Paper summary: Can the West Save Africa? by William Easterly

Question 1 - What is the evidence that aid works/has worked at all? That it has caused reductions in infant mortality, economic growth, or anything else?

• Little evidence connecting aid to growth; most analysis is highly problematic and the best analysis shows no effects. (1,2,3,4,5,6,7,8)
• "Micro/macro paradox": projects often show promising results for individuals, but this doesn't translate to the macro level. (10,11,13)
• Evidence weakly suggests that donors have been able to improve education rates, but with no corresponding impact on growth, perhaps due to "little job creation in the African formal private sector, which would normally be the employer of skilled labor." (39,40,41,42)
• Child mortality has fallen dramatically, and case studies / success stories imply that health aid and Western health technology deserve credit. (43,44,45)

Question 3 - Can we expect health aid to create economic growth? Can we expect economic aid to work in areas where health is poor?

• No clear connection between health aid (or any aid) and economic growth - see Question 1.
• Economic empowerment initiatives appear a failure at the macro level (43,44,45).
• No "success stories" in economic empowerment are mentioned. Microfinance and irrigation are not mentioned.

Question 4 - Why have some parts of the world emerged from poverty while others haven't? Is there anything aid can do to make the former more likely?

• Other parts of the world have emerged because of events within them, not thanks to the West. Details of these emergences are not discussed. (9)
• Aid has had little impact on growth (see Question 1) and the most ambitious "transformational" initiatives have basically no successes to point to (24-38).
• Easterly feels that the focus is best put on "marginal" rather than "transformational" projects - aim to improve lives/communities, not to "solve root causes." This is expressed in the abstract, conclusion, and throughout.

Question 5 - What are the risks of aid causing harm, and what evidence is there for their severity?

• Effect on corruption: studies are mixed but basically point to no effect. (29,30)
• Effect on democracy/accountability: studies are mixed (one showing no effect, another showing negative effects). "more money available to those who control the state will make them less likely to permit any democratic threat to their stay in power." (33)
• Other potential negative effects not discussed.

Question 9 - What are the interventions that are supported by rigorous evidence?

Education-related:

• A variety of school programs: meals, scholarships, extra textbooks & teachers, remedial education, citizens' report cards, the hiring of contract teachers, or increased oversight of local school committees. Not clear whether these have been shown to improve later-life outcomes or just school attendance and/or test scores. (16)
• PROGRESA in Mexico: cash grants to parents conditional on keeping their children in school and on regular health/nutritional checkups (accompanied by micronutrient supplements). Led to more years in school, and improved health outcomes including height, probability of illness, anemia. (17,45)
• Deworming & vitamin supplements (iron, A) have increased school attendance. (17)

Health-related:

• PROGRESA - see directly above
• Deworming has led to lower worm infection rates and improved height-for-age and weight-for-height, though the results are a bit mixed and potentially problematic. (45)
• "Breastfeeding, immunization against diarrheal diseases, micronutrient supplementation and oral rehydration therapy (ORT) have all been found to work in randomized trails in the fight against diarrhea. " (45)

Other:

• Agriculture: Selling a voucher earmarked for fertilizer purchases to the farmers right after the harvest has increased fertilizer usage. (18)
• Auditing village road projects has decreased "missing expenditures." (20)
• Newspaper campaigns on intended government transfers to schools improved the likeliness of the funding actually coming through, and also "showed up in increased enrollment and test scores for those schools." (20)
• "A random sample of Indonesian villages was given the right to democratically choose which aid projects would be implemented. Aid democracy did not have much effect on which projects were actually chosen, but it did dramatically improve villagers' satisfaction with the projects and their willingness to contribute." (21)

Banerjee 2008 and Duflo/Kremer 2008 cited as more thorough reviews on this question.

Question 10 - Which interventions are the big success stories?

Vertical health interventions, including "the elimination of smallpox, the near-eradication of river blindness and Guinea worm, the spread of oral rehydration therapy for treating infant diarrheal diseases, DDT campaigns against malarial mosquitoes (although later halted for environmental reasons), and the success of WHO vaccination programs against measles and other childhood diseases. " (43, 45)

Question 11 - What other interventions are considered to be extremely promising by the aid community, and why?

• Vertical health interventions are the biggest success story (see q's 9 and 10 for specifics).
• Maintenance of infrastructure (e.g., roads) is mentioned as important and underfunded. (15)

Question 12 - What is the track record of vertical vs. horizontal interventions?

• The aid community has cycled back and forth between the two a lot. (51)
• Vertical interventions have tangible successes to point to. (43) Track record of horizontal interventions not discussed.

Question 14 - To what extent has aid been able to create sustainable (as opposed to continually-donor-financed) improvements? What has held it back or worked well?

"This hope [sustainability] has turned out to be an illusion." (23)

Question 15 - What is the track record of aid through the govt vs. aid through NGOs?

• "Fungibility" of government aid (the extent to which it ends up paying for other government spending) is significant though not 100%. (22)
• Otherwise this question is not addressed.

Question 16 - What else should be kept in mind?

• Water infrastructure projects have underperformed simpler, behavior-based approaches to combating waterborne disease. (53)
• Privatization of water infrastructure has had promising effects according to one study. (54)
• User fees significantly reduce consumption of health services. (52)
• Corruption in the health & education systems is a major concern. (49,50)

Notes

1. 25: The literature only really became meaningful when the severe problem of reverse causality was addressed with the use of instrumental variables measuring political motivations for aid flows, as well as population size (a promising instrument, since as already noted, there is an exogenous small-country bias in aid such that smaller countries get higher aid per capita and higher aid as a ratio to their income). Boone 1996 was among the first to use such instruments and found zero effects of aid on investment and growth.

2. 25: Boone provoked further testing of the claim that aid raised growth. By far the most cited aid and growth study in the entire literature was Burnside and Dollar 2000 in the American Economic Review (BD). BD also found that aid had no effect on growth. However, they also tested an interaction term between aid and government policy, which was significantly positive in some of their regressions. Hence, they concluded that raised growth when the recipient had good policies (measured by the Sachs-Warner openness index, low inflation, and low budget deficits).

3. 26: What is notable given this strong policy influence is that the original results were both weak and fragile. BD used similar instruments as Boone for aid. Curiously, however, the significant positive effect of aid on growth (with "good policies") held only in their OLS regressions, not in 2SLS (they argued this was not a problem because they failed to reject exogeneity of aid). And even for the OLS coefficients, the positive growth effect of aid was significant (under good policies) in just 2 out of the 4 regressions they presented. Even this was after they excluded some outliers that went against the hypothesis (as they made transparent). Furthermore, Easterly, Levine, and Roodman 2003 (ELR) subsequently showed that the significance of the Burnside-Dollar aid-policy interaction term even in the OLS regressions where it was significant was not robust to some basic checks, such as adding new data that had become available since the original study. The distinguished academic panel led by Angus Deaton that reviewed World Bank research singled out the Burnside and Dollar results for criticism for lack of robustness and unconvincing identification strategy, and criticized the World Bank for overselling this particular result in its advocacy for more foreign aid (Deaton et al. (2006), pp. 52-57).

