1. Introduction

The above quote suggests that foreign direct investment (FDI) is important for economic development and that HIV/AIDS impedes FDI.³ Indeed, the negative effect of HIV/AIDS on FDI is consistent with the 2003 World Economic Forum (WEF) survey, where about 33% of business leaders in developing countries that participated in the survey reported that HIV/AIDS has “affected their country’s access to FDI in the past five years.”⁴ It is important to note that examining the effect of HIV/AIDS on FDI is crucial for countries in Sub-Saharan African (SSA) for the following two reasons. First, FDI is crucial for poverty reduction in SSA, however, the region has received very little FDI and the investments are concentrated in only a few countries (Asiedu and Gyimah-Brempong, 2008; see Table 2). Second, a majority of the people infected with HIV/AIDS live in SSA (about 66% of the 33 million people infected with the disease live in SSA); the region has the highest new infection rates (in 2009, about 71% of newly infected adults lived in SSA); and the disease is the leading cause of death among adults in the region (UNAIDS, 2010).⁵

This paper theoretically and empirically examines the relationship between HIV/AIDS and FDI. Specifically, we construct a model that examines the relationship between HIV/AIDS and FDI and employ panel data from 41 countries in SSA over the period 1990–2008 to test the implications of the model. For the theoretical model, we show that HIV/AIDS has a negative and non-linear effect on FDI, and we derive the conditions under which the non-linear effect is negative and diminishing. For the empirical analysis, we find that HIV/AIDS has a negative effect on FDI, the adverse effect occurs even when the prevalence rate is low (as low as 0.1%), and the negative effect diminishes as the HIV/AIDS prevalence rate rises. We also show that our results are robust: they hold when we control for market size, openness to trade, infrastructure development, and natural resource intensity in host countries.

The idea that HIV/AIDS may have an adverse effect on FDI is not implausible, for the simple reason that healthy workers are more productive than sick workers. As a consequence, one would expect the health status of workers in host countries to have an impact on FDI. It is therefore surprising that although there is an extensive empirical literature on the determinants of FDI to developing countries in general, and to Sub-Saharan African countries in particular, research on how the health status of workers in host countries affects FDI is scant and also recent.⁶ We found only three papers that have included a measure of health as a determinant of FDI: Alsan et al. (2006), Azemar and Desbordes (2009), and Ghosh and Francesco (forthcoming). Alsan et al. (2006: 613) note that “To date, however, a relationship between population health and FDI has not been established in the empirical literature.” They also assert that “To the best of our knowledge, this represents the first empirical investigation evaluating whether health directly affects FDI, ceteris paribus,” (2006:614). This paper contributes to the thin literature on this important topic. Alsan et al. (2006) analyze the effect of life expectancy on FDI flows to developing countries.

The remainder of the paper is organized as follows. Section 2 presents the literature survey and Section 3 discusses some stylized facts about HIV/AIDS and FDI flows to SSA. Section 4 presents the theoretical model, Section 5 describes the data and the variables employed in the regressions, Section 6 reports of the empirical results and Section 7 concludes.

2. Literature review

As indicated in the introduction, our literature review revealed only three papers that have included a measure of health as a determinant of FDI. Alsan et al. (2006) analyze the effect of life expectancy on FDI flows to developing countries. Azemar and Desbordes (2009) analyze the indirect effect of HIV, tuberculosis and malaria on FDI to SSA countries. Specifically, they first analyze the impact of HIV, tuberculosis and malaria on life expectancy, and then examine the effect of life expectancy on FDI. Ghosh and Francesco (forthcoming) analyze the effect of communicable diseases on FDI to developing countries. All the studies conclude that good health promotes FDI.

