Transcript

Cold open [00:00:00]

Paul Niehaus: What economists mean by general equilibrium is that one person’s actions have broader ripple effects and ramifications on other people through the economy, because we live in this interconnected system. And if we do a study where we just look at what happens when one household gets money and another doesn’t, and we compare the two, we might see differences between those two households in terms of what stuff they have or how they behave or spend their time, but we aren’t picking up those broader impacts: What does that all mean for the rest of society?

With cash transfers, the most obvious example to think about, to make this very concrete, is that whenever people get cash transfers they use them to transact. So there’s somebody else that they go out to and they buy food, or they buy an asset, or maybe they even put money in a savings account — that actually rarely happens — but whatever it is that they do, there’s some counterparty to that transaction, so it’s almost mechanical that that counterparty is going to be affected at least to some extent by it. And that means if we want to think about the total consequence of the transfers, we want to take that into account.

Luisa’s intro [00:00:58]

Luisa Rodriguez: Hi listeners, this is Luisa Rodriguez, one of the hosts of The 80,000 Hours Podcast.

In today’s episode, Paul Niehaus makes the case that rather than trying to predict what kinds of aid programmes might benefit the global poor, we should give them cash so they can decide what they most need for themselves.

This is the argument that led Paul to cofound GiveDirectly, the nonprofit giving cash directly to the poor back in 2009.

Since then, they’ve conducted dozens of studies to understand the benefits of giving cash. In this interview, we discuss the findings from some of the most important studies they’ve done so far.

Specifically, Paul and I discuss:

• How the impacts of GiveDirectly compare to USAID employment programmes.
• Whether there have been any USAID programmes that outperform GiveDirectly.
• The empirical evidence on whether giving cash directly can drive meaningful economic growth.
• The case for universal basic income, and GiveDirectly’s UBI studies in Kenya, Malawi, and Liberia.

We also spend some time discussing GiveDirectly’s first and only major case of fraud, in which just under $1 million of GiveDirectly’s funds were stolen from beneficiaries in DRC.

Finally before we get to the interview, a quick reminder that over on our other feed, 80k After Hours, we’re releasing 20-30 minute highlights episodes. These aren’t necessarily the most important parts of the interview, and if a topic matters to you I do recommend listening to the full episode — but for those who understandably don’t have time to listen to our full episodes, we think these will be a nice upgrade on skipping episodes entirely.

At the moment we’ve released highlights for 14 interviews from earlier in the year, including some of my personal favourites:

• Michael Webb on whether AI will soon cause job loss, lower incomes, and higher inequality — or the opposite
• Rohin Shah on DeepMind and trying to fairly hear out both AI doomers and doubters

If you’d like to try out those highlights episodes you need to subscribe to our second feed 80k After Hours.

Without further ado, Paul Niehaus.

The interview begins [00:03:11]

Luisa Rodriguez: Today I’m speaking with Paul Niehaus. Paul is an economist at UC San Diego and entrepreneur working to end extreme poverty. He’s also cofounder, former president, and a current director at GiveDirectly, a nonprofit that gives cash directly to some of the world’s poorest. Thanks for coming on the podcast, Paul.

Paul Niehaus: Hey, thanks for having me.

The basic case for giving cash directly to the poor [00:03:28]

Luisa Rodriguez: I hope to talk about whether giving cash to the poor beats conventional aid, and whether it might do even better than that by boosting an entire region’s economic growth. But first, can you basically just explain the key case for giving cash directly to the poor?

Paul Niehaus: Yes. But I would flip the question, because I think that in fact the better question is, “Why not?” So let me explain what I mean by that. We’ve been at this project of trying to accelerate global development poverty reduction for quite a long time now — 50 or 60 years or so since that became a thing. And we’ve spent many trillions of dollars over that time period — some of it well; I think there have been some big successes.

But the thing I want to emphasise is that almost all of that money has been spent by people other than the folks living in extreme poverty who we’re trying to help. And yet, if you look at the data, I would say that in aggregate, if you were to look across the evaluations that we’ve done — and it’s now been 20 years or so since we started to do these credible experimental tests that really tell us what works and what doesn’t — generally speaking, people living in extreme poverty have a better track record of putting money to use in ways that improve their lives than the top-down approach where we have programme designers plan and allocate money. There are of course exceptions in both directions, but as a general observation, I think that’s a pretty fair characterisation.

