5. Main Results

Performance Equation: When does AIDS-aid work?
My final results do not provide a strong case that “good” policies are necessary for aid efficacy, however they ultimately support Easterly’s contention (2006) that large scale top down aid efforts will not be effective in decreasing the burden of HIV/AIDS.
Table 5 (Regressions 1 and 2) examines the effect of pure AIDS-aid on HIV rates and pure Policy conditions, without any controls. The results demonstrate significance between the Policy Index and HIV rates, and a lack of significance in the relationship between AIDS-aid and HIV rates. The trend accurately reflects reality, for example, in Uganda where despite the decline in HIV prevalence since the 1990s, recent surveillance data shows that HIV rates are presently increasing. The recent increased HIV rate coincides with reports of mismanaged Global Fund aid by Ugandan corrupt government officials in 2006, suggesting that political conditions of the recipient do affect the ability to successfully use allocated AIDS-aid.
However when controls are added, two of the Policy Index variables change signs (Regression 4) and the Index itself loses significance (Regression 5). When the Interaction Term of Aid and Policy is introduced it is significant within World Bank data (Regression 6) but not Global Fund (Regression 7), however it maintains its positive sign, as does the Policy Index, in all regressions in which it is included (Regressions 6-9). Bolded t-scores illustrate that Interaction Term significance is not far off, so the Term suggests that countries with the best policies (which will be large negative numbers) can benefit from a large decrease in HIV rates once aid is given. Most importantly, Aid alone and the Policy alone maintain even less significance, supporting the Easterly contention that neither aid nor policy alone will solve the problems of the poor.

Several factors may be at the root of insignificant coefficients. This study lacks necessary micro-level data, such as the effects of non-governmental and faith-based organizations that have dominated grassroots efforts to address HIV/AIDS, who are not included and may in fact affect the significance. I also use macro country-level data that may not be appropriate to measure the progress of a disease that disproportionately strikes poorer communities and at-risk populations within countries, for which we lack sufficient micro-level data. Missing data and variables may better explain the situation than the macro indicators I have employed. Without good macro measurements it is difficult to establish the effect of large-scale aid. However most importantly, HIV prevalence may be a poor choice for dependent variable in estimating the efficacy of AIDS-aid, as the power of aid and control variables can send the rate in different directions. By definition, a decline in HIV prevalence can mean more AIDS deaths (less people living with HIV) or more effective prevention that leads to less new HIV infection. Similarly, increases in HIV can imply poor health services exist and transmission continues unchecked, or may be due to the results of exceptional health services that prolong the lives of the HIV positive and thus have more living with the disease. I use prevalence as the dependent variable, despite the potential confusion, because the data is available and because it is this broad scale of national prevalence that is most often cited as an indicator of AIDS burden on a nation. A more precise study would have regressed aid, Policy, controls, and the Interaction Term on two dependent variables: the incidence of HIV, or number of new infections reported per year, and the AIDS-attributable mortality per year. Unfortunately this data is not collected in a standard way nor does a methodology exist that has been implemented globally. Social stigma and poor health systems keep new infections from being accurately reported. The convoluted implications of changes in HIV rate may also help explain the signs and significance of values within the Policy Index. While all start off with negative values (Regression 3), only the measure of Women’s Political Rights maintains its negative direction. Health Expenditure may become positive once aid is added because more strongly supported health budgets result in more people living longer with HIV. Women’s Political Rights expectedly remains negative, supporting Farmer et al contentions as well as instinctive assumptions that women as decision makers will address risks to females, including traditional practices of female circumcision, gender disparities, violence, spousal abuse and rape, and mother to child transmission. Logically, women as child-bearers and caretakers have inherently greater power to decrease HIV infection. For example HIV positive mothers can use an ARV called Nevirapine during and after childbirth to prevent transmission of HIV to their child. If women’s political rights increase, more pressure may be placed on programs and politicians to provide these specific medications to the many poor communities still lacking access, and potentially prevent all new infant HIV infections. I expected corruption to maintain its negative sign, as it is so prevalent in the literature as a barrier to aid efficacy. Even the former President of the Global Fund, Richard Feachem, recounted in 2005 an instance of Fund embezzlement, stating, “Can we bring the risk of corruption to zero? No, of course we can't….World Bank money is stolen; USAID [United States Agency for International Development] money is stolen. Anyone in the business of development finance has their