-screening programme is an obvious candidate, particularly at low initial rates of HIV infection where individual responses for prevention may be lacking. Less clear are the implications for a programme such as HIV/AIDS information provision, which may not influence behaviour at risk for HIV infection if people perceive the risks from such behaviour to be small, as is likely to be the case during the early stages of the HIV/AIDS epidemic. In (relatively) high prevalence settings, information provision may actually influence behaviour towards reducing risky behaviour, and this might suggest the introduction of such a policy a bit later. However, to the extent that people are more likely to undertake “private” preventive action, including the acquisition of information about HIV, once the AIDS epidemic starts becoming more visible, and not during its early stages, publicly supported prevention programmes may have to be introduced early. The net effect on the timing of policy will depend on the relative strengths of these two effects. 

Analyses conducted for Sri Lanka (a low HIV prevalence country) reported earlier suggest that it may be cost-beneficial to set up blood screening programmes and to introduce disposable instead of reusable equipment even when HIV prevalence rates in the population are extremely low (0.08 per cent), provided secondary infections are included in calculations (Bloom et al 1997). This suggests introducing these programmes very early in the epidemic. 

Address for correspondence: This email address is being protected from spambots. You need JavaScript enabled to view it.  

Notes I am grateful to David Bloom and Jaypee Sevilla for their help and comments at several points in the research for this paper, and to the United Nations Development Programme for financial support. The usual disclaimer applies.] 

1. The history of tuberculosis provides a salutary lesson here, as complacency over tuberculosis in the US in the 1980s reversed the positive progress that had been made in eradicating the disease over the previous 50 years. See Bloom, River Path Associates and Fang (2001).

2. However, this could be counteracted somewhat by a need to acquire precautionary savings to meet the increased risk of health care expenditures associated with the risk of HIV/AIDS [Bloom and Mahal 1997].

3. They may also experience reduced levels of nutrition as well.

4. For instance, on equity grounds.

5. Specifically, their empirical equations can be derived in a Solow-Swan growth framework, provided countries are close to their steady state, and that steady state outputs depend on the explanatory variables in question.

6. In any event, estimates for different years are not independent and are typically derived from epidemiological models. For econometric implications, see for example, Bhargava et al (2001).

7. Some qualifications to this claim are necessary. It is not obviously true that HIV prevalence rates among women visiting ante-natal clinics are representative of prevalence rates among men. Are these rates reasonably representative of HIV-prevalence rates among women in the reproductive age-group? Many people do not visit ante-natal clinics, many of which are located in urban areas, so they are likely to have lower proportions of rural women. Second, young women at high risk for HIV infection do not visit ante-natal clinics on account of stigma. Third, HIV seems to lead to reduced fertility rates, so that visitors to ante-natal clinics would disproportionately represent individuals with lower HIV prevalence, relative to the whole group. Finally, most of the sentinel surveillance sites are located in public facilities, so that there would be socio-economic differences between women whose blood is tested for HIV in the sentinel surveillance sites, and those who visit private facilities, and are not covered. These caveats suggest that sentinel surveillance data would underestimate HIV prevalence among women in reproductive age groups. The relatively greater ease with which HIV is transmitted to women, compared to men, and the increasing role of heterosexual sex in HIV transmission in developing countries, suggests however, at least one factor leading to a bias in the other direction when using ante-natal clinic HIV data to assess HIV prevalence rates among all adults, men or women.

8. Results available from the author. 

References

Arndt, Channing, and Jeffrey Lewis (2000): ‘The Macro Implications of HIV/ AIDS in South Africa: A Preliminary Assessment’, Draft, The World Bank, Washington, DC.

– (2001): ‘The HIV/AIDS Pandemic in South Africa: Sectoral Impacts and Unemployment’, Journal of International Development, 13:427-49.

Arora, Suchit (2001): ‘Health, Human Productivity and Long-term Economic Growth’, Journal of Economic History, 61(3): 699-749.

Barnett, Courtney, Manjiri Bhawalkar, A K Nandakumar and Pia Schneider (2001): ‘The Application of the National Health Accounts Framework to HIV/AIDS in Rwanda’, Special Initiatives Report No 31, Abt Associates, Bethesda.

Barnett, Tony and Piers Blaikie (1992): AIDS in Africa: Its Present and Future Impact, The Guilford press, New York.

