Corruption and Pay-for-Performance Aid – William Savedoff and Charles Kenny

Summary

---

Podcast Summary

The CGD Podcast

Episode: Corruption and Pay-for-Performance Aid – William Savedoff and Charles Kenny
Date: January 13, 2014
Host: Lawrence MacDonald
Guests: Charles Kenny (CK) and William Savedoff (WS), Senior Fellows at the Center for Global Development

---

Episode Overview

This episode delves into a critical debate in the international development community: Is results-based “cash on delivery” (COD) or pay-for-performance aid more or less vulnerable to corruption than traditional input-tracking models? Host Lawrence MacDonald speaks with Charles Kenny and William Savedoff about their research challenging conventional assumptions about corruption and aid modalities. Through real-world examples, relevant research, and memorable metaphors, the discussion advocates for a shift in focus—from accounting for every input to measuring and rewarding actual development outcomes.

---

Key Discussion Points & Insights

1. Framing the Corruption Question in Aid

• Many donors are wary of COD/pay-for-performance aid, worrying that “the money might get stolen” because outputs (not inputs) are being paid for.
• The assumption: input-tracking (every receipt, robust audits) is the best guard against corruption.

Notable Quote:

"Making corruption...the reason why poor countries are poor is a bit like in the United States saying, oh well, you know why these people can't get a job? It's because they're lazy...I think it's blaming the victim and I think that's bad."
— Charles Kenny [01:21]

---

2. Problems with Existing Anti-Corruption Controls

• WS and CK contend that standard accountability practices (audits, procurement, documentation) are costly and often ineffective.
  • Example: The World Bank spends at least $30 million/year directly on audits, with additional hidden costs as staff focus on compliance [03:03].
  • Recipient countries often assign their most capable people to compliance rather than development.

Notable Quotes:

"It's really a massive burden on the aid agencies. But more importantly, it's a massive burden on recipient countries who put some of their best people on making sure they comply...rather than, I don't know, actually trying to get more kids through school and tested at the end of them as it might."
— Charles Kenny [03:27]

---

3. Are Input-Tracking Controls Effective?

• “Red flag” detection systems (few bidders, incomplete documentation, etc.) don’t reliably indicate real corruption.
  • Research finds that projects with more “red flags” aren’t reliably the problematic ones [05:00].
• Cross-country data: World Bank’s identical oversight mechanisms failed to equalize costs in road projects—costs were still higher in high-bribery countries [06:03].

Notable Quotes:

"If the control mechanisms were working, you'd hope, you'd assume the prices would be the same...In countries that are more corrupt...the World Bank pays more to build that road."
— Charles Kenny [06:36]

---

4. Incentive Structure: Status Quo vs. Results-Based Aid

• Status Quo: Honest implementers are overburdened with compliance, while corrupt actors just need to fabricate documentation.
  • The metaphor: Counting beans planted vs. counting beans harvested.
  • Dishonest actors evade detection by producing correct paperwork even with no meaningful outputs [08:05].

Notable Exchange:

"All they had to do was show the records that said they bought the thing or they planted it or whatever and they get the money. So it's much easier for the dishonest person to divert money and not do any work than for the honest person..."
— William Savedoff [09:08]

---

5. Does Outcome-Based Payment Reduce Corruption?

• Paying for deliverables (e.g., students tested or “beans harvested”) aligns incentives better.
  • Honest actors focus on results.
  • Dishonest actors must still deliver results to get paid, raising the bar for effective theft and reducing opportunities for corruption [10:01].
  • Even if some money is diverted, the loss is proportional to lost outcomes—making the scope of corruption more visible and relevant.

---

6. Case Study: Afghanistan’s Health Program [12:37]

• The Special Inspector General for Afghanistan Reconstruction (SIGAR) recommended shutting down a highly effective, low-cost health program due to inadequate record-keeping, despite strong evidence of positive impact on health outcomes (reduced child, maternal, and adult mortality).
  • Example of the perverse effects of prioritizing documentation over real-world results.