4. 26: This survey does not make more of an effort to survey all corners of this gigantic literature on aid and growth because the quality of most articles is poor. Most aid and growth articles fail to have a serious identification strategy.

5. 27: Population size is another promising candidate for an instrument because of the exogenous and pronounced small country bias in aid. Of course, it may not satisfy the exclusion restriction as population size might directly affect growth. The growth regression literature has extensively looked for population scale effects and has generally failed to find them. (Easterly and Kraay (2000) found no evidence that small population size affected growth performance on average. ) This is not a valid test of the exclusion restriction, but it does give some important reduced form information. Given that aid received as a ratio to income is strongly affected by population size, then if aid affected growth, we would expect smaller population size would be associated with higher growth. This is not there in the data, which partly reflects the poor growth performance of many small Pacific
and Caribbean islands and small African nations (all on average also very aid-intensive), so this is indirect evidence against a positive growth effect of aid.

6. 27: Werker et al 2008 is a recent paper that seems to have a believable and original identification strategy (aid from OPEC members to their poor Muslim allies, with the instrument being the price of oil interacted with a Muslim dummy). They also find a zero effect of aid on medium-term growth. There is still the doubt about this is whether it extrapolates to non-intra-Muslim aid.

7. 27: There is no clear theory as to what other control variables should be included, which
also weakens confidence in knowing what instruments satisfy the exclusion restriction. There is even doubt how the aid variable itself should be included (variants in the literature have included quadratic terms for aid/GDP, the log of aid, separating out aid loan repayments as a linear term combined with a log aid term, interacting aid with other variables, and many others), there is a serious data mining problem. Control variables in the literature have included such non-intuitive entries as Ethnic Fractionalization* Assassinations (BD). This is on top of the general data mining problem in growth regressions, in which Durlauf, Johnson, and Temple 2005 showed that 145 separate variables had been found to be significant in growth regressions with a typical sample of around 100 observations – and aid was not even one of the 145! The constructive thing that one can say is that data mining would manifest itself as a lack of robustness of results – changes in both the magnitude and significance of the aid coefficient. The failure of ELR to confirm BD is suggestive of this lack of robustness.

8. 28-29: To believe in a positive growth effect of aid, one needs to believe in the counterfactual that African growth would have been even worse in the absence of aid (not impossible, but harder to believe than if growth had been respectable). Given the figures shown above where the median aid received since independence (around 1965) was around 10 percent of GDP (Figure 4) and the per capita growth outcome was roughly zero percent (Figure 1), the implausible counterfactual implied by the "Big Push" coefficient of [0.2,0.5] is that the median African growth would have been -2 to -5 percent per capita in the absence of the aid "Big Push" since independence.

9. 29: even proponents of more rigorous randomized evaluation methods (to be discussed below) like Banerjee (2007) have some intuition about the limited role of aid in successes outside of Africa: "my sense is that {the dramatic reduction in world poverty between 1981 and 2001} was driven largely by events in India and China, where donors had very little impact." (Note 20 notes: Banerjee contrasts his interpretation to that of Goldin, Rogers, and Stern 2007, who attribute global poverty reduction to foreign aid, as an example of how stylized facts fail to induce consensus. However, Goldin, Stern, and Rogers 2002 was the original source of their conclusion, and this was not a research study but a World Bank advocacy effort ("The Case for Aid") to increase foreign aid in the run-up to the
UN Monterrey Conference in 2002.)

10. The literature discussed the “micro-macro” paradox, in which project returns to aid were high and yet as we have seen, the literature often failed to find an overall growth payoff to aid (see discussion in Doucouliagos and Paldam 2008). Later evidence on projects was not as favorable. The World Bank commissioned a study (known as “the Wapenhans report,” World Bank 1992) of World Bank project performance, as measured by the percent of projects classified as successful (again done by project managers and thus probably biased upward). Even with the probable upward bias, only 59 percent of projects in Africa were classified as “successful,” compared to 74 percent worldwide for World Bank projects.

11. 33: This literature has found many aid project interventions to have positive benefits and to be cost-effective (Banerjee 2008 and Duflo and Kremer 2008). Based on this encouraging evidence, Banerjee has written positively about the potential of such
(marginal) aid in his book Making Aid Work (2007).

12. As even a World Bank handbook said “Despite the billions of dollars spent on development assistance each year, there is still very little known about the actual impact of projects on the poor” (Baker 2000).

13. 39: Rajan and Subramanian 2008 pointed out that the micro-macro paradox still holds with the new randomized evaluation literature, with positive returns to micro projects yet apparently still zero macro growth payoff.

14. Duflo (2004) has argued that REs present a simple form of unambiguous evidence that is more likely to influence policy than other kinds of empirical development work. Here, the great success story is PROGRESA in Mexico, which was scaled up and continued under two different administrations due in part to the positive results from REs evaluating PROGRESA (Levy 2006). Of course, there were also political factors. Green (2005) found that, despite the attempt to de-politicize PROGRESA, municipalities that had previously voted for the party in power were more likely to have their localities enrolled in the program. Diaz-Cayeros et al. (2008) dispute that finding, but found that even a non-discretionary PROGRESA/ OPORTUNIDADES program paid off at the polls for the incumbent in both the 2000 and 2006 elections. They also point out that President Vicente Fox's decision to expand OPORTUNIDADES from rural areas to the cities
made political sense since his party's political base was urban.

15. 42: The bias against operations and maintenance in infrastructure has been known for
decades -- highlighted for example in World Bank (1981, 1988, 1994), with each succeeding report bewailing the failure to make progress since the previous report -- and it remains a problem today. The results are chronically potholed and cratered donor-financed roads, for example, always being reconstructed and then deteriorating again. This is another example of inability to learn from past mistakes.