This paper extends the health-FDI literature in three ways. First, unlike previous studies, the empirical analysis is based on a theoretical model with microfoundations. In addition, the model reflects two of the most important channels through which HIV/AIDS can affect FDI: a decline in productivity and an increase in absenteeism.⁷ Second, the paper addresses some specification and endogeneity issues not considered in previous studies. Specifically, we find that the relationship between HIV/AIDS and FDI is non-linear, while in Azemar and Desbordes (2009) HIV is treated as linear in their regressions. With regard to endogeneity, the analysis of Oster (2009) and Herzer and Nunnenkamp (2012) points to the possibility that FDI may have a causal effect on HIV/AIDS, suggesting that the relationship between FDI and HIV/AIDS may be bi-directional.⁸ Also, the data on HIV prevalence rates are likely to exhibit measurement errors (Bloom and Canning, 2008).⁹ These two potential problems suggest that endogeneity may be a concern. In our empirical analysis, HIV/AIDS is treated as endogenous. Another issue is that several studies have found that past FDI is correlated with current FDI as in Busse and Hefeker (2007) and Asiedu et al. (2009). However, Alsan et al. (2006), Azemar and Desbordes (2009) and Ghosh and Francesco (forthcoming) do not take into account the persistent nature of FDI. We introduce the lagged FDI into the regression to capture its persistence. The system GMM estimator that we employ for our estimations accounts for unobserved country-specific effects, mitigates any potential endogeneity problems, permits the inclusion of lagged dependent variable as well as endogenous explanatory variables, and also accommodates panel data with short time periods.

The third contribution of the paper to the health-FDI literature is that it is the first paper to analyze the direct effect of HIV/AIDS on FDI. We focus on the effect of HIV/AIDS on FDI for the following reasons. First, HIV/AIDS is a global epidemic and the disease “has inflicted the single greatest reversal in human development in modern history” (UNDP, 2005: 10). Furthermore, the adverse effect of ill health on FDI is more profound for HIV/AIDS than other chronic and infectious disease.¹⁰ Finally, we argue that life expectancy is not appropriate for analyzing the impact of the health of workers on FDI for the following reasons: (i) it reflects the health status of the total population, including children and the elderly, two groups that are not part of the labor force; (ii) it measures mortality rates—does not take into account morbidity (sickness) rates (Bloom et al., 2003); (iii) it is highly correlated with other country characteristics that affect FDI flows (e.g., income per capita, infrastructure development), and this makes it difficult to isolate the effect of health on FDI.

The paper also contributes to the small but increasing literature that examines the effect of HIV/AIDS on the growth rate of income per capita. The findings from these studies are mixed. Some studies find a positive effect of HIV/AIDS on growth (e.g.,Young, 2005; McDonald and Roberts, 2006), others find a negative effect (e.g., Juhn et al., 2008; Kalemli-Ozcan and Turan, 2010), and some conclude that HIV/AIDS has no significant effect on income per capita growth (e.g., Bloom and Mahal, 1997).¹¹ Haacker (2004) asserts that the existing studies underestimate the effect of HIV/AIDS on growth. He argues that the existing studies analyze only the direct effect of HIV/AIDS on growth, and do not take into consideration the fact that the disease can impact growth indirectly by decreasing FDI. This paper contributes to the literature by examining one of the channels through which HIV/AIDS may indirectly affect economic growth.

3. HIV and FDI to Sub-Saharan Africa: some stylized facts

This section provides some stylized facts about HIV and FDI flows to SSA from 1990–2013. Table 1 shows the HIV prevalence rate for ages 15–49, and Tables 2 and 3 show net-FDI flows and FDI as a share of GDP to SSA, respectively.

