Aid to be paid on delivery?

The idea behind cash on delivery is that aid recipients are paid for achieving various development benchmarks, plus bonuses for additional achievements.

March 18, 2011


For the past few months, the New York Times has featured an excellent series of online articles by Tina Rosenberg looking at solutions to social problems, many of them development-related. Each time she writes a first post presenting an idea, and then a follow-up post a few days later responding to comments.


In January, she wrote about conditional cash transfers to the poor (here and here); a few weeks later she discussed microcredit and microconsignment (here and here); in mid-February she discussed the possibility of drastically improving healthcare in developing countries without needing to train (and retain!) new doctors (here and here). All of these are well worth reading.


Her most recent entries in the series appeared on March 14th and 18th, and address the idea of changing the protocol for foreign aid to a cash-on-delivery model. The articles discuss the promise and some of the potential pitfalls of using aid to pay for past, rather than future, performance. Much of the first article draws on a fine book by Nancy Birdsall and William Savedoff, Cash on Delivery: A New Approach to Foreign Aid, which was just published in a revised edition by the Center for Global Development.


The idea behind cash on delivery is that aid recipients are paid for achieving various development benchmarks (for example, $20 for every child that finishes elementary school), plus bonuses for additional achievements. The book highlights several important features of the COD scheme: it is likely to be more transparent than traditional aid, it requires less heavy handed oversight on the part of donors (who traditionally want a say in how various goals are achieved), and it can make more efficient use of the available aid funds.


Rosenberg highlights another potential contribution: it may increase public support in donor nations for foreign aid. After all, donor state citizens would know precisely what they got for their money, so to speak. Rosenberg suggests that this might contribute to eliminating the persistent misperception among most Americans that the United States gives large amounts of aid and that most of it is wasted.


COD is an intriguing notion, and it will be exciting to see it put into practice (apparently the British government will begin a pilot project soon). However, as Rosenberg hints in her posts, this may be unexpectedly complicated. Birdsall and Savedoff claim in the preface to their book that "the only true preconditions for this new approach are a good measure of progress and a credible way to verify it," suggesting that this is fairly straightforward. But of course it isn't.


Rosenberg cites a similar project by the Global Alliance for Vaccines and Immunization, which learned that they had underestimated the required management skills to implement a COD vaccination scheme and that, moreover, there was a big temptation to exaggerate accomplishments (since remuneration was linked to those accomplishments). There is no reason to think these problems will be any less for more traditional development assistance projects.


At the moment, COD in development assistance is little more than an intriguing notion. It may turn out that it simply will not work in practice. On the other hand, it may turn out that it will work much better than traditional development assistance, as some of its advocates suggest. I suspect, however, that it will work reasonably well under certain conditions and for certain projects, but most of the time will not measurably outperform traditional aid.


For COD to be preferable to traditional forms of aid, reliably measuring progress must be cheaper and/or less invasive than is the case with oversight mechanisms for the latter. There is no reason to believe that this will hold true in practice; at least not as a general rule.


Advocates of COD also emphasize benefits in terms of accountability and transparency, for recipient and donor populations alike. For such benefits to become reality, information about these projects has to be both widely disseminated and easy to understand. Here, too, it is not obvious that this will be easier to achieve for COD than it is for traditional aid.


Traditional aid programs have not been all that transparent or accountable, but that is not inherent in the method. Over the past few years, AidData and others have been changing this, making extensive project information available to anyone who is interested, in donor countries as well as recipient countries. Of course, data availability is not the same as awareness and understanding, and it is possible that COD projects are somehow inherently easier to understand. Rosenberg's account suggests that she believes they are.


Initial tests of COD will probably simply see whether the approach can work at all. A real test, however, ought to compare two projects with the same goals, one set up along traditional

lines, and one along COD lines, each of which invested the same quantity of resources both in the project itself and in disseminating information about it. Only then could we ask:

1. Does the COD project achieve better (or at least the same) outcomes with the same resources?

2. Are donors and recipients better able to understand the COD project than the traditional one?

3. Are donors and recipients more likely to seek information about COD projects than traditional ones?


In fact, it might be possible to investigate the second and third questions before even testing COD in practice. Take a few projects from the AidData database for which post-project evaluations have been done, so we know performance outcomes. Translate the project description into a COD format: for example, instead of saying the project funded the construction of a school, say that it paid the recipient 20$ for each student attending the school for a year. Then conduct a survey (ideally in both donor and recipient countries), and gauge how people interpret and react to these competing descriptions of the same project:

Does one seem easier to understand than the other?

Does the traditional format sound more or less efficient / wasteful?

Do donors and recipients react in the same way to these descriptions?


Rosenberg argues that one benefit of COD aid might be eliminating persistent misperceptions among American citizens about how much aid the United States gives, and how much good it does. A survey along the lines just described might go some way towards telling us whether Rosenberg is correct on this point. Even if she is wrong, COD might still be preferable from the point of view of development outcomes, but it will be valuable to know what its other implications are before investing large sums of money.