Barnett, Tony and Alan Whiteside (2000): ‘HIV/AIDS and Development: Case Studies and a Conceptual Framework’, Draft, University of Natal, Health Economics and HIV/AIDS Research Division, Durban, South Africa.

Barro, Robert (1991): ‘Economic Growth in a Cross Section of Countries’, Quarterly Journal of Economics, May, 407-43.

Barro, Robert and Xavier Sala-I-Martin (1995): Economic Growth, McGraw Hill, New York.

Basu, Alaka, Devendra Gupta and Geetanjali Krishna (1997): ‘The Household Impact of Adult Morbidity and Mortality: Some Implications of the Potential Epidemic of AIDS in India’ in David Bloom and Peter Godwin (eds), The Economics of HIV and AIDS: The Case of South and Southeast Asia, Oxford University Press, New Delhi.

Bechu, Natalie (1998): ‘The Impact of AIDS on the Families of Cote d’Ivoire: Changes in Consumption among AIDS Affected Households’ in Martha Ainsworth, Lieve Fransen and Mead Over (eds), Confronting AIDS: Evidence from the Developing World, The World Bank, Washington, DC.

Bharat, Shalini (1999): HIV/AIDS Related Discrimination, Stigmatisation and Denial in India, Tata Institute of Social Sciences, Unit for Family Studies, Mumbai, India.

Bhargava, Alok, Dean Jamison, Lawrence Lau and Christopher Murray (2001): ‘Modelling the Effects of Health on Economic Growth’, Journal of Health Economics, 20:423-40.

Biggs, Tyler and Manju Shah (1997): ‘The Impact of the AIDS Epidemic on African Firms’, RPED Discussion Paper #72, The World Bank, Africa Region, Washington, DC.

Bloom, David and Sherry Glied (1991): ‘Benefits and Costs of HIV Testing’, Science, 239(4840): 604-10.

– (1993): ‘Economic Implications of AIDS in Asia’ in David Bloom and Joyce Lyons (eds), Economic Implications of AIDS in Asia, Oxford University Press, New Delhi. 

Bloom, David and Joyce Lyons (eds) (1993): Economic Implications of AIDS in Asia, Oxford University Press, New Delhi.

Bloom, David, Neil Bennett, Ajay Mahal and Waseem Noor (1996): ‘The Impact of AIDS on Human Development’, Draft, Columbia University, Department of Economics, New York, NY.

Bloom, David and Ajay Mahal (1996): ‘Economic Implications of AIDS in Asia’, Draft, Columbia University, Department of Economics, New York, NY.

– (1997a): ‘Does the AIDS Epidemic Threaten Economic Growth?’, Journal of Econometrics, 77:105-24.

– (1997b): ‘HIV/AIDS and the Private Sector’, Draft, Harvard School of Public Health, Department of Population and International Health, Boston, MA.

Bloom, David, Ajay Mahal, Lene Christiansen, Amala de Silva, Soma de Silva, Malsiri Dias, Saroj Jayasinghe, Swarna Jayaweera, Soma Mahawewa, Thana Sanmugam and Gunatillake Tantrigama (1997): ‘Socioeconomic Dimensions of AIDS in Sri Lanka’ in David Bloom and Peter Godwin (eds), The Economics of HIV and AIDS: The Case of South and Southeast Asia, Oxford University Press, New Delhi.

Bloom, David and Jeffrey Williamson (1998): ‘Demographic Transitions and Economic Miracles in Emerging Asia’, The World Bank Economic Review, 12(3): 419-55.

Bloom, David and David Canning (2000): ‘The Health and Wealth of Nations’, Science 287, February 18, 1207, 1209.

Bloom, David, Ajay Mahal, Jaypee Sevilla and River Path Associates (2001): ‘AIDS and Economics’, Draft, Harvard School of Public Health, Department of Population and International Health, Boston, MA.

Bloom, David, David Canning and Jaypee Sevilla (2001a): ‘The Effect of Health on Economic Growth: Theory and Evidence’, Working paper #8587, National Bureau of Economic Research, Cambridge, MA.

– (2002b): ‘Economic Growth and the Demographic Transition’, Working paper #8685, National Bureau of Economic Research, Cambridge, MA.

Bloom, David, Ajay Mahal, Larry Rosenberg, Jaypee Sevilla, David Steven and Mark Weston (2004): Asia’s Economics and the Challenge of AIDS, Asian Development Bank, Manila.

Bollinger, Lori and JohnStover (1999): ‘The Economic Impact of AIDS in Swaziland’, Draft, Futures Group International, Washington, DC.