4. A model of HIV/AIDS and FDI

As pointed out earlier, two of the most important channels through which HIV/AIDS can affect FDI is a reduction in labor productivity and an increase in absenteeism. We incorporate the productivity/absenteeism effect of the epidemic on FDI by including three features in the model. First, we assume that infected workers have lower productivity than uninfected workers. Second, infected workers take days off from work when they fall sick.¹² The third feature is related to a negative externality of the disease. The morbidity (sickness) rate and the mortality (death) rate of HIV/AIDS are quite high. The sickness and death of infected co-workers may affect the working environment and lower morale at the workplace. In addition, uninfected workers may be reluctant to interact with infected workers for fear of being infected. Clearly, these factors lead to a reduction in overall productivity at the workplace. We incorporate such a negative externality by assuming that infected workers lower the productivity of all workers. Thus the total (labor-augmented) productivity can be decomposed into two parts: the “idiosyncratic” part which is determined by the individual’s HIV status and “non-idiosyncratic” part which is determined by the overall health of the entire labor force. Therefore, the HIV prevalence rate enters the goods production function as a determinant of effective units of labor.¹³ Later we show that the existence of such an externality is crucial in establishing the non-linear effect of HIV/AIDS on FDI.

We now describe the model. Consider an environment where a host country maximizes utility by choosing consumption, labor and foreign capital (i.e., FDI). Specifically, the country uses capital, k, and labor, n, to produce output according to the following production function:

Here, k = k_d + k_f, where k_d is domestic capital and k_f is foreign-owned capital. Assume that capital depreciates completely, thus k_f is simply FDI. We assume that k_d is exogenously determined and k_f earns a rate of return r, determined by the world capital market.

The population of workers in the host country is normalized to unity. Let h ∈ (0,1) be the share of the workers that are HIV positive. We use subscripts “1” and “2” to refer to uninfected and infected workers, respectively. Denoted by n₁ the labor supply of uninfected workers, and n₂ the labor supply of infected workers. Then the effective labor supply is equal to (1 − h)n₁ and αhn₂, respectively, with α ∈ (0,1) reflecting the assumption that infected workers have lower productivity. To take into account the negative externality, we introduce a discount factor on the quality of labor supplied by all workers, γ(h), with γ(h) ∈ (0,1). We assume γ′(h) < 0. This assumption implies that the negative externality gets bigger as the prevalence rate increases. Note that α pertains to only infected workers (i.e., the “idiosyncratic” productivity parameter) but γ(h) affects all workers (i.e., the “non-idiosyncratic” productivity parameter). Thus, the aggregate effective labor supply (adjusted for quality) is given by

For simplicity, we assume that the utility function is separable in consumption, c, and leisure time, l, and takes the form:

where η > 0. The time endowment is normalized to be equal to one. Let s be the sick time taken off by infected workers, then the effective time endowment of infected workers is 1 − s. Therefore the time constraint faced by healthy and infected workers are given by Eqs. (4) and Eqs. (5), respectively:

Finally the resource constraint for the host country is:

The host country chooses c₁, c₂, l₁, l₂, and k_f to maximize:

subject to Eqs. (1)–(6).

Let λ be the Lagrangian multiplier associated with the resource constraint (Eq. (6)). The first order conditions are given by:

Clearly Eqs. (8) and Eqs. (9) imply that c₁ = c₂ = c.¹⁴ Eq. (10) shows that the marginal product of capital equals the world interest rate, yielding a constant capital–labor ratio:

The next two first-order conditions reflect that workers’ marginal rate of substitution between consumption and leisure equals their marginal product of labor. It is straightforward that these two equations imply

Since infected workers have lower productivity (α < 1) and take sick time off (s > 0), thus n₁ > n₂, infected workers supply less labor. Combining Eqs. (6) and (10), and substituting Eqs (4), (5), and (13) into Eq. (2), we end up with the following two equations determining n and l₁ in the equilibrium:

Totally differentiating the above with respect to h yields And the effect of HIV/AIDS on FDI is summarized by and

Clearly dk_f/dh < 0 as long as any one of the following three conditions is satisfied: (i) α < 1; (ii) s > 0 or (iii) γ′ (h) < 0. This implies that HIV/AIDS has an adverse effect on FDI: (i) when infected workers have a lower productivity relative to uninfected workers, α < 1; or (ii) when infected workers take sick time off, s > 0; or (iii) when the negative externality generated by HIV/AIDS increases with the prevalence rate, γ′ (h) < 0.