Notable Quotes:

"I can tell you one thing that program has done. It has saved thousands upon thousands of Afghan lives...But because we don't have all the receipts in a nice little book, SIGAR is suggesting shutting it down."
— Charles Kenny [13:13]

---

7. Rethinking Aid Accountability Systems

• Agencies (like the World Bank, Global Fund, or USAID) often add outcome measurements on top of legacy input-tracking, increasing, rather than decreasing, the reporting burden.
  • Real gains come from shifting more focus to outcomes and away from onerous input-tracking, not layering both [18:36].

Notable Quotes:

"We're going to do everything we did before and require everything we required before, but now we're going to add on top...what results you had. There's some gain to that. But as long as you continue to have very detailed requirements about where the money went...that tends to overwhelm any attention to...outcomes."
— William Savedoff [18:36]

---

Notable Quotes & Memorable Moments

• "We still need bean counters. We just need them counting, you know, the beans that we want delivered, not the beans we put into the process in the first place."
  — Charles Kenny [21:56]

• "It's a mindset change...There's a proposal here to align the incentives in a way that's going to maximize good outcomes and minimize corruption. But there's still going to be some."
  — Lawrence MacDonald [20:13]

---

Timestamps for Key Segments

• 01:21 – Charles Kenny critiques the narrative that corruption is the cause of poverty.
• 03:03 – Costs of audits and procurement compliance in traditional aid.
• 05:00 – Why “red flag” tracking doesn’t identify corruption reliably.
• 06:36 – Road project costs and input controls' failure to equalize outcomes.
• 08:05 – How traditional compliance makes life harder for the honest, easier for the unscrupulous.
• 10:01 – Results-based aid: honest actors benefit, dishonest face higher hurdles.
• 12:37 – Afghanistan health program case study: perverse outcomes of input-focus.
• 18:36 – Adding result tracking to inputs increases, not decreases, reporting burden.
• 21:56 – “We still need bean counters…just different beans.”

---

Language & Tone

Throughout the episode, the tone is clear, wonky but accessible, and sometimes gently irreverent. The guests use memorable analogies (“bean counting”) to demystify complex issues, and the style encourages curiosity and critical reflection.

---

Conclusion

This episode presents a persuasive argument, grounded in research and practical examples, that traditional input-tracking does not reliably prevent corruption and may come at a significant opportunity cost. Instead, a shift toward results-based aid—measuring and rewarding true outcomes—can better align incentives, improve development impact, and make the cost of corruption more visible and relevant. The hosts leave the listener with food for thought and pointers to further reading.

---

For more details, see Charles Kenny and William Savedoff’s CGD working paper and policy brief on this topic at www.cgdev.org.

Transcript

[A]

Welcome to the Global Prosperity wonkcast. I'm Lawrence MacDonald. With me in this gorgeous new studio at our new offices in the center for Global Development are Charles Kenny and Bill Savadoff. They're senior fellows here at CGD and they have written a CGD working paper and now an accompanying brief in which they try to take on a difficult question. And here let me pause for a minute and say that some of our listeners, those who follow the center for Global Development, know that we've been putting forward a new idea led by our president Nancy Birdsel, of cash on delivery aid. That is the assistants would pay for results, not for input. So instead of paying for school buildings, you pay when the kids actually come complete school and take a test. That's the classic example. A lot of people have said, well that sounds like a good idea, but how do you know the money's not going to get stolen? Corruption is a big problem in many societies, particularly in developing countries where people might be trying to help. If they weren't corrupt, they probably wouldn't be poor to begin with. So shouldn't we make sure that the money really gets spent on the school buildings? Charles has winced when I said that. Why?

[B]

I think corruption is a big problem for development, but I also think child mortality is a big problem for development and a whole load of problems are large. And making corruption putting corruption on a pedestal and suggesting it's, you know, the reason why poor countries are poor is a bit like in the United States saying, oh well, you know why these people can't get a job? It's because they're lazy. Or you know why these people can't hold down a job? It's because they didn't study hard enough in school. It's their fault. I think it's blaming the victim and I think that's bad.

[A]

I stand corrected. They might be poor even if they had no corruption. There's a lot of reasons for poverty. However, as you say, corruption is one of the problems. People who are providing assistance naturally worry about it. And some of them have said we can't do cash on delivery aid because if you only focus on paying for the outcomes, you won't know if the money got stolen. And you too have unpacked it in a very interesting way in which you contend that in fact there's less space for corruption in a results based payment than in something that tracks payments for the inputs. Bill, how could that be? Don't I do better if I want to reduce corruption? If I insist on seeing a Receipt for everything?

[C]

No, not really what we came. We tackled this in sort of two different ways in the paper. The first thing is we asked whether the existing methods for tracking every single dollar or trying to make sure that it was used for the right things, that the methods we have for doing that don't really work or we don't know if they work. We can't say that putting in more audits and more documentation and more procurement procedures actually does assure that that money gets used for what.

[A]

So let's unpack that one. Before you go to your second point. One of the things in the brief that I thought seems obvious, but I haven't seen it stated that way is these things are really expensive, right. In the case of the World bank, they spend how much per year in auditing and tracking and making sure that things are not corrupt? $30 million. Am I remembering it correctly about that?

[B]

Yeah. And it's hard to do a full accounting because that's sort of the people who are. That's in their job description. But as former workers in aid agencies, I wasn't a procurement expert, but I spent a lot of time worrying about were my projects following the right procurement steps. So it wasn't in my job description, if you will, but I was certainly doing it. So if you add up not only the sort of the official staff working on procurement issues and financial management and then on top of that all of the other staff who run around worrying about it, and we all did a lot, it's really a massive burden on the aid agencies. But more importantly, it's a massive burden on recipient countries who put some of their best people on making sure they comply with every dotted I and crossed t of the World bank procurement regulations, which are big and complicated, rather than, I don't know, actually trying to get more kids through school and tested at the end of them as it might.

[A]

Be, because if they don't comply, then they could be at risk of being alleged corruption and then these flows could stop. So both the funder and the recipient is spending a lot lot of time and energy and as you say, maybe the best and the brightest people involved in preventing the leakage of funds rather than focusing on the outcomes.

[B]

Well, and again, as Bill says, involved in an effort to prevent the leakage of funds that we don't really know even if it works.

[A]

How do we not know if it works? Can you give me an example?

[C]

Well, for one thing, the things that the World bank other aid agencies look at to try to see is this a suspicious project, maybe Money was diverted. Here they have these things that they call red flags. So they'll say, you know, were there too few bidders in something or was the documentation looking poor or something like that? And work that Charles has done with other people has shown that those really aren't correlated. You look at a lot of projects have these red flags, so it looks like everything's corrupt, but then when you compare that to the subset of projects where you actually found problems, they actually don't have more red flags than the other ones. So it's like you're putting a lot of effort in to detect something, something you don't know its actual scale and your detection method isn't all that good.

[A]

And I think there's an example in the brief in the bank where the same detection methods applied across different environments yielded different results depending on the pervasity of bribery within that culture. So countries that tended where the bribes were larger got worse results even though the control mechanisms were identical. That tells you the control isn't working.

[B]

That's absolutely right. So this was on World bank financed road project projects. And the World bank has actually done a great job of putting together a database that allows you to compare exactly the same kind of road and what it costs to build across countries. So, you know, rehabilitating a two lane highway using tarmac. And those costs of doing exactly the same thing are higher in countries where survey evidence of firms suggests that firms are paying higher bribes to government for government contracts. Now if the, even though the control.

[A]

Mechanisms were the same.

[B]

Right. These were all under World bank projects, so the control mechanisms were exactly the same. If the control mechanisms were working, you'd hope, you'd assume the prices would be the same. Nuh. In countries that are more corrupt on this measure of paying bribes for contracts, you pay more bribes for a contract. The World bank pays more to build that road.

[A]

Okay, so we have a situation where those who look at cash on delivery or other performance outcome based payment methods would say, we can't do this because the money can't get stolen. And you two argue that a, with the conventional approach, the status quo, you're spending a lot of money on something that the evidence suggests may not be working very well. So it's kind of a question of what is your counterfactual. Right. It's not as if the current method is working great. And we're suggesting trying something new and untested against something that works pretty well. It's like it's not. If it ain't broken, don't fix it, it is in fact broken. Right. There's something about the incentives for graft within the old system that I want you to explain to me before we go to our new proposal. Can you unpack that? Let's say I'm well intentioned, I don't intend to steal. You've shown that the showing that I'm complying is expensive and may not work. But let's say I'm pretty corrupt and I want to steal. How does the status quo present me with incentives that make it more or less likely that I'm going to steal?

[C]

Well, one of the things we point out in the paper is that in a situation with a traditional project where you're tracking the inputs that go into the program, somebody who's honest has a lot of work to do because they have to try to figure out how they're going to build the program or deliver the services and do it in their difficult circumstances, whatever. So they're working really hard to do that. And at the same time, they have to understand and report to and report back on all these procurement procedures and other kinds of mechanisms that are being required by the aid agencies. So they've got a big, got two jobs. What's ironic is, by contrast, somebody who actually wants to steal money from a project like that has a much easier time because they're not worried about delivering the services or building the bridge. They're mostly focused on how can I get the money to flow and then divert it to some other place. So it's kind of like, you know, we were talking about simple examples like bean counting. You give somebody beans and you say, you know, we want you to count for all these beans. And you come back and they tell you, here's the book where I recorded all the beans that I planted. And you say, okay, well, looks like you planted all the beans, but at the end of the season, there's no harvest. What happened? And they can come back and say, oh, well, you know, the worms ate it, birds came and pulled it out, or it was a bad year for rain, so they didn't grow. You know, there's all kinds of excuses for why the inputs didn't turn into outputs. But all they had to do was show the records that said that they bought the thing or they planted it or whatever and they get the money. So it's much easier for the dishonest person to divert money and not do any work than for the honest person who's trying to get their job done and then also report on all this Stuff.

[A]

So if I'm a mother and I give my daughter some beans to plant, if I ask her to account for all the beans, am I going to get a different outcome than if I tell her I'm going to count the harvest?

[C]

Well, there's been a lot of work on results based or incentive programs in those terms. And it's more, and it's focused mainly on if you actually pay for the beans that are harvested, will the person put in more effort, will they innovate or try to find better ways to grow these things? And what we're doing with this paper was saying that if you pay for the outcome that you want, the mother says she'll, yeah, the mother says, I want you to tell me how many beans you actually grew. Well, she may never know how many of the beans her daughter actually planted. The daughter might have planted half the beans, sold the rest, but she found other ways to water them or cut the shade trees down or something else that got the output that made it seem like, wow, we actually got got the production that we wanted. So it's not that results based things eliminate corruption or stop corruption, it's that it reduces the room for diverting money. And back to the earlier point about dishonest and honest people, what it means is that if you're being paid for the amount of beans you produce, then the honest person has an easier time because all they have to do is their job. They know how to plant, they know how to cultivate and hoe and do all the things to get the outputs up. So they focus on getting the output, the outcomes to happen, whereas the dishonest person has a double job. Now they have to figure out not only how to get all the beans produced, but now they've got to figure out how to do it even more efficiently so that they can skim some of it off.

[A]

Bill, I find this example so helpful. We're going to take a short break. When we come back, I'm going to ask Charles to give us another real world example. Not the parable of the bean counters, but a real story of what happened with U.S. foreign assistance in Afghanistan. This is the Global Prosperity Wonkast. I'm speaking with Bill Savitoff and Charles Kennedy, senior fellows here at CGD, about whether paying for outputs, things like cash on delivery aid, is more or less likely to lead to corruption than traditional aid systems. We will be back in a bit. Welcome back to the Global Prosperity Wonkast. Charles during the break we were discussing this work that our colleague Justin Sandiford did on US Foreign assistance in Afghanistan. I'm familiar with that work, but I hadn't really thought of it in the context of paying for performance. What's the connection?

