Chasing Success: Health Sector Aid and Mortality SVEN E. WILSON * Brigham Young University, Provo, USA Summary. — As many cases studies show, successful public health measures are being implemented in many places around the globe, and country-level mortality has fallen signiﬁcantly in recent decades in all but a few countries. Are the two linked? Does development assistance for health (DAH) improve, on balance, recipient countries’ mortality trajectory? Using a new data source containing DAH on 96 high mortality countries, the regression analysis shows no eﬀect of DAH on mortality. Other types of aid, including water develop- ment, also have no eﬀect. Economic growth, on the other hand, has a strong negative eﬀect on mortality. These ﬁndings conﬁrm and build upon recent work by Williamson (2008) and are shown to be robust to a variety of sensitivity analyses and alternative model speciﬁcations and estimation methods. This analysis also shows that the eﬀectiveness of DAH has not increased over time, even as the level of that funding has increased four- fold, though spending on infectious diseases and family planning may have caused small reductions in mortality. Furthermore, even though it is encouraging that DAH has tended to go where the need is highest, it also goes to states that have experienced the greatest mortality reductions in the recent past. In other words, DAH appears to be following success, rather than causing it. Ó 2011 Elsevier Ltd. All rights reserved. Key words — aid eﬀectiveness, mortality, health, public health 1. INTRODUCTION Eﬀective public health measures can save lives. That fact is not in question. The rich, healthy countries of the world were, at one time, as unhealthy as most impoverished nations today and, in many cases, even worse. Clean water, eﬀective sanita- tion, immunizations, antibiotics, rehydration therapy, malaria prevention and treatment, and better nutrition have, among many other tools, dramatically reduced mortality among in- fants, children and adults across the globe, both historically and in recent decades. In general, the public health practices necessary to implement wide-scale reductions in mortality are not particularly complex—though they can be multi-fac- eted—nor particularly expensive. Health, many believe, is an area where development assistance is likely to see positive results. Bold titles such as Millions Saved: Proven Success in Global Health (Levine, 2004) highlight apparent success at virtually eliminating mea- sles in southern Africa, in eradicating smallpox globally, in preventing STDs in Thailand, in reducing child mortality through vitamin A in Nepal, in successfully implementing rehydration therapies and reducing diarrheal deaths in Egypt, and many others. These types of case studies are very encouraging. 1 But these successes might lead to misplaced faith because they tell us little about whether development assistance for health (DAH), in aggregate, has had a positive impact on health outcomes. To answer that question correctly, we need to look at all DAH programs in all places, rather than identi- fying only highly successful cases. When all the DAH projects are added up, do they result in a meaningful increase in public health or health care in the recipient country and is DAH actu- ally improving health beyond where it would be without the assistance? Using a variety of speciﬁcations and alternative assumptions, I search diligently for statistical evidence that aggregate DAH reduces mortality. I ﬁnd—over and over again—no correlation. These investigations point overwhelm- ingly to noneﬀectiveness of DAH on mortality, whether using infant mortality (IMR), child (under 5) mortality (CMR), or life expectancy at birth (e 0 ). Williamson (2008) was the ﬁrst to look at DAH and mortal- ity, and she also found no eﬀect of DAH. But the analysis here extends what she did in many ways. The AidData database contains many large donors not found in the standard OECD-CRS data used by Williamson, and with that extended data, I explore extensive sensitivity analyses and alternative speciﬁcations she did not use, especially the latent growth model. 2 Furthermore, this analysis is the ﬁrst to estimate changes in aid eﬀectiveness across time and the eﬀectiveness of the various components of DAH, such as spending on infec- tious diseases. Encouragingly, there is some evidence that spending on HIV/AIDS, other infectious disease, and family planning have had statistically signiﬁcant (but very small) ef- fects on mortality. The most important innovation of this analysis is that I ex- plore the impact that endogenously determined DAH may have on the aid eﬀectiveness story in a manner that goes be- yond GMM models. 3 In particular, I estimate a simple aid allocation model and ﬁnd that reductions in mortality lead to large and signiﬁcant increases in incoming DAH ﬂows. This ﬁnding has two important implications. The ﬁrst is that DAH eﬀectiveness is not hiding behind the endogeneity problem, since the correlation between DAH and unobserved factors aﬀecting mortality appears to be in the opposite direction than would be necessary to undermine the noneﬀect of DAH. Sec- ond, the increasing ﬂows to countries that have already expe- rienced reductions in mortality suggest a possible reason for aid ineﬀectiveness that has received little attention: DAH ﬂows are responsive to country-level variables, but money is ﬂowing to where mortality is improving rapidly for reasons other than eﬀective aid programs. In sum, even though the mortality trajectories of countries have almost universally and nontrivially improved since 1975 across the globe, countries receiving high levels of DAH have done no better, on average, than countries receiving low levels of DAH. There are certainly many public health programs and projects that reduce mortality among those treated (I empha- * Final revision accepted: May 6, 2011. World Development Vol. 39, No. 11, pp. 2032–2043, 2011 Ó 2011 Elsevier Ltd. All rights reserved. 0305-750X/$ - see front matter www.elsevier.com/locate/worlddev doi:10.1016/j.worlddev.2011.07.021 2032