money stolen on some occasion in some places, so we cannot make this a zero-risk game.” But once again the unique challenge of HIV/AIDS-aid can convolute what would otherwise seem basic logic. The provision of free life-extending antiretroviral drugs (ARVs) is at the core of many World Bank, Global Fund and PEPFAR initiatives. Since demand consistently outweighs supply of ARVs, it is those who bribe and not those who are legally eligible for free drugs, who receive treatment. A black market for ARVs has emerged, presently estimated by the World Bank to be worth $32 billion, because poor oversight of ARV distribution allows patients to register on multiple programs for free ARVs and sell their excess, and allows corrupt official to siphon drugs out of the supply chain. Akunyili (2006) explains that the market thrives because the anonymity (no program enrollment) circumvents the social stigma and long clinic lines that deter people from accessing available resources. This massive corruption sends HIV rates in both directions. Those who buy illegally miss out on the clinical counseling that explains necessary lifestyle changes to prevent future transmission, so prevalence increases. However the rate may simultaneously decrease because counseling also explains how to properly take drugs, so those who take them improperly may be dying. Also those who are incentivized to sell their drugs to buy other more necessary commodities for their families may die without the medication, also decreasing the HIV rate. The most significant control variables are Women’s Economic Rights and whether or not the Country is a PEPFAR Focus Country. Because Economic Rights is linked to financial wellbeing, and Political Rights is not, it may be the case that more women are able to afford ARVs and therefore live longer and increase the number of people living with HIV. It could also mean more new HIV cases if greater participation in and exposure to the labor force has increased faster than gender bias. Women may simply be more at risk of sexual abuse and HIV transmission. As expected, HIV Surveillance ratings imply that countries with more effective systems for surveying HIV will also help to decrease levels, and that countries with higher HIV rates will be more likely to see bigger increases in HIV than countries with low HIV rates. GDP and GDP per capita are unsurprisingly insignificant. Higher GDPs do lead to longer life spans and result in people being more risk averse than someone with a short lifespan and therefore will be more likely practice safe sexual behavior. However another recent study from Zimbabwe suggests that recent inflation has decreased GDP and GDP Per Capita, making the hiring of prostitutes and maintaining of multiple partners too expensive for many men. This results in less risky sexual behavior and less HIV transmission. Aggregate results are mainly driven by the World Bank aid data, as the Global Fund results were not at all statistically significant, which is most likely due to the small sample set. Alternative Explanation: Selection Analysis After determining policy conditions in the recipient nation that have the potential to affect AIDS-aid efficacy, the Treatment Effects Model determines the potential of a selection bias. Results indicate that there are ulterior motives to aid giving, and that pure aid, without supporting policies, may be ineffective not because of the lacking political capacity, but because donors systematically choose who to give money to, and for purpose other than HIV/AIDS relief. Table 6 shows results of the preliminary logit regression and measures of marginal effects that support the findings of Mascarenhas and BDM: significantly more aid is given to countries with higher terror ratings and high energy production, and aid is not significantly given based on need.
The Treatment Effects model calculates the decision to give AIDS-aid as an endogenous decision, conditional on the independent variables that determine change in HIV. Table 7 includes both equations of the Treatment Effects Model, the Outcome Equation and the Selection Equation, and the two sets of independent variables determine, respectively, the factors that actually influence the decision to give aid, and whether or not a country will receive aid.
The likelihood ratio test reveals a small probability that the chi-squared statistic is greater than the critical value for statistical significance, therefore I can reject the null hypothesis that the error terms in the selection equation and the outcome equation are uncorrelated. I therefore reject the null hypothesis that the independent variables do not determine AIDS contributions and confirm that there is indeed a selection bias in deciding the recipients of AIDS-aid. However we can see that this bias has less to do with the need, the HIV rate, and more to do with potential for policy concessions: terror and energy production. The results support BDM’s (2004) suggestion that aid is given bilaterally to facilitate policy concessions and Mascarenhas and Sandler (2006), who find no evidence of cooperative behavior among donors.
The HIV rate is insignificant as a driver of the decision to give aid, and we can eliminate the possibility of endogeneity, or that AIDS-aid is only given to those with high HIV rates. However, the results do not conclude that AIDS-aid is therefore meaningless. In fact, the significance of the binary AIDS-aid variable in the outcome equation proves that countries who have received aid may have been worse off had they not received any aid, which here correlates with a significant decrease in HIV prevalence.