16. 48-49: The randomization literature has found a number of aid interventions (both inside and outside Africa) to be effective in education. Kremer, Miguel, and Thornton (KMT) 2007 found that a merit scholarship for high school girls in Kenya seemed to induce greater study effort and increased the girls' test scores, and even had some externalities to boys' performance in the same classroom. In contrast, a program to give textbooks to students in Kenya did not increase test scores on average (Glewwe, Kremer, and Moulin 2007), a result that contrasts sharply with the previous literature (see e.g. Lockheed and Hanushek 1988; even education skeptics like Filmer and Pritchett 1999b argued there was a high payoff from textbooks). The authors argue that Kenyan schools were oriented towards the strongest students (whose test scores did improve), while the weaker students suffered from lack of English skills (textbooks were in English) and greater absenteeism of both pupils and teachers. Vermeesch 2003 found that a school meals program in pre-schools in Kenya raised attendance rates from 21 percent to 29 percent. It did not raise test scores on average, but did raise scores in schools with better-trained teachers. Note that the conditional conclusions in this paragraph are examples of ex-post slicing of the sample that this article discussed above as a sacrifice of econometric rigor in REs (including KMT 2007, where the merit scholarship worked in one sample site and not in the other). As noted earlier, Angrist et al. 2002 studied the effect of vouchers for private school distributed via a lottery in Colombia. The lottery winners had 0.12-0.16 additional years of schooling, test scores higher by 0.2 standard deviations, and higher secondary school completion (the latter confirmed in a follow-up study by Angrist, Bettinger, and Kremer 2006). This only scratches the surface of randomized studies on education inputs as shown by a statement like Kremer and Holla 2008:
Evidence is also now accumulating on the effectiveness of certain school inputs like extra
teachers and textbooks (Banerjee et al, 2005; Duflo, Dupas and Kremer, 2007; and Glewwe et al, 2007), and provider incentives (Glewwe at al, 2008; and Muralidharan and Sundaramanan, 2007), remedial education (Banerjee et al, 2007; Duflo et al, 2007; He et al, 2007), citizens' report cards, the hiring of contract teachers, or increased oversight of local school committees (Bjorkman and Svensson, 2007; and Duflo, Dupas and Kremer, 2007), school choice programs (Angrist et al, 2002, 2006; Bettinger et al, 2007).
It looks like the RE literature has offered a lot of particular aid interventions that “work” in education. Yet there is an air of randomness about which interventions work and which don't, since the intuition distinguishing the two is not compelling. One has the worry stated earlier that such laundry lists of results tend to select out significant coefficients without enough information about how many different outcomes were tested, what results were based on ex post slices of the sample, and how many results depended on inclusion of covariates. Also the worry about how RE findings are very sensitive to context remains relevant.

17. 50: Other RE studies seem more convincing and more robust. The famous Progresa program in Mexico to give cash grants to poor families in return for them keeping their children in school (subsequently known as conditional cash transfers (CCT)) has led to several influential studies using a randomized design. Schultz (2004) found that schooling among the beneficiaries did increase significantly, estimating the long run effect as 0.66 additional years of schooling on top of a baseline of 6.8 years of schooling. Behrman et al. 2005 came up with a similar estimate of 0.7 additional years of schooling using different methods (including effects on dropout, re-entry, and grade repetition rates). There are a few pilots of CCTs underway in Africa that are being evaluated with a randomized design, but none of the evaluations are available yet as of September
2008.
These studies seem more persuasive because they align well with theory – a sufficiently
large incentive to keep kids in school, created by PROGRESA, trumps the incentive for families to use children as workers to earn income.
Another famous RE finding on education offers some claims to robustness. The Kremer
and Miguel 2004 study on treatment of children for worms in Kenya found that it reduced school absenteeism by one-quarter (although it did not improve test scores). An interesting historical confirmation of this result is Bleakley 2007, who discusses the Rockefeller Foundation campaign against hookworm in the American South in the early 20th century. Bleakely also found strong effects on school attendance from decreasing worm infection. Bobonis, Miguel, and Puri-Sharma 2006 found that treatment of children with iron supplements, Vitamin A, and deworming for anemia reduced student absenteeism in pre-school by one-fifth in a district in India. This finding is thus an example of one that was successfully replicated in different settings.

18. 72: In contrast, RE studies have shed light on some of the marginal steps that could pay off in agriculture. One problem that is being studied is the chronically low use of fertilizer by African farmers, compared to other regions. Duflo, Kremer and Robinson 2008 study the hypothesis that the return to fertilizer on real world maize farms in Kenya is lower than the high returns on pilot farms, using REs of actual farms at different dosages of fertilizer. They found high returns also on real farms, although it required a kind of Goldilocks conclusion – too little or too much fertilizer makes the return unfavorable, but using just the right amount yields a large positive return. The official Kenyan government recommendation is NOT at the right amount and would yield poor returns. These results could suggest the problem with fertilizer under-utilization is due to missing technical knowledge on how much fertilizer to apply, but this is puzzling given the high private returns to acquiring such knowledge. In contrast, Conley and Udry 2007 document farmers learning how much fertilizer to apply from their successful neighbors in a new technology, pineapple growing, in Ghana (using spatial econometric techniques on a unique dataset of social connections among farmers). An earlier study by Duflo, Kremer, and Robinson 2007 tests for behavioral explanations of low fertilizer use. They find what seems to be a savings commitment problem – farmers do not set aside money for fertilizer for the next season when they are flush with funds from the harvest in the current season. Selling a voucher earmarked for fertilizer purchases to the farmers right after the harvest seems to correct the problem.

19. 72: Other micro studies (not REs) also show some potential in tackling some of the chronic problems of agriculture one at a time. A common concern about aid and agriculture is that food aid – giving food in kind for free – could harm local agriculture by driving down producer prices. If the poor are disproportionately small farmers, the longstanding fear in aid policy circles was that this kind of aid would perversely hurt the poor. Levinsohn and McMillan 2007 address this concern by analyzing a dataset with tens of thousands of household in Ethiopia, one of the chronic recipients of food aid. They find that the poor are disproportionately likely to be net buyers of wheat, and buyers outnumber sellers at all levels of income. Hence, the food aid fears are misplaced at least in Ethiopia – food aid is helping more households than it is hurting, and even more so among the poorest households. However, they note that the same objectives could be attained and local farmers also benefited if food were purchased locally and then distributed to the poor. This has been a frequent recommendation of aid analysts and there has been some movement in that direction, but there is still considerable pressure to source food purchases from the donor economies because of rich country agricultural lobbies.

20. 87-88: The micro randomization literature has also addressed corruption and possibly points the way to more successful marginal aid interventions. One influential study is Olken 2007, who found that official audits reduced corruption in Indonesian village road projects. When the villagers were told in advance that they would be subject to an audit by the central government audit agency (increasing the probability of an audit from 4 to 100 percent), an estimate of “missing expenditures” decreased from 28 percent of expenditures to 19 percent. The effect was statistically significant, but the magnitude is modest. Olken suggests the probability of punishment did not increase anywhere near as dramatically as the probability of an audit. Olken also tested whether grassroots monitoring (such as distributing anonymous complaint sheets to villagers) decreased corruption, and found no effect. Other micro empirical studies (not using randomization) also shed light on the effectiveness of some kind of transparency or auditing. In a famous paper, Reinikka and Svensson 2004 found from a tracking survey that only 13 percent of central government transfers to local primary schools in Uganda arrived at their destination. This research was itself a form of transparency, as the release of the study by itself prompted improvement in the transfer ratio. The Ugandan central government took the bold measure of publishing the intended transfers by school in the local newspapers where they could be monitored by parents and local officials. Reinikka and Svensson 2005 showed that the newspaper campaign successfully increased the proportion of transfers that arrived at the schools, one of the factors behind an increase in this proportion to 80 percent. The newspaper campaign also showed up in the increased enrollment and test scores in these schools. Also possibly supportive of the monitoring and transparency approach to reducing corruption is the finding by Besley, Pande, and Rao 2005 that higher education among the voter population is associated with less corruption, using a natural experiment of elections to village councils in South India.

