Burnside and Dollar related growth rates in developing countries to foreign aid received, as figure 2 does for Africa. However, their new twist distinguished between aid recipients who had “good” policies (measured by things such as low budget deficits, low inflation, and free trade) and those with “bad” policies. Their hypothesis was that good policy increased the payoff to aid, so growth should be related to aid among countries with good policy. This was intuitively appealing, because it recognized that bad government could be the problem, as discussed in the previous section. If poor countries had good governments, then perhaps aid would increase growth after all.
Their sample consisted of six four-year time periods running from 1970–1973 to 1990–1993. In many of their tests, they found that when a country both got more foreign aid and had good policy, growth went up. They summarized: “We find that aid has a positive impact on growth in developing countries with good fiscal, monetary, and trade policies but has little effect in the presence of poor policies” (p. 847).
Their paper reinforced the hope that aid could accomplish great deeds, which fed a policy recommendation to increase foreign aid to a country only if that country’s policies were good. In early 2002, The Economist rebuked then U.S. Treasury secretary Paul O’Neill for his skepticism about foreign aid on the grounds that “there is now a strong body of evidence, led by the research of David Dollar, Craig Burnside[,]…economists at the World Bank, that aid does boost growth when countries have reasonable economic policies.” An article in the New Yorker in 2002 chimed in that “aid can be effective in any country where it is accompanied by sensible economic policies,” based on the Dollar and Burnside study.
President George W. Bush was apparently reading the American Economic Review as well. On March 14, 2002 (any coincidence in timing with the war on terror is purely intentional), he announced a five-billion-dollar increase in U.S. foreign assistance, about a 50 percent increase.9
The White House followed up on November 26, 2002, with the creation of the Millennium Challenge Corporation (MCC), whose job is to administer the five-billion-dollar increment in foreign aid. Arguing that aid works only with good policies, the administration announced sixteen indicators of country performance to guide the selection of countries to receive MCC aid—three of the indicators were versions of the Burnside and Dollar policy measures (most of the rest were measures of quality of institutions). On its Web site, the White House said that the new aid was motivated by the idea that “economic development assistance can be successful only if it is linked to sound policies in developing countries.10
In May 2004, the Millennium Challenge Corporation announced the selection of sixteen “good policy” countries eligible to apply for its aid grants from fiscal year 2004 funds.11 In March 2005, the MCC reached its first agreement with a “good policy” country, a Millennium Challenge Compact with Madagascar.
How much can we rely on the original study that sent this freight train down the tracks? A study I did with Ross Levine (Brown University), and David Roodman (Center for Global Development) used the exact same techniques and specifications as Burnside and Dollar, but added new data that had become available since Burnside and Dollar did their study. We also hunted for more data in their original sample period (1970–1993). We found more data even over their sample period by consulting the original sources rather than secondary sources. Using updated data, we did the same statistical exercise with four-year averages with the same control variables, including terms for aid/GDP, and their policy index (a weighted average of budget deficits/GDP, inflation, and an index of openness to trade). We found no evidence that aid raised growth among countries with good policies, indicating no support for the conclusion that “aid works in a good policy environment.” Our study was published as a comment on Burnside and Dollar in the American Economic Review.
The original researchers and other researchers may have tried many different statistical exercises, but the aid policy community is tempted to select the study that confirms its prior beliefs (known as “confirmation bias”)—even though other statistical exercises may have found no evidence for it. Applying new data to the old statistical exercise is a good test of whether the original result really holds and is not just confirmation bias. The statistical exercise with the new data is constrained by the old statistical exercise, so you are not searching among many different exercises for the one confirming prior beliefs. Even good first-round research can suffer from confirmation bias.12
The cycle is now starting all over again. After my co-authors and I found no evidence for the “aid works in a good policy environment” conclusion, a new study came out by Michael Clemens, Steven Radelet, and Rikhil Bhavnani (hereafter denoted CRB) of the Center for Global Development. I respect these authors a lot and think they were following high academic standards. Their new twist on the statistical exercise was to separate aid that could be expected to have an impact on growth in the short run from aid that had either a humanitarian purpose or could work only in the long run, such as health or education aid. They found a strong growth effect for their preferred category of aid (“short-impact aid”)—and not only when there was good policy in the recipient country.
Again, the original research was scientific; the use of it was less so. Aid advocates once again regarded the new finding as supporting their recommendations. The UN Millennium Project Report in January 2005 cited the CRB study as providing support for the project’s proposal of massive increases in aid.13 The Blair Commission for Africa, in March 2005, recommended an immediate doubling of aid to Africa, and cited the CRB findings as support for its recommendations.14 Unfortunately for these recommendations, researchers again subjected the positive aid findings to further scrutiny and found them wanting. The chief economist of the International Monetary Fund, Raghuram Rajan, and IMF researcher Arvind Subramanian subjected the CRB finding to statistical testing. They used the simplest specification to control for possible reverse causality from adverse country characteristics to aid receipts, and a standard specification for the determinants of growth. In their May 2005 study, Rajan and Subramanian found no evidence that either “short-impact aid” or any other type of aid had a positive effect on growth.15 For good measure, they also tested the Burnside-Dollar hypothesis yet again, and found no evidence that “aid works in a good policy environment.”