Bonnel, Rene (2000): ‘HIV/AIDS and Economic Growth: A Global Perspective’, South African Journal of Economics, 68(5):820-55.

Chin, James and S K Lwanga (1991): ‘Estimation and Projection of Adult AIDS Cases: A Simple Epidemiological Model’, Bulletin of the World Health Organisation, 69-399-406.

Cuddington, John (1993a): ‘Modelling the Macroeconomic Effects of AIDS, with an Application to Tanzania’, World Bank Economic Review ,7(2):173-89.

– (1993b): ‘Further Results on the Macroeconomic Effects of AIDS: The Dualistic Labour Surplus Economy’, World Bank Economic Review, 7(3):403-17.

Cuddington, John and Hancock, John (1994a): ‘Assessing the Impact of AIDS on the Growth Path of the Malawian Economy’, Journal of Development Economics, 43:363-68.

– (1994b): ‘The Macroeconomic Impact of AIDS in Malawi: A Dualistic Labour Surplus Economy’, Journal of African Economies, 4(1):1-28.

Dayton, Julia (1998): ‘World Bank HIV/AIDS Interventions: Ex-ante and Ex-post Evaluation’, Discussion paper #389, The World Bank, Washington, DC.

Gallup, John and Jeffrey Sachs (2000): ‘The Economic Burden of Malaria’, Working paper #52, Centre for International Development, Harvard University, Cambridge, MA.

Gillespie, Stuart (1989): ‘Potential Impact of AIDS on Farming Systems: A Case Study of Rwanda’ Land Use Policy 6(4):301-12.

Giraud, Patrick (1993): ‘The Economic Impact of AIDS at the Sectoral Level: Developing an Assessment Methodology and Applying It to Thailand’s Transport Sector’ in David Bloom and Joyce Lyons (eds), Economic Implications of AIDS in Asia, Oxford University Press, New Delhi.

Guinness, L, and A Alban (2000): ‘The Economic Impact of AIDS in Africa: A Review of the Literature’, UNAIDS background paper for ADF 2000, UNAIDS, Geneva.

Hamoudi, Amar, and Jeffrey Sachs (1999): ‘Economic Consequences of Health Status: A Review of the Evidence’, Working paper #30, Centre for International Development, Harvard University, Cambridge, MA. Haney, D (2001): ‘AIDS Rampant among Young Black Men’, Associated Press, February 5.

Haslwimmer, M (1994): ‘The Social and Economic Impact of HIV/AIDS on Nakambala Sugar Estate’ Food and Agricultural Organisation and the Zambia Sugar Company Limited, Lusaka.

Janjareon, Wattana (1998): ‘The Impact of AIDS on Household Composition and Consumption in Thailand’ in Martha Ainsworth, Lieve Fransen and Mead Over (eds), Confronting AIDS: Evidence from the Developing World, Oxford University Press, New Delhi. Kambou, Gerard, Shanta Devarajan and Mead Over (1992): ‘The Economic Impact of AIDS in an African Countary: Simulations with a Computable General Equilibrium Model of Cameroon’, Journal of African Economies 1(1): 109-30.

Kremer, Michael (1996): ‘Integrating Behavioural Choice into Epidemiological Models of AIDS’ Quarterly Journal of Economics, May: 549-73.

Kwaramba, P (1997): ‘The Socioeconomic Impact of HIV/AIDS on Communal Agricultural Systems in Zimbabwe’, Working paper #199, Farmers Union, Friedrich Ebert Stiftung Economic Advisory Project, Harare, Zimbabwe.

MacFarlan, Maitland, and Silvia Sgherri (2001): ‘The Macroeconomic Impact of HIV/AIDS in Botswana’, Working paper #WP/01/80, International Monetary Fund, Washington, DC.

Mahal, Ajay (2002): ‘HIV and Human Development: An Analysis’, Draft, Harvard School of Public Health, Department of Population and International Health, Boston, MA.

Mankiw, Gregory, David Romer and David Weil (1992): ‘A Contribution to the Empirics of Economic Growth’, Quarterly Journal of Economics 107(2): 407-37.

McDonald, Scott, and Jennifer Roberts (2001): ‘AIDS and Economic Growth: A Panel Data Analysis’, Department of Economics, University of Sheffield, Sheffield, UK.