21. 92: Olken 2008 uses the randomization methodology to assess the impact of introducing more democracy into the aid process itself. A random sample of Indonesian villages was given the right to democratically choose which aid projects would be implemented. Aid democracy did not have much effect on which projects were actually chosen, but it did dramatically improve villagers' satisfaction with the projects and their willingness to contribute. Again, a micro study points to a small step whereby donors could directly introduce a modest level of democracy and it would have some effect on outcomes. It is pure guesswork to assess how such steps affect the larger agenda of making a country more democratic, but at least it makes a village a little more democratic!

22. 39: Swaroop and Devarajan 2000 and Feyzioglu, Swaroop, and Zhu 1998 both find significant but less than 100 percent aid fungibility across sectors. Even with partial
fungibility, unfortunately, the rate of return to an aid-financed project is not the same as the general equilibrium rate of return to aid spending.

23. 42: Kremer and Miguel 2007 suggest the problem is the donors' obsession with
“sustainability,” in which they envision the recipient government or local communities providing the financing of recurrent costs (operations and maintenance) after donors finance the capital costs of infrastructure, so that the project will be “sustained” once donor financing ends. This hope has turned out to be an illusion, as the failure to cover recurrent costs has been nearly universal. Kremer and Miguel suggest donors should be willing to permanently bear the recurrent costs of their projects if they really want those projects to be effective.

24. 43: The World Bank did 70 civil service reforms in Africa during 1987-97, and over a quarter of World Bank lending to Africa is currently devoted to “capacity-building.” Yet political scientists specializing in analyzing African states see little sign of effect of these Herculean efforts at making civil servants perform better, even seeing some signs of decline (Moss et al. 2008).

25. 75-76: However, the poor growth outcomes in Africa in the 1980s and 1990s caused much blame to be heaped on structural adjustment. The attempt to attain East Asia's growth (or even respectable per capita growth) did not succeed, not only in Africa but also in Latin America and the Middle East. With this failure, the “imitate the stars” approach fell into disrepute with many academic observers, such as Dixit (2007): At any time, some country is doing well, and academic as well as practical observers are tempted to generalize from its choices and recommend the same to all countries. After a decade or two, this country ceases to do so well, some other country using some other policies starts to do well, and becomes the new star that all countries are supposed to follow.
Hence, the backlash against structural adjustment coincided with a loss of confidence in the academic literature that researchers could identify policy actions that would raise growth. The early hope that growth regressions would identify growth-promoting policies ran afoul of concerns about data mining (see discussion above). So many of the world's leading macroeconomists concluded in a conference called the Barcelona Development Agenda (2004): “there is no single set of policies that can be guaranteed to ignite sustained growth.”40 The World Bank (2005) itself accepted this agnosticism: “different policies can yield the same result, and the same policy can yield different results.” The World Bank followed this up by sponsoring a Growth Commission, whose final report appeared in May 2008, and contained a similarly agnostic conclusion: “It is hard to know how the economy will respond to a policy, and the right answer in the present moment may not apply in the future” (Commission on Growth and Development 2008, p. 29). This is not to say that these cited reports embraced development nihilism – they all contain plenty of useful insights about development – but the confidence of the SAL era that economists can say what policies will reliably pay off in growth rates is gone.

26. 77-78: there is a lot of variation within Africa as to who received SALs, and this variation was unrelated to the improvement in policies. The variation consists of whether countries received SALs at all, and for those who did, how many they received. The biggest surprise in the way that SALs evolved was that many countries received an awful lot of them, topped by the 26 in Cote d'Ivoire over 1980-1999. There are two different ways to interpret the frequent repetition of adjustment lending to the same country: (1) policy dysfunction requires a gradual, multi-stage treatment, so each additional SAL was taking a salutary step in the right direction, or (2) previous SALs were ineffective at changing policies (or raising growth, to be discussed below), so new SALs tried again, which also helped repay the previous ineffective SALs. In the view according to (2), over time a serious problem of moral hazard in adjustment lending developed. The evidence seems inconsistent with (1), since Easterly (2005) found no evidence of policy improvements from one SAL to the next within countries. Collier et al. 1997 also pointed out the lack of evidence that SAL conditions were kept. World Bank 1998 (p. 51) pointed out that the same agricultural policy reform in Kenya was the subject of a condition in five different SALs, violated each time. The IMF's own evaluation office (Independent Evaluation Office (IEO) 2002) harshly criticized the IMF's repeat lending as counterproductive, apparently finding no evidence that it was part of a salutary multi-step process. The IEO 2007 confirmed the general failure of IMF SALs to change economic policies over 1995-2004 (a period including the successor instrument to SALs after 1999). IEO 2007 found that about half of structural conditions were not kept. Yet Easterly 2005 found an exogenous trend in improvement in most of the policies described above, unrelated to the number of SALs received. It could be that the intellectual influence of the IMF and World Bank was important in convincing countries to improve their policies, but if so, this was not mediated through SALs. Why were SALs ineffective at inducing policy change? Svensson (2003) argues that there was a problem of time inconsistency in conditionality. Since aid recipients knew that each country department in the World Bank, for example, was under pressure ex-post to fully disburse its budget, the threat of withholding disbursements if conditions were unmet was not credible ex-ante. The donors also seemed to lack appreciation for internal political incentives to sometimes keep pursuing policies that benefited local elites. The weak incentives to change from donor conditions paled by comparison.Other initiatives: education, health, agriculture