From Eq. (15) we can see that d²k_f/dh² > 0 if and only if

This condition is easy to satisfy. For example, if γ(h) is concave, then γ″(h) > 0 and hence d²k_f/dh² > 0 Note that defining γ as a function of h is critical for the non-linearity result. It is still possible to have dk_f/dh < 0 as long as α < 1 or s > 0. However d²k_f/dh² will be equal to 0 once γ is independent of h, suggesting that the effect of HIV/AIDS on FDI, although still negative, will be the same across different values of the prevalence rate. In summary, our model implies that HIV/AIDS has a non-linear negative effect on FDI, and the effect is diminishing under certain condition.

5. The data and the variables

The empirical analyses utilize panel data from 41 countries in Sub-Saharan Africa over the period 1990–2008. We focus on the post-1990 years because although the HIV virus was identified in the 1980s, global awareness of the disease began in the 1990s. Furthermore, the HIV data prior to 1990 are deemed unreliable (UNAIDS, 2010).

The dependent variable is fdi where fdi = ln(1 + FDI), and FDI is net FDI inflows.¹⁵ As it is standard in the literature, we average the data over three years to smooth out cyclical fluctuations. We use the percentage of adults that are HIV positive in a country, hiv, to capture the severity of the HIV/AIDS epidemic. Table 1 shows the value of hiv averaged from 1990 to 2008, for the countries in our sample.

6. Empirical results

We proceed in two steps. We first estimate the benchmark model where we control for openness to trade, market size and infrastructure in host countries. In addition, we determine the share of countries for which HIV/AIDS is negatively correlated with FDI. In step 2, we test whether our results are robust when we control for natural resource intensity in host countries.

We estimate a linear dynamic panel-data (DPD) model to capture the effect of previous FDI flows on current flows. In particular, we estimate the equation:

where i refers to countries, t to time, fdi = ln(1 + FDI), hiv is HIV adult prevalence rate, Z is a vector of control variables, and θ_i and λ_t are country-specific and time-specific effects, respectively.

We now provide some details about our estimation strategy. First, we use the two-step GMM system estimator proposed by Blundell and Bond (1998), which is asymptotically efficient and robust to all kinds of heteroskedasticity. Second, as noted earlier, hiv, and therefore hiv² are likely to be endogenous. Therefore in our regressions, we specify hiv and hiv² as endogenous variables. Third, the control variables are treated as strictly exogenous. Finally, our regressions utilize only internal instruments—we do not include additional (external) instruments.¹⁷ Note that the effect of HIV/AIDS on FDI can be derived from Eq. (16) as ∂fdi/∂hiv = α + 2β × hiv. Therefore, the parameters of interest are α and β.

7. Conclusion

This paper theoretically and empirically examines the relationship between HIV/AIDS and FDI. We derive two main results. First, HIV/AIDS deters FDI to SSA countries. If FDI promotes growth and reduces poverty (Asiedu and Gyimah-Brempong, 2008), then our results suggest that policies that reduce the infection rate will enhance development in SSA countries. We also find that the adverse effect of HIV/AIDS on FDI is non-linear and that the negative effect diminishes as the HIV/AIDS prevalence rate decreases. This implies that all else equal, a reduction in the HIV/AIDS prevalence rate will be more beneficial to countries with low prevalence rates than countries with high prevalence rates. Thus, for high prevalent rate countries, an initial reduction in the prevalence rate would not generate sizable FDI; however, the flows will increase as the prevalence rate decreases.

With regard to policy, we draw from the 2014 World AIDS Day document which specifies five ways to reduce the epidemic: encourage HIV testing, reduce discrimination, promote HIV/AIDS education, encourage proper health care, and affirm support for people living with HIV/AIDS.¹⁸ We also note that access to good health care, in particular, access to antiretroviral (ARV) drugs boosts the immune system of HIV patients. In addition, there is some evidence that ARV drugs help increase labor supply, raise productivity and reduce absenteeism at the workplace (Rosen et al., 2008; Thirumurthy et al., 2008). Thus making ARV drugs available to infected workers may also mitigate the adverse effect of HIV/AIDS on FDI.