Meltzer, David (1992): ‘Mortality Decline, the Demographic Transition and Economic Growth’, PhD dissertation, Department of Economics, University of Chicago, Chicago, IL.

Menon, Rekha, Maria Wawer, Joseph Konde-Lule, Nelson Sewankambo, and Chuajun Li (1998): ‘The Economic Impact of Adult Mortality on Households in Rakai District, Uganda’ in Martha Ainsworth, Lieve Fransen and Mead Over (eds), Confronting AIDS: Evidence from the Developing World, The World Bank, Washington, DC.

Merson, Michael (1992): ‘The AIDS Epidemic in Asia: The Reality, the Opportunity, and the Challenge’, Keynote Address, Delivered at the Second International Congress on AIDS in Asia and the Pacific, New Delhi, November 8-12.

Misra, Geetanjali, Ajay Mahal and Rima Shah (2000): ‘Protecting the Rights of Sex Workers: The Indian Experience’, Health and Human Rights 5(1):88-115.

Nampanya-Serpell, Namposya (2000): ‘Social and Economic Risk Factors for HIV/AIDS Affected Families in Zambia’ Paper presented at the AIDS and Economics Symposium, Durban South Africa, July 7-8.

Over, Mead (1992): ‘The Macroeconomic Impact of AIDS in Sub-Saharan Africa’ Technical working paper #3, Africa Technical Department, The World Bank, Washington, DC.

Philipson, Tomas, and Richard Posner (1993): Private Choices and Public Health: The AIDS Epidemic in an Economic Perspective, Harvard University Press, Cambridge, MA.

Pitayanon, Sumalee, Sukontha Kongsin, and Wattana Janjareon (1997) ‘The Economic Impact of HIV/AIDS Mortality on Households in Thailand’ in David Bloom and Peter Godwin (eds), The Economics of HIV and AIDS: The Case of South and Southeast Asia, Oxford University Press, New Delhi.

Preston, Samuel (1976): Mortality Patterns in National Populations, Academic Press, New York.

Rugalema, G (1999): ‘HIV/AIDS and the Commercial Agricultural Sector of Kenya: Impact, Vulnerability, Susceptibility and Coping Strategies’, Food and Agricultural Organisation of the United Nations, Sustainable Development Department, Rome.

Sala-i-Martin, Xavier (1996): ‘The Classical Approach to Convergence Analysis’, The Economic Journal, 106(July): 1019-36.

Shepard, Don (1998): ‘Levels and Determinants of Expenditures on HIV/ AIDS in Five Developing Countries’ in Martha Ainsworth, Lieve Fransen and Mead Over (eds), Confronting AIDS: Evidence from the Developing World, The World Bank, Washington, DC.

Solon, Orville, and Angelica Barrozo (1993): ‘Overseas Contract Workers and the Economic Consequences of HIV and AIDS in the Philippines’ in David Bloom and Joyce Lyons (eds) Economic Implications of AIDS in Asia, Oxford University Press, New Delhi.

Solow, Robert (1956): ‘A Contribution to the Theory of Economic Growth’ Quarterly Journal of Economics 70(1): 65-94.

– (1957): ‘Technical Change and the Aggregate Production Function’ Review of Economics and Statistics 39(August): 312-20.

Strauss, John and Duncan Thomas (1998): ‘Health, Nutrition and Economic Development’, Journal of Economic Literature 36(June): 766-817.

Summers, Robert and Alan Heston (1991): ‘The Penn World Tables (Mark 5): An Expanded Set of International Comparisons, 1950-1988’, Quarterly Journal of Economics 106(2): 327-68.

United Nations Development Programme (UNDP) (1992): ‘The HIV Epidemic as a Development Issue’, Pamphlet, United Nations, New York, NY.

van Zon, Adriaan, and Joan Muysken (2001): ‘Health and Endogenous Growth’, Journal of Health Economics 20:169-85.

Viravaidya, Mechai, Stasia Obremsky, and Charles Myers (1993): ‘The Economic Impact of AIDS on Thailand’ in David Bloom and Joyce Lyons (eds), Economic Implications of AIDS in Asia, Oxford University Press, New Delhi.

Yang, Bong-Min (1993): ‘The Economic Impact of AIDS on the Republic of Korea’ In David Bloom and Joyce Lyons (eds), Economic Implications of AIDS in Asia, Oxford University Press, New Delhi. 

Reprinted from the Economic and Political Weekly, March 6, 2004 1049-63 with the permission of the author and editor