So I think we want to flip it around and say: There are lots of obvious reasons and lots of good data why it makes sense to let people living in poverty decide how this money that’s meant to benefit them gets used — when do we think we can do better than that by coming in and doing something different from what they would have done? I think there are cases where that’s true. But that’s the way we’ve tried to position GiveDirectly: as saying, first, we think we should do a lot of this; but second, we think this should be a bit of a prompt and a challenge to us when we think about other ways of helping people living in poverty, to ask why do we really think that we, as outsiders, can do better?

Luisa Rodriguez: Right. So the default is we have these resources that we’d like to give to help people in poverty. Let’s give them the resources as a default. Then if we can confidently do much better for them by thinking through some alternative intervention, great, that could make sense. But whoever wants to give this intervention should have to be confident that that’s actually going to beat cash.

Paul Niehaus: Yes. One of our earliest supporters and a dear friend of mine, Mark Lampert, once said to me, “The way I think about it is, imagine that this money were already in the hands of people living in poverty. If I could, would I want to tax it and then use it to finance other projects that I think would benefit them?” I think that’s an interesting thought experiment — and a good one — to say, “Are there cases in which I think that’s justifiable?”

Luisa Rodriguez: Yeah, that does help pump that intuition for me. For the most part, I can’t really imagine wanting to tax the ultra poor. There are very few things that I’d be confident enough I could do better with that money than them choosing to spend it in ways that make sense for them. I guess to make it more concrete, how do people spend the money they receive via cash transfers? I’m sure there are loads of things, but what are some examples?

Paul Niehaus: So first, it’s very important: How do we know? What data do we look to or how do we answer your question? Just to underscore this point, I think up until around 2000 or so, the answer is we actually didn’t know too much about that, or about other approaches, because there had been very little high-quality empirical testing. It’s only since then, with the advent of the randomised controlled trial movement, that we have generated an enormous amount of data on that. So we’ve learned a lot in the last 20 years. And partly I say that just to say that if you came up, like I did, hearing from people that you can’t just give money to people living in poverty — that that doesn’t solve the problem, or it’s not really going to do any good — then I heard that too. But that comes from a time period in which we actually didn’t have much rigorous data to speak to the question.

So what actually happens in all of these studies, these experimental trials — where some people are given money, others are not, and then we compare outcomes and compare their budgets and their spending patterns to see what they’ve done — a lot of different things happen. And that’s part of the point: cash transfers give people a lot of flexibility to do what they want with it. So there is no one answer to the question of what do people living in poverty do with cash transfers? Each person’s going to do something different — and that’s intentional, and that’s by design. Just as if I were to say, “How does Luisa spend money?” — we’d say lots of different things, right? And probably different from the things that Paul spends money on.

But that having been said, there’s pretty consistent evidence that people spend money on things that are positive, that improve their lives. Some of those are things that improve their lives today, like better food, better nutrition. Some of those are things that are going to improve their lives in the future — whether because they’re buying durable goods; or building better housing that’s going to last for a while; or because they’re investing in productive assets that are going to increase their income; or because they’re spending on schooling for their kids, which will improve their lives in the long run.

And then we don’t see evidence of the negative things that people were most worried about. Things like what if people just stop working and don’t try to improve their lives on their own? Or what if people misuse the money in ways that are actually harmful for them, like spending it on alcohol and tobacco and things like that? We generally haven’t seen evidence of those things.

Luisa Rodriguez: Yeah. To get even more concrete, are there specific kind of case studies that you could give? Maybe one or two? I have a vague sense of what “durable goods” might mean, but are there a couple of people or cases that come to mind?

Paul Niehaus: I think within that category of durables, housing is really the thing that stands out — in the sense that for a lot of GiveDirectly projects, for example, that’s been a thing that a sizable share of people have invested in. So very concretely, literally concretely, that might mean going from having a mud floor to a concrete floor, or having wattle-and-daub mud walls to concrete walls, or going from having a thatch roof to a metal roof over your head. Or just having a bigger house with enough space for more people in your family. Things like that. So that’s very common.