[B]

Well, the connection is how badly things can go wrong when you focus on receipts and forget all about results. And the baddie in this story, the bad guy, is the Special Investigator General for Afghanistan who is charged by the President to try and reduce fraud and corruption in Afghan aid programs. Providing aid to Afghanistan, Department of Defense, usaid, all parts of the American government.

[A]

You say the guy. In fact, it's an office of like 150 or 200 people or something. There's one budget.

[B]

Yes, it's a big. It's a big office sucking up a lot of money right there. Especially as they fly back and forth to Afghanistan a lot now. And cigar, that's the acronym, cigar, took a look at a program in Afghanistan which is run through the Afghan Ministry of Health and is providing basic health services to about 90% of the population of Afghanistan. And he said we should shut this program down now. Why shut the program down? Because he wasn't confident that the Ministry of Health in Afghanistan was keeping records well enough. We couldn't really tell if all of the money was getting from point A to point B and exactly the way it should do under the various procurement regulations of the United States. I can tell you one thing that program has done. It has saved thousands upon thousands of Afghan lives. If you look at the mortality, child mortality rate, the maternal mortality rate, the adult mortality rate in Afghanistan, it's been dropping of late and it's been dropping of late, in no small part thanks to this health program. It's saving people's lives at an incredibly low cost. This is, you know, Obamacare is golden pearls compared to this. This is really cheap stuff. This is some of the most cost effective health interventions anywhere in the world ever. We know that the program is working, but because we don't have all the receipts in a nice little book, the cigar is suggesting shutting it down.

[A]

And I think when Justin wrote about this the first time, his headline in his blog was something like the biggest US victory in Afghanistan you have never heard of. Yeah, I mean, all we ever hear out of Afghanistan is how terrible things are. Things are surely plenty bad. But meanwhile, there's this small, relatively low cost compared to the military intervention. This is like rounding errors, right?

[B]

Yes, indeed. It's peanuts. And it is very effectively spent. Peanuts in that we do have not perfect, but good enough survey evidence to say, look this is having real impacts on health outcomes on the ground. And this program that we actually can honestly say is working is the bit that we're going to shut down, as opposed to a lot of the other stuff we're doing in Afghanistan, which, as you say, the record on how well we can say it's working is perhaps not quite so strong.

[A]

But again, so what does it take in a situation like that where you have, in this case, good survey evidence showing that the health program really is reducing infant mortality, reducing maternal mortality, extending life expectancy, but the bureaucracy is set up to track the receipts. How do you change something like that?

[B]

You get the bureaucracy to focus on what the US Taxpayer actually cares about. And what the US Taxpayer in this case actually cares about in a health program in a developing country is, is it saving lives? And you measure that and you make payment based on that. You make payment based on the outcome you are trying to achieve. And then corruption can't have the biggest nasty effect corruption can have, which is to stop your program working.

[A]

So in the short term, it'd be nice if the folks in this agency, the acronym CARGAR, would sort of back off, but in the longer term, it would be better if the programs were set up to pay for the outcomes.

[B]

Absolutely.

[C]

Or even if they don't pay for the outcomes, to make clear to have very good ways of measuring those outcomes and demonstrating them. And in all fairness to Cigar as well, if they were reviewing the program and USAID had done a good job of saying this program cost this much and this is what we achieved, that can be part of the auditing process. They're like, is, is the US Government getting some output from these programs? A balanced report would have started with the results and then you'd have a context for understanding if, you know, if half percent of the money didn't go where you wanted it to, it'd still be nice to know and do something about that. But it's a very different context. When you see a program that's having mainly huge positive impacts and has a few problems, you don't know the scale. This is what happened with this is what continually is happening with the Global Fund is that they have a system in place. They'll start reporting on money that they can't account for, something like that. But the scale of those discoveries are so small relative to the overall program that they have that it's never discussed in a proper perspective.

[A]

So the agencies themselves, whether it's the Global Fund, USAID or the World bank, can, without any outside authorization, just in terms of how they set up the projects themselves and what they present in terms of the. They can shift the focus at the margin to focus more on outcome and less on tracking. But I want to, I'm thinking about the World bank here and as some listeners will know, they have. It's sort of, if you will, payment for outcomes is the flavor of the month. It's a very trendy thing. So every agency wants to have one. The Bank a couple of years ago put in place something they call payment for results, P4R. My understanding of that is that people like Nancy Birdsall and perhaps you, Bill, who've looked at it, worry that they've kept in place all of the old input tracking mechanisms and then they add the outcome tracking mechanisms on top of that. And so you aren't really changing the incentives, you're just creating an additional burden both for those providing the funds and those who are receiving it, is that the. Have I captured that right?

[C]

You've captured the issue right. My understanding with the P4R mode in the World bank is that they actually have possibly separated that out, but that is the standard. If you look at the history of results based payment approaches in the last 30 years, it tends to be we're going to do everything we did before and require everything that we required before, but now we're going to add on top of it that you have to report about what results you had. And that's not really. There's no real. Well, there's some gain to that. But I think the cigar experience shows you that as long as you continue to have very detailed requirements about this, reporting about where the money went and exactly what it bought, that tends to overwhelm any attention to the results and to the, to the outcomes. And that's the other part we were trying to get at in the paper was to get people to think that the cost of corruption isn't. I mean, there is a cost of corruption. If you think about counting how much money was diverted or how much money was stolen or whatever you want to call it. But what is invisible in all these programs, what we call the foregone benefits, the benefits that didn't happen because of the corruption. And in that case, you can steal, you know, the bolts from a bridge and have massive impact on a project, you've lost 100% of the benefits you may have. Bridge falls down, bridge falls down. You may only, you know, you may have only diverted 1% of the funds, but the entire bridge falls down. That's a much bigger problem than if you skimmed off a quarter inch of the asphalt, but the road got built and it lasted for maybe not the 20 years it was programmed for, but it lasted for 10.

[A]

Charles, how do we make this change? I think it's a mindset change where you say, okay, there's the potential for a certain amount of corruption in whatever you do. There's a proposal here to align the incentives in a way that's going to maximize good outcomes and minimize corruption. But there's still going to be some. But I can imagine if I'm a taxpayer watchdog group, I'm still going to say, yeah, but I want my money tracked. I want to see those receipts, and by God, I want to make sure that every receipt is genuine. And I don't want to see my money wasted because it's my money.

[B]

I think the last thing you said is what people really care about when they think through things again. The reason that take the United States, the reason that programs like PEPFAR are popular is that Americans can say, I know my aid dollars are going to keep people alive who would otherwise be dead from aids. That's what they care about. That is the sort of the moral push behind spending the money. What we need to be measuring is are we keeping people alive? Now, in the case of pepfar, we know the answer to that. The answer is yes. It may not be the most cost effective way to spend health dollars, but by golly, we know it to save lives. And it's fairly easy to see if we focus in on that. Honestly. Honestly. It reduces the scope for corruption better than the input tracking process, which has sort of no record of success in terms of stopping corruption.

[A]

So everybody will be happy. Except maybe the accountants.

[B]

Yeah, but not only the accountants can do something else.

[A]

They can count the outcomes.

[B]

The accountants, indeed. We still need counters. We still need bean counters. We just need them counting, you know, the beans that we want delivered, not the beans we put into the process in the first place. It's different beans they need to be looking at.

[A]

I can't think of a better place to leave it than that. Bill, do you have any final comments?

[C]

No, I think we're just redefining what bean counters should be counting. And I think that's a great idea.

[A]

Thank you both for joining me on the show. This is the Global Prosperity Wonkcast from the center for Global Development. My guests today have been Bill Savitoff and Charles Kenney, and we've been discussing whether payment for performance or cash on delivery is more or less conducive to corruption than the way that we typically provide assistance. I leave listeners to draw their own conclusion. And if you're not persuaded by our bean example, do come to the CGD website and take a look at both the working paper and the brief that Charles and Bill have written. You can find the Wonkast online on itunes and on Stitcher. Just search for Wonkast or CGD and sign up to hear a new interview every week. Until next time, I'm Lawrence MacDonald. Thank you for listening.