They also considered some alternative explanations as to why foreign aid does not raise growth. One well-justified complaint about aid is that it is often tied to the purchase of goods and consultants from the donor country, which may prevent the aid from bringing much growth to the recipient country. Another possibility is that the donor country gives the aid for political reasons, which again may limit the aid’s effectiveness. There is one simple test of these explanations—only aid from national aid agencies (bilateral aid) is tied, while aid from the World Bank and regional development banks (multilateral aid) is not. Similarly, bilateral aid is far more politicized than multilateral aid. Rajan and Subramanian found, however, that there was no difference between the effects of bilateral and multilateral aid on growth. Another test they did was to see if having a high share of aid coming from Scandinavian countries (which are less motivated by political alliances and do less aid tying) was associated with faster growth—they found it was not.
With so little light shed by statistical studies of growth, the big picture is perhaps still useful in evaluating the aid and growth relationship. Do we believe that African growth would have declined even more sharply from the mid-seventies to the present but for the tripling of aid as a percentage of income?
There is another aspect to both the Burnside-Dollar and CRB studies that aid agencies and advocates have chosen to emphasize much less. To the extent that they found any growth effect at all, both Burnside-Dollar and CRB found that the larger the aid already was, the smaller was the additional growth benefit from that additional injection of aid. In the CRB study, their category of aid had a zero effect
on growth when it reached 8 percent of the recipient’s GDP, and after that the additional aid had a negative effect on growth. This feature of their results directly contradicts the Big Push reasoning, which is that small sums don’t help because you need a sufficiently large mobilization of aid to fix all the big problems simultaneously (that’s why it had to be a Big Push). This theory implies that the larger the aid is already, the larger the additional growth benefit from an additional injection of aid. This is contrary to CRB. There are already twenty-seven countries with aid receipts over the 8 percent of GDP at which the CRB-estimated effect of additional aid turned negative; if the donors adopt the current Big Push proposals, virtually all low-income countries (forty-seven of them) will be far above that level.16 Unfortunately, the Blair report and the Millennium Project report select research results to support a Big Push idea that is contradicted even by the selected studies themselves.
We can also check on some of the intermediate steps in the aid and growth story. Jeffrey Sachs and co-authors previously predicted that large aid increases would finance “a ‘big push’ in public investments to produce a rapid ‘step’ increase in Africa’s underlying productivity, both rural and urban.17 Alas, we have already seen this movie, and it doesn’t have a happy ending. There is good data on public investment for twenty-two African countries over the 1970–1994 period. These countries’ governments spent $342 billion on public investment. The donors gave these same countries’ governments $187 billion in aid over that period. Unfortunately, the corresponding “step” increase in productivity, measured as production per person, was zero. Perhaps part of the reason for this was such disasters as the five billion dollars spent on the publicly owned Ajaokuta steel mill in Nigeria, begun in 1979, which has yet to produce a bar of steel.18
What about the elusive “takeoff” into self-sustained growth? If we define “takeoff” as a one-time shift from zero growth to sustained positive growth, there are surprisingly few countries whose development experiences fit this description. Most countries that escaped from extreme poverty did so with gradually accelerating growth, sometimes punctuated by crises of zero or negative growth. Japan is the only rich country that became rich by means of a takeoff. In more recent data, there are only eight countries (all in South and East Asia) that had a takeoff in the period 1950–1975: China, Hong Kong, India, Indonesia, Singapore, South Korea, Taiwan, and Thailand. Three of the eight countries had aid-to-GDP ratios above the norm: Indonesia, South Korea, and Taiwan; in the others, aid did not play an important role in their takeoff. Moreover, other countries got high foreign aid over this period and did not take off. Statistically, countries with high aid are no more likely to take off than are those with low aid—contrary to the Big Push idea.
So the aid Planners keep pouring in aid resources with the fixed objective of stimulating higher growth, although evidence does not support an effect of aid on growth.
The Problem of Evaluating the White Man’s Burden
One thing that makes the aid debate so contentious is that it is not easy to evaluate the effect of Big Pushes. Actually, one argument against Big Push programs is that they are so hard to evaluate. All of the major interventions of the White Man’s Burden have similar evaluation difficulties.
My daughter Grace asked me several years ago as we were driving on the Washington Beltway, “Daddy, why do ambulances make so many accidents?” Of course, now that she is nine, Grace knows that the presence of an ambulance at every accident is a consequence rather than a cause of the accident. The presence of the IMF and World Bank and aid agencies at country crises is surely a consequence rather than as a cause of the accident. This is the selection effect—ambulances show up at car wrecks, not at tailgate parties. This is the same as the reverse causality problem just mentioned about foreign aid. Once you control for the selection effect, you find that things could have been even worse without the aid. This is what is called the counterfactual question: How does what happened with the White Man’s Burden compare to what would have happened without the White Man’s Burden?
There are several approaches that can partially (but not completely) resolve the selection problem and address the counterfactual question for the big programs of the White Man’s Burden. One is to find factors that are not themselves determined by an economic crisis and to ask if the variation in the White Man’s Burden programs associated with those factors had positive or negative effects. If some ambulances just patrolled a neighborhood because the mayor lived there, we could evaluate the effect of the ambulance patrol on survival from heart attacks by comparing what happened to the heart attack victims who lived next door to the mayor with what happened to victims elsewhere. All the statements I make earlier about “controlling for reverse causality” are based on some method like this. The method is never perfect. For example, it won’t work if being the mayor’s neighbor has a direct effect on your survival fitness that has nothing to do with the ambulance patrol—that would contaminate the comparison between the mayor’s neighbors and others.
Another is to analyze cases where there were repeated White Man’s Burden efforts. If ambulances keep showing up at the accident, but the injured still do not get any help for their injuries, you would question how good the ambulance service is. Unfortunately, these methods are not always available, but we still need some way of judging real-world programs that are going ahead anyway. The last resort, which is far from perfect but still provides insight, is simply to describe the results of a program or intervention. If a program is associated with a disastrous outcome, you need to believe that things would have been even more disastrous without the program. If all the ambulance patients are always DOA at the hospital, it’s hard to believe that the ambulances are doing any good. This book will use all of these methods.
Alternative to the Legend of Development
Fortunately, there are some people who work on aid and poverty who do have a more neutral, modest mind-set. These are mainly academic economists, who are woefully short of a plan to eradicate poverty or achieve world peace. They are not good visionaries and are terrible at public relations. They experiment and come up with smaller but more useful things that outsiders can do to help the poor, which they subject to ruthless testing to see if they really work.
With smaller interventions, more rigorous evaluation is available to address the counterfactual question. One scientific method used is the controlled experiment. The control group represents what would have happened to the treatment group without the treatment. The difference between the two groups is the effect of the treatment.
The researcher must choose both groups randomly—say, a lottery determines who is in the treatment group and who is in the control group. If you assign people based on some other criteria, then the difference between the treatment and the control groups could reflect the selection criteria rather than the treatment. For example, if you assigned people with more severe problems to the treatment group, then you could get a spurious negative effect of treatment. (You don’t want to test the effect of ambulances on health by comparing the health of ambulance patients to that of the man on the street.) Conversely, if you assigned those with the most potential to benefit from the treatment to the treatment group, then you would get an overestimate of the treatment effect.
The U.S. Food and Drug Administration (FDA) follows this approach when it decides if new drugs work. It first does randomized treatment and control groups. If the drug works for the treatment group compared with the control group, then everyone gets the drug.19 The FDA may feel a stronger incentive to use scientific methods than aid agencies because it is democratically accountable to voters, who are the same group that will be using FDA-approved drugs. If the drugs do not actually work among the general population, or if they generate side effects that kill off the patients, the new drug users (or their survivors) will complain to the politicians. The politicians will put the heat on the FDA, which will then take more care to test scientifically what reall
y works but does not include bad side effects. The intended beneficiaries of the aid agencies—the poorest people in the poor countries—don’t have a similar way to put heat on the agencies.
The Dutch aid organization International Christian Support Fund (ICS) distributed deworming drugs to schoolchildren in southern Busia district, Kenya, where 92 percent of children were infected with intestinal worms that caused listlessness, malnutrition, and pain. Economists Michael Kremer of Harvard and Edward Miguel of Berkeley took the randomized approach in assessing the effects of deworming drugs. Kremer and Miguel studied programs that administered drugs and that conducted worm-prevention education for schools in Busia district, Kenya. The ICS project phased in the programs over three years, so there were three groups for Kremer and Miguel to study. In the first phase, phase I schools could be compared to phase II and III schools. In the second phase, phase I and II schools could be compared to phase III schools. Kremer and Miguel were able to identify a positive effect of deworming drugs on school attendance and a zero effect of deworming education on worm infection rates. The deworming drugs decreased school absenteeism by one quarter. “Pupils who had been miserable now became active and lifeful,” said schoolteacher Wiafred Mujema.20
Kremer and Miguel’s practical scientific approach identified a way to help children stay in school (give them deworming drugs) and also identified other methods that didn’t work (educate children on behavior to prevent worm infection). After the results came in, ICS expanded its deworming program; it now covers all of Busia district plus neighboring Teso district. Other aid organizations have imitated the deworming program around the world. If this practical, critical approach spreads, much more of the foreign aid dollars available could actually reach the poor! And then maybe aid advocates could make the case for more foreign aid.
Not all scientific work is on randomized trials of individual interventions; some is on statistical analysis of aggregate data. And not all findings are positive; some tell policymakers and aid officials what not to do. Researchers Thorsten Beck, Asli Demirgüç-Kunt (both at the World Bank), and Ross Levine (Brown University) studied whether small and medium enterprises (SMEs) were catalysts for poverty reduction. The aid community believes in SMEs’ catalytic role, with the World Bank having lent $10 billion to support SMEs over the last five years.21 USAID spends about $170 million a year on microenterprise promotion.22
B000QJLQXU EBOK Page 6