You can also kind of pick individual people and go and talk to them, and say, “Hey, what did you do with the money?” That’s a little tricky because then you have to think about the counterfactuals, and you have to think about maybe they just know what sort of things Western NGOs like to hear. So I would always take that with a grain of salt. But you can also get some very colourful things. One of my favourites was somebody who used transfers from GiveDirectly to buy musical instruments and start a band, and they started touring around. So it’s actually a programme to stimulate the arts, it turns out.

Luisa Rodriguez: Wow, cool. So I guess the key theme here seems to be GiveDirectly just values recipient choice super highly. My sense is that one of the core values of GiveDirectly is giving the recipients of charitable efforts a choice around how they use donated resources. Is that right? And if so, why does that seem so important?

Paul Niehaus: That’s right, and there are two reasons. One is instrumental, which is that we often think that actually people who are there on the ground living their own life are going to have more insight and more perspective on how to use money than we as outsiders would. Not always, but often.

So the housing, the metal roofs, I think are a good example of that for me personally. I have a PhD in development economics, so you sort of feel comfortable saying that as far as expertise goes in what to do about poverty, I’m as well trained as anybody. I never would have guessed that so many people wanted to replace their thatch roof with a metal roof.

And when we saw so many people doing that, and looked into it to try to understand why they were doing that, you learn interesting things. Like if you have a thatch roof, you have to replace it or repair it every so often and that costs money; if you have a metal roof, it lasts longer, so you save that money — so it ends up looking like a pretty good long-term investment to build a metal roof. Or you can use a metal roof to collect clean drinking water from the rainwater, and you don’t have to travel a long way to a lake or a river, and you’re less likely to get sick from things that are in the groundwater. Things like that, that’s all stuff that was complete news to me as an outsider, but completely obvious to the people living on the ground. So I think it’s partly in order to be able to tap into that local information.

But I do also think this may vary a little bit, depending on the donor. I personally put a lot of value in people’s ability to make choices per se. I would say that in my description of an ethically good world would be one in which a lot of people have more autonomy and more self-determination than they do now — even if they do sometimes make mistakes, or use it in ways that I disagree with to some extent. I put a lot of intrinsic value in that.

Luisa Rodriguez: Let’s talk about the empirical evidence a bit more. Unconditional cash transfers have been studied empirically many times in a range of contexts, as you’ve noted. Can you summarise what we know about the return on investment recipients get?

Paul Niehaus: I think it’s tricky, because people are going to spend money on such a wide array of things. So if you wanted to provide a full accounting for all of that stuff, you really need to get into all these different categories of saying there’s some benefit of people eating better today, kids being better nourished — some of that might be today benefit; some of that may be long-term benefit. There’s spending on education, you have to think about the long-term returns to that. There’s the durables, like housing. There’s some flow of value that you get from having a better house, but that’s not easy to quantify. There are productive assets — that’s perhaps the one where it’s easiest, where we can say they spent so many dollars on a motorcycle or on livestock, and now they have a business and they earn some more revenue from that.

So there are all these sorts of things which I don’t think anybody’s really made a serious effort to put all of them on an equivalent scale, and say here’s the bottom-line number. That having been said, there are certainly cases you can pick out where a large share of the money got invested in some sort of asset and business got better, and the return on capital in that business was maybe 20% per year, or 30%, or even up to 50%. So there are certainly cases like that, where in a very narrow financial sense, we can say that we’ve learned from this that people have access to high-return investments, and it’s great that we’re able to finance them.

But I would actually push back a little bit about that instinct of trying to kind of put everything into one number — because I think once you get into the reality of how diverse life is, it’s too complicated for that.

Luisa Rodriguez: Yeah, it must be frustrating. It seems like there are all these randomised control trials on a bunch of interventions like this, including unconditional cash transfers. And many of them, in some ways, have it easy. They’re tracking the effect of bed nets on malaria, and it’s pretty easy to measure malaria — at least relative to how difficult it seems to be to measure how people spend money, when there are dozens, hundreds, in some sense an infinite number of potential options for them. And how do you measure the benefit they get from that?