27. 80-81: 41 HIPCs in the end). Of 18 IDA countries that received above average number of SALs, 17 became HIPCs. Today, debt ratios are rising once again in these same countries – UN (2008, p. x) notes that 21 HIPC countries “are considered to be at moderate-to-high risk of falling back into debt distress.” The World Bank and IMF do not seem to have learned sufficient lessons about the dangers of lending to the poorest countries, with the history of debt forgiveness now creating an obvious moral hazard problem. 81World Bank and IMF adjustment loans using as instruments strategic variables like a dummy for former French colonies, US military assistance, and log of population size; this regression finds a positive but insignificant effect of SALs on growth. Of course, the same concerns about identification assumptions (do SALs to Francophone countries have the same effects as others, for example?), unclear specifications, and data mining could be leveled against this literature as much as the aid and growth literature. This literature differs from the aid literature, however, in that there are very few academic claims of positive effects of SALs on growth (although there are such claims in non-academic publications of the IMF and World Bank itself). The repetition of the loans to the same country alleviates, but does not eliminate, the selection bias problem. If the same patient is re-admitted on a daily basis to the emergency room and fails to improve, one is inclined to think the emergency room is ineffective or the wrong form of treatment. SALs were supposed to be “emergency” loans that enabled countries to correct problems over the life of the original loan – their repetition was not envisioned in their design. It could be designers of SALs did not realize that they needed to be a multi-stage process in which different loans would address different problems. However, as we have seen, the macroeconomic policies did not improve from one loan to the next. Another indirect piece of evidence on the outcomes from SALs is that the loans ultimately were forgiven (the Heavily Indebted Poor Countries –HIPC-- initiative in 1996 partially forgave the SAL debt, then the Multilateral Debt Relief Initiative created in 2005 forgave virtually all of it, which was intended to end over 20 years of incremental debt relief). Since SALs to Africa were heavily concessional (zero interest and 40 year maturity), the payoffs to the loans were not good enough to avoid a crippling debt crisis even with debt that was mostly a grant.43 Only countries borrowing at the World Bank's International Development Association (IDA) concessional rates were eligible to become HIPCs receiving debt forgiveness (there were 43 Another possibility was unwillingness to repay, as opposed to inability to repay. However, the World Bank and IMF would have been less likely to forgive the loans if it had been too obviously the first. The empirical literature on IMF/World Bank structural adjustment lending and growth outcomes faces many of the same issues as the aid and growth literature. There is an obvious selection bias in whom the World Bank/IMF treats with adjustment lending, just as there is a selection bias in whom an emergency room treats. Some of the complaints by NGOs about SALs are based on correlations between SALs and outcomes that are the equivalent of the negative correlation between admission to an emergency room and a person's health, with the implication that the emergency room is bad for your health. Przeworski and Vreeland 2000 address this problem by doing a selection equation for entering an IMF program involving when a country is under pressure from low foreign exchange reserves, high budget deficits, and high debt service. Variables affecting the IMF's side of the loan decision are also significant, such as the balance of payments deficit, whether the government is a dictatorship (favorable for getting the IMF to give you a loan), and the number of loans the IMF is making to other countries. Controlling for selection bias, they find that an IMF program lowers growth by 1.5 percentage points. Barro and Lee 2005 find that IMF “loans tend to be larger and more frequent when a country has a bigger quota and more professional staff at the IMF and when a country is more connected politically and economically to the United States and other major shareholding countries of the IMF.” Using these variables as instruments, Barro and Lee find that IMF loans have a negative effect on growth. Easterly 2005 does an IV growth regression for the number of

28. 84: The aid community had two levers available to try to induce decreases in corruption. It could withhold aid from corrupt governments, and it could use its technical advice to control corruption. However, using aid money as leverage was subject to the same time inconsistency problem as conditions on SALs, and did not turn out to be conspicuously effective. In fact, there is no evidence of increasing responsiveness of aid allocation to corruption (or to democracy), as demonstrated earlier in Alesina and Weder (2002) and reaffirmed and updated in Easterly (2007).

29. 85: There is no trend in corruption in Africa relative to the rest of the world over 1996-2006. This conclusion is derived from the measure of Kaufmann, Kraay, and Mastruzzi 2007 (KKM), who do a sophisticated averaging over all available corruption indicators, correcting for selection bias and other problems. The KKM measure is relative each year, standardized as a Normal (0,1). African countries on average are a little over 0.6 standard deviations worse than the world average on corruption, a measure that showed little change over 1996-2006 (especially considering the wide confidence intervals). The trend on the relative measure seem most relevant to the aid-corruption question when comparing a region above average in aid-intensity to the rest of the world in making progress against corruption.

30. 86-87: What about more formal empirics on the relationship between aid and corruption? One of the most well-known regression studies finds that aid worsens corruption for ethnically diverse countries – which includes most African countries (Svensson 2000). This finding resonates with Banerjee and Pande's (2007) story that voters that are more polarized along ethnic lines are more likely to elect corrupt candidates, because they care more about electing a candidate from their ethnic group than electing an honest candidate. Banerjee and Pande showed in the state of Uttar Pradesh in India that a rise in ethnic politics went together with increasingly corrupt politicians (as measured by voter perceptions – and by the criminal record of the candidate!) Other studies find that ethnic diversity (which is the highest in the world in Africa according to standard measures) tends to increase the demand for redistribution towards one's own ethnic group using channels like public employment (Alesina, Baqir and Easterly 2000) and public transfers (Besley, 87Pande, and Rao 2004). A related literature shows that less resources are allocated to public goods in ethnically diverse environments (see Alesina, Baqir, and Easterly 1999 and Luttmer 2001 for the US, Banerjee, Iyer, and Somanathan 2005 for India, and Miguel and Gugerty 2005 for Kenya). Introducing more additional resources such as aid into such a political equilibrium is likely to raise misappropriation of public funds. Knack 2001 found that aid unconditionally worsened “governance,” an average of ratings of corruption, bureaucratic quality, and rule of law. Svensson and Knack instrumented for aid with the usual population size, initial need, and strategic variables. However, Tavares 2003 finds that aid decreases corruption, using a similar IV specification for aid. It is rather frustrating that later authors did not themselves try to relate disparate results to previous studies. Of course, these aggregate studies are subject to the same critiques as with the aid and growth literature, with unclear specification of other control variables and identifying assumptions that are always somewhat problematic for aggregate outcomes in which most factors are endogenous.

31. 89-90:Africa had more of a democratic transition than other developing countries, at about the same time (but not quite to the same degree) as ex-Communist countries moved away from autocracy. However, Africa's democratic transition preceded the heyday of democracy promotion efforts by donors, so it would be hard to attribute the former to the latter (although general aid may still have played some role). The KKM governance indicator on African democracy relative to the rest of the world is available only for the decade 1996-2006 (which would correspond more closely to the timing of democracy promotion efforts). There has been little sign of Africa converging to the rest of the world on the KKM democracy measure.

32. 91-92: In Africa, the uncomfortable reality was that many voluntary groups formed along ethnic lines, which politicians often exploited at election time in a way that increased ethnic animosity. Even aside from the ethnic issue, other voluntary groups in Africa were economic self-help organizations that were relatively apolitical. Donors attempts to fund Western-style NGOs that promoted political participation and issue lobbying often created artificial NGOs with few roots in the community, which would immediately collapse if donor support was withdrawn (Ottaway 2000).

33. 92-93: A small cross-country regression literature has analyzed the effect of aid on democracy. Knack 2004 finds no association between aid and the change in democracy from 1975 to 2000, including when he instruments for aid. Djankov, Montalvo and Reynal-Querol (2008, 2006) found a causal negative relationship from aid to the change in democracy, using the usual suspects as instruments for aid. They labeled this the “aid curse,” in which aid is as bad for democracy as oil is in the well-known “oil curse.” The intuition is similar: more money available to those who control the state will make them less likely to permit any democratic threat to their stay in power. Moss et al. 2008 provide further intuition for such results when they argue that states beholden to donors for most of their revenues have less incentive to be accountable to their own citizens compared to states dependent on domestic tax revenue. Their argument is most relevant for Africa, since the median African country got aid equal to 37 percent of government expenditures over 1990-2006 (compared to 4 percent for non-African aid recipients). They also argue that goods provided by donors such as four-wheel drive vehicles or “sitting fees” for attending donor seminars (which can exceed monthly salaries of civil servants) can become objects of political patronage, reinforcing the “patrimonial state” and further undermining the prospects for democracy.

34. 94: Despite decades of attempts to register land titles, during both the colonial and independence eras, today only about 1 percent of land in Africa is registered under the formal system (Blair Commission for Africa 2005, p. 231). In Africa, there has been a long historical evolution of customary rights to land. As Pande and Udry 2005 say in a study of the Democratic Republic of the Congo, Cote d'Ivoire, Gambia, and Ghana, such rights can be complex: what matters for rural land rights is the country's community-based mechanisms as exemplified by customary law. The use of almost all land in these four countries is governed by customary tenure arrangements, not formal sector rules....The same piece of land can be subject to multiple claims which relate to the ways in which it is used by separate groups and individuals at different levels. For example, one individual may have the right to cultivate annual crops on a plot, while another retains rights to the tree crops that exist on the same land. An elder might have the right to allocate a plot to a family member for temporary use, but not the right to rent the plot to an outsider on a commercial basis.
Outside donors paid little heed to the pre-existing local arrangements. Under these
circumstances, issuing a land title to yet another party can increase rather than decrease
uncertainty about who has what rights to the land.

35. 94: Indeed, a number of empirical studies show little effect of outsider-directed formal land titles on the incentive to invest in Africa. Jacoby and Minten 2007 found no effect of land titles on plot-specific investment in rice fields in Madagascar. Deininger and Jin 2006 have recently summarized the literature on land titles in Africa as showing little or no effect of titles on investment or access to credit, although they found evidence that a more general measure of “tenure security” in Ethiopia (not dependent on titles, which did not exist) fostered land investments.

36. 95: Why was the payoff to formal land titling in Africa so hard to find? Migot-Adholla et al. 1991 long ago presented evidence that indigenous property systems in Africa, far from being static, have themselves spontaneously evolved towards more individualized land rights in response to increased population pressure. They argued therefore that the indigenous systems do not constrain investments in increased land productivity. Platteau 1996 also argued that there is little evidence of any benefit of formal land rights compared to indigenous systems. Cotula 2007 and Boudreaux and Aligica 2007 provide more recent statements of this same view, albeit with some variations and cautionary notes that indigenous evolution of land rights is not a panacea for optimal outcomes. Udry and Pande 2007 have a more recent empirical result reinforcing the picture of population pressure-induced evolution of customary land systems; they found that commercial land transactions were more likely in matrilineages in Ghana that had higher population pressure.

37. 95-96: RE studies from other contexts give a somewhat more positive picture for the effect of formal titling. Erica Field has performed a number of studies based on a natural experiment of the semi-randomized timing of a government program giving formal titles to urban squatters in Peru. Field 2005 found that these titles increased title-owners investment in their urban shantytown dwellings (a 68 percent increase in the rate of home renovation in the four years after receiving a title). Field 2007 also found that title holders were able to reallocate work away from that performed at home and to increase total work hours, plausibly because greater tenure security reduced the need to have someone always at home to protect the property. Field and Torero 2006 have a more ambiguous result on the De Soto prediction that formal titles would unlock access to credit. They found newly entitled households got more credit from government banks, but not from private banks. If the household did manage to get a loan, the interest rate for households with titles was 9 percentage points lower. Dower and Potamites 2007 find that an Indonesian titling program had a positive effect on credit access, but not necessarily through collateral for the loans (which is how De Soto thought titles would increase credit access). Their data on rural Indonesian households showed when a title was used as collateral, compared to having a title but not using it as collateral on a loan. The latter had an effect on access to credit, but the first did not. Their interpretation was that title was more useful as a signal of creditworthiness to banks, not through collateral. Sixty percent of titled households with formal loans do not use the title as collateral. Further evidence against the “title as collateral” story was that the banks also accepted non-formal claims to land as collateral (and they accounted for 58 percent of the land collateral cases). Udry and Pande 2007's world wide survey of the land titling literature concludes “Land titling and registration typically increase agricultural productivity and farm investment,” but their list of studies in Africa mirror the same non-results described above.

38. Civil War: 96-104 long, dense discussion tearing apart Collier

39. 47: There is an obvious long run global trend towards increasing enrollments (Kenny 2008c); developing countries since 1960 have raised enrollments faster than today's rich countries did in their history (Clemens 2004). The brute stylized fact is that donors intended to increase education with aid, they spent money building schools, and
enrollment did increase – such evidence is suggestive even if far from definitive.

40. 51: At best, the multiplication of “interventions that work” shifts priors that “donor efforts can pay off in education.” If so, then together with the stylized fact that enrollments rose in Africa at the same time as there was extensive donor involvement in African education, perhaps does shift priors that “aid works in education.” Otherwise, one is left with the feeling that aid could improve education, but the literature is not always that clear on when, why, or how.

41. 51-52: Despite Africa's success on raising primary enrollment, there has been disappointment that growth in education has not paid off in higher economic growth, as stressed in Pritchett (2001). Education has its own micro-macro paradox, as Mincer regressions usually show a positive impact of an individual's educational attainment on their wages, but results from growth regressions and growth accounting suggest little or no aggregate payoff to society-wide education. Africa plays a large role in Pritchett's results, as it contributes several dozen observations with low economic growth and rapid percent growth in schooling attainment. Krueger and Lindahl 2001 contradicted Pritchett with much more positive results showing a positive association between the absolute change in years of schooling and economic growth -- this put the African low growth observations more in the middle of the sample compared to being at the top of the sample on percent growth in years of schooling, since initial schooling in Africa was so low. However, Pritchett (2006) lets micro and macro data arbitrate the functional form and
finds that the best fit is closer to percent change than to absolute change. The poor outcome of educational improvements in Africa is consistent with the stylized fact that there is little job creation in the African formal private sector, which would normally be the employer of skilled labor (Pritchett 2006). Poor institutions could explain such an outcome, and poor institutions could also divert skilled labor into rent-seeking rather than productive activities. Even those who argue strongly for a positive effect of education on growth concede that poor institutions and policies, as in Africa, prevent education from paying off (Hanushek and Wößmann 2008).

42. 52-53: Another well-known and long-standing finding in the growth regression literature is between initial schooling (usually the primary enrollment rate) and subsequent growth rate, controlling for per capita income (Barro and Sala-i-Martin 2003). Doppelhoffer et al. 2004 find that initial primary enrollment is the single best performing variables in a Bayesian exercise to decide what variables belong in the growth regression
. Hanushek and Kim 2000 and Hanushek and Wößmann 2008 stress initial quality (as measured by test scores) of education, and get stronger results with their test score variable than those for initial enrollment. Low primary enrollment quantity could have also been proxying for low schooling quality, since a dysfunctional education bureaucracy would plausibly produce both low quantity and low quality. However, Bils and Klenow 2000 had already raised some doubts about whether the relationship between education level and growth was causal, noting that the coefficient magnitude could be explained entirely by individuals' investing in education in anticipation of high growth
(which obviously raises future returns to skills). Easterly 2001 and Pritchett 2006 pointed out that a causal relationship between initial schooling and growth would predict accelerating growth with rising education in all developing regions, whereas the sample mean growth rates actually fell instead from the 60s through the 90s. Even if we accept as credible worldwide evidence on some growth payoff to initial level of schooling, there is considerable disappointment for Africa that this payoff has not materialized despite successful efforts at expanding schooling, which again could reflect poor quality of schooling and/or low demand for skills related to poor institutions.
We are left with little reason from the aggregate empirical literature to believe that rising
education in Africa has paid off in higher per capita income or growth. This disappointment weakened the arguments of advocates of “marginal” project interventions and strengthened the case for “transformational” systemic changes, as we will see in the next section.

43. 53: Health is an even more clear success story than education in Africa, as child mortality has improved dramatically over time (Figure 8). There are well known and striking donor success stories, like the elimination of smallpox, the near-eradication of river blindness and Guinea worm, the spread of oral rehydration therapy for treating infant diarrheal diseases, DDT campaigns against malarial mosquitoes (although later halted for environmental reasons), and the success of WHO vaccination programs against measles and other childhood diseases. The aid campaign against diseases in Africa (known as vertical health programs, see discussion below) is likely the single biggest success story in the history of aid to Africa (see Levine 2007).

44. 54: There is another sense in which the West had a major effect on health in Africa. The
major technological breakthroughs in health – e.g. antibiotics, vaccines, the germ theory of disease, the identification of mosquito transmission of malaria, later the discovery of the AIDS virus – originated in the science of the West (see discussion in Cutler, Deaton, and Lleras-Muney 2006). The health improvements in Africa would have been impossible without Western science; this is one important way in which Western outsiders did indeed “save Africa,” at least in one specific area. Acemoglu and Johnson 2007 show empirically the strong effect of the international epidemiological transition on changes in life expectancy after 1940.

45. 55-56: Finally, randomized evaluations of also found positive impacts of a number of health interventions adopted by aid agencies or NGOs. First, many of the education interventions discussed above also had a health component. Gertler 2004 checked whether the PROGRESA cash-for-schooling program also had a major health impact, since the cash rewards were also conditional on families receiving micronutrients and protein supplements, and bringing their children to clinics for regular health and nutritional checkups. For children covered by the program compared to the randomized control group, Gertler found significant effects of a 22-25 percent decrease in probability of illness in the 4 weeks preceding the checkup, an impact on child height of 1 centimeter (although, puzzingly, not a significant decrease in probability of stunting), and a 25 percent decrease in probability of anemia. The nutritional success is notable when we remember that knowledge of the large payoff to cheap nutritional supplements has been
around for decades (like Vitamin A in Table 5), and yet these still remain underutilized. These findings comprised another large part of the “Progresa Success Story” discussed above. The Bobonis et al. 2008 study on anemia and school participation also found that iron supplements and deworming drugs were effective in increasing children's weight-for-height and weight-for-age scores. This might be thought to be obvious, except the impacts on direct measures of anemia and worm infection were surprisingly insignificant. (Again, one worries about a pattern of some outcomes being significant – hence the intervention “works” - but other equally plausible ones are not. This makes it more difficult to interpret the significance level of a conclusion that an intervention “works.”)
In contrast, the well-known Kremer and Miguel paper showed a strong effect of deworming on worm infection rates in a district in Kenya, which reflected not only direct effects on children receiving the drugs but also surprisingly strong externalities to others in the same school or nearby schools. Bleakley 2007 also noted the strong and immediate effects of the Rockefeller deworming campaign in the American South.
Another area where REs point to success is in preventing or treating infant diarrhea
(Zwane and Kremer 2007). Breastfeeding, immunization against diarrheal diseases, micronutrient supplementation and oral rehydration therapy (ORT) have all been found to work in randomized trails in the fight against diarrhea. Unlike the education interventions, we know from case studies that these interventions were pursued by donors. Case studies suggest ORT is another health aid success story, accounting for a substantial drop in diarrheal mortality since 1980. REs seem to be more persuasive in health, but for reasons that also make them less necessary. The link between
medicines and health is often so obvious that it doesn't require an RE to verify it. Still to be as generous as possible, taken together, the various kinds of evidence support some positive effect of aid on health.

46. 66-67: African agricultural aid is also unusual in that virtually all those involved agree that it has been a failure, amidst much recrimination and finger-pointing. The stylized facts on food production per capita certainly influence this pessimism, with a decline in Africa contrasting with the general Asian rise (Figure 10).

47. 67-68: World Bank (1997) called for movement “From Vision to Action.” World Bank (2003) is the report of another task force called “Reaching the Rural Poor,” which noted the agricultural development portfolio has not yet met the 80% satisfactory development outcome rating at completion, as targeted by “From Vision to Action.” The quality of the poverty focus, and the sustainability and quality of the institutional development still leave much to be desired.
Reaching the Rural Poor will address these concerns.
The latest report, an internal evaluation of all World Bank work in agriculture over the
period 1991-2006 (World Bank Independent Evaluation Group 2008), was again scathing about failure. The 2008 World Bank WDR (p. 15) in turn noted the stagnant cereal yields in Africa in contrast to rising yields in all other regions.
WDR 2008 noted the existence of “'agroskepticism' of many donors” which “may well be related to their experience with past unsuccessful interventions in agriculture.” Similarly Eicher and Kane (2004) noted “The failure of past initiatives in agriculture led to a reduced confidence among donors in agriculture in the 1980s ...and many donors have since turned to other sectors.”
The UN system has followed a similar progression, with a World Food Summit in 1996
another installment in a long line of efforts to make progress on hunger in Africa through
agricultural development. The FAO (2006) passed judgment on that effort: “Ten years later, we are confronted with the sad reality that virtually no progress has been made towards that objective.”
As far as the “green revolution” specifically, Eicher (1999) had already noted that:
Much energy has also been wasted in trying to replicate Asia's Green Revolution model
in Africa before the completion of pilot studies. Over the past decade, many instant experts on Africa have talked glibly about the ease of replicating Asia's Green Revolution model in Africa. Many of these experts have overlooked Africa's early stage of scientific development, falsely assuming that Africa had the requisite infrastructure, irrigated land, trained scientists, technology, and national and local institutions to replicate the Asian model.
The Wapenhans report (World Bank 1992) confirmed this picture, with only 40 percent
of World Bank agriculture projects in Africa judged as successful (compared to 59 percent for all projects in Africa, and 72 percent for African education projects).

48. 67 Note 36: One exception to the general gloom on African agriculture was the success of commercial maize production in southern Africa.

49. 40: While educational enrollments have expanded rapidly, the quality of education is hampered by missing inputs like textbooks and other school materials, weak
incentives for teachers to show up or teach effectively, corruption in education bureaucracies, appointment of unqualified teachers for patronage reasons, and disruption of schooling by political events (Filmer and Pritchett 1999). Donors have long recognized the quality problems in education (for example, World Bank WDR 1980), but these problems are remarkably persistent (World Bank WDR 2007 again stressed quality problems in education). In health, corruption in the health system (studies in Guinea, Cameroon, Uganda, and Tanzania estimated that 30 to 70 percent of government drugs disappeared before reaching the patients), absenteeism of health workers, and sheer bureaucratic inefficiency are chronic problems. Some widely-cited regressions find no impact of health spending on health outcomes (Filmer, Hammer, and Pritchett 2000, Pritchett and Woolcock 2004).

50. 41: Chaudhury, Hammer, Kremer, Muralidharan, and Rogers 2006 surveyed studies
in Bangladesh, Ecuador, India, Indonesia, Peru, and Uganda where unannounced visits to schools and health clinics found teachers absent 19 percent of the time on average and health workers absent 35 percent of the time. Even this was an understatement as some who were present were not working. We can understand this as reflecting weak sanctions for absence: in a sample of 3000 Indian schools, there was only one report of a teacher fired for repeated absences. The problem of teacher and health absence is worse in poorer countries or states within countries (e.g. the one African country, Uganda, has a worse problem than richer countries in the study). There is some evidence of response to incentives. Teachers in an NGO program of non-formal schools in India that were required to take a date-stamped picture of themselves with students everyday, with a pay bonus for each additional day of attendance, had a much lower absence rate – 22
percent compared to 42 percent in the control group (Duflo and Hanna 2005).

51. 58-59: The history of health aid is a cycling between these two alternatives. After the early “vertical” health successes described above ran into diminishing returns, there was a switch to the “horizontal” approach. By 1980, the World Bank had shifted towards recommending an “integrated approach” in health (i.e. horizontal), which continued for the next two decades. The 1993 World Bank WDR on health, for example, stressed the health system problems described above as a critical bottleneck in improving health.
By the new millennium, however, the prominent health crisis of AIDS in Africa induced a shift back towards vertical, disease-specific programs, such as the creation of the Global Fund to fight AIDS, TB, and Malaria (GFATM) in 2002, the US President's Emergency Plan for AIDS Relief (PEPFAR) in 2003, the President's Malaria Initiative (PMI) in 2005, and the Gates Foundation's well-publicized efforts on these same diseases, which implied large increases in health aid but mainly in these vertical programs. There have been some successes from these programs, such as the life-saving treatment of more than 1 million HIV positive Africans (Sachs 2008). However, critics have complained that the concentration of foreign aid on AIDS, in particular, has crowded out more cost-effective approaches to more common diseases, not least because the AIDS initiatives may have overwhelmed the still dysfunctional public health systems. 59 For example, a group of health experts wrote in the prestigious medical journal the Lancet in July 2003 about how 5.5 million child deaths could have been prevented in 2003, lamenting that “child survival has lost its focus.” They blamed in part the “levels of attention and effort directed at preventing the small proportion of child deaths due to AIDS with a new, complex, and expensive intervention.” (Gareth Jones et al. 2003) England (2008) points out that while AIDS causes 3.7 percent of mortality, it gets 25 percent of international healthcare aid. Even within AIDS programs, prevention is neglected relative to treatment, even though the former has far better cost/benefit ratios (Canning 2006). Moreover, AIDS funding is increasing even further – President George W. Bush signed a bill in July 2008 giving an extension of his original 2003 five-year $15 billion PEPFAR program for another five years at $30 billion (still heavily skewed towards treatment). AIDS is a good example of how the vertical approach is vulnerable to capture by lobbies for particular diseases that are “fashionable” causes in rich countries but don't necessarily match the aid recipient's priorities. The World Bank (2007) responded by again fervently advocating the horizontal approach. The large new vertical programs would not work unless there was an “urgent effort ... made to strengthen health systems” (p. 15). But the G8 Summit in July 2008 in its discussion of health in Africa stubbornly stuck with vertical: “G8 members are determined to honor in full their specific commitments to fight infectious diseases, namely malaria, tuberculosis, polio and working towards the goal of universal access to HIV/AIDS prevention, treatment and care by 2010.”

52. 61: This is the kind of debate where the RE literature claims to deliver a clearer message to policymakers and does not allow them to cherry-pick studies for support for their favorite policies or interventions. It is also supposed to deliver a clearer verdict than the sometimes inconclusive debates between academics about empirical findings. Unfortunately, things did not work out so cleanly, as pointed out by Rodrik (2008) and others. For example, advocates of providing development goods for free often quote the Cohen and Dupas (2007) study that finds that going from free provision of bed nets to charging 75 cents per net (still heavily subsidized) reduced 61uptake by 75 percent. Supporters of charging for development goods have cited the Ashraf et al 2007 finding that charging for water purification tablets was successful in screening out those less likely to use them. The general conclusion that demand for health inputs is very sensitive to price in Africa seems on firmer ground. Kremer and Miguel 2007 found that modest user fees for deworming drugs reduced take-up rates by 80 percent in Kenya. Kremer and Holla 2008 argue that the pattern of take-up being very sensitive to price is consistent with a number of RCT studies of interventions: not only deworming and bed nets, but also learning the results of HIV tests. They point out that the evidence from Progresa of significant health, nutrition, and education responses to relatively small subsidies is also consistent with the same high price elasticity hypothesis. Even the Ashraf et al. 2007 study did show a high price elasticity for water purification tablets (even if price is successful at screening out those less likely to use them, there remains the question of why there are so many who don't want to use water purification tablets).

53. 63-64: As already noted, Ashraf et al. 2007 noted that water purification tablets in Lusaka, Zambia were an inexpensive way of avoiding water-borne illness. Zwane and Kremer 2007 suggest behavioral changes such as hand-washing and disinfecting the household's own water were more effective in rural areas than formal infrastructure. Hand-washing and other hygiene behaviors may be necessary even if there is clean water infrastructure, to avoid re-contaminating the water, although evidence on this is ambiguous (Kremer and Zwane 2007). However, how to 64induce such behavior change is still unclear. Kremer et al. 2008 showed that investments in protecting naturally occurring springs from contamination led to dramatic improvements in water quality in rural Kenya (as measured by the fecal indicator E. Coli). However, the higher communal water quality at the springs did not seem to pay off at the household level, perhaps because of recontamination through household behavior, as there was no effect on diarrhea, or child height and weight. We have already seen that there was high price elasticity for water purification tablets, and Zwane and Kremer describe how behavior changes such as hand-washing and purifying water were surprisingly difficult to achieve in poor households. Another study by Kremer et al. 2008 in rural Kenya estimated household willingness to pay for clean water (estimated through transportation costs to protected springs that were clean compared to those that were not) as coming out only to $0.86-$1.72 per case of diarrhea avoided (surprisingly low when diarrhea from water-borne diseases is a life-threatening condition for infants.)

54. 66: the privatization of water services in Argentina was associated with a 5 to 7 percent drop in infant mortality in Argentina according to one recent study (Galiani, Gertler, and Schargrodsky (2005)).