Paul Niehaus: And there are all of these knock-on things, like you see impacts on mental health, or recently there have been papers that found reductions in rates of suicide or rates of all-cause mortality. So do you also think about that? Is that a separate thing that I need to value separately, or is that the result of all these other things that I was just talking about? So I think it’s really hard.

And actually, I think that the way economists have traditionally thought about it, which to me makes more sense, is to say we’re actually going to think of this as like the numéraire, right? The value to giving someone a dollar is a dollar. And then we’re going to use that as a reference point in a comparison to other things — and say, relative to that, how great is a bed net, or deworming, or any of these other things we want to think about?

Luisa Rodriguez: I see. And at least part of the thinking behind GiveDirectly is that, in surprisingly many cases, the value of giving someone something that you’ve decided in advance might be best for them, that costs a dollar, might actually be less than a dollar — because people have such different needs, and it’s hard for us living in other countries to predict them.

Paul Niehaus: That’s the thing we want to watch out for. And the issue there is that in the aid or philanthropic system, there isn’t any built-in feedback loop that prevents us from doing that, right? So think about it: By comparison to a commercial business, if I’m trying to sell something for a dollar, and people value it at less than a dollar, then nobody buys it — and I learn quickly that this isn’t working; I don’t have product-market fit. In the philanthropic world, if it costs you a dollar to produce something, and people value it at less than a dollar, they’re going to say, “Oh, thank you. This is better than nothing.” You don’t get that feedback loop of people telling you that there’s something better that you could have done with your money. So we have to be very intentional about building that in.

Comparing GiveDirectly to USAID programmes [00:15:42]

Luisa Rodriguez: Yeah, OK. I agree. It seems like for most of the history of charitable aid, it doesn’t seem like there have been the feedback loops that allow us to learn things like “When I bought this thing for this person in poverty, the value of it was actually less than the amount I spent, and giving that person the cash would have actually been better.”

But a very cool thing that’s started happening over the last five years is that USAID has started testing its Africa-based programmes to see how they compare to giving cash. Can you explain the overall approach there?

Paul Niehaus: Yeah, and I want to give a lot of credit to USAID as we enter into this topic. USAID has a huge budget to think about. At the same time, they face a lot of constraints — because when Congress gives them money, they give it to them with very specific instructions, like, “Use this for nutrition in Rwanda,” let’s say. So they have to be very thoughtful about being observant of their congressional mandates and those constraints that are placed on them, and they’ve, over time, developed lots of different programmes that they think — based on everything we know about development, about poverty — are going to be the best at achieving those goals, given the budget they have.

What happened a few years ago, which I think is super exciting, is this partnership that emerged between USAID and GiveDirectly. Staff within USAID heard a bit about GiveDirectly on some podcasts like this that we’re doing, and I think heard specifically some questions about, would you be up for comparing this head-to-head with other more traditional approaches, like giving people cows or things like that? And we had said, yeah, of course, that would be great, and we should do it. And USAID staff said that would be really interesting, and maybe we could try that with our programming as well.

So working with them, and with some partners at Google and at Open Philanthropy, we set up an initiative within USAID, where any country office that wants to can opt-in to design a project where they give some people their conventional programming — designed to achieve whatever goals has been set for it — and they give some other people cash transfers that are equivalent, and cost the same amount of money to deliver. And then we just look at the two things side by side, and see which one does more to achieve the goals that USAID have set for themselves.

To emphasise, we’re fixing the goalposts here as being the goals that USAID has set. So if the goal is to increase employment, cash transfers are not an intervention designed to increase employment — they’re an intervention to let people do whatever the heck they want with the money. And so we’d expect that to be a bit of a lower bar, right? Presumably, if your goal is just to increase employment, there’s some way of doing that that’s more effective than cash transfers. But hey, let’s test it and find out.

So USAID has now done a series of these, five studies in four different countries, comparing head-to-head the impacts of their conventional programming to cash transfers. And that’s just totally revolutionary for USAID, because historically, USAID has done very little experimental evaluation at all — let alone this sort of head-to-head testing of two different alternative approaches to try to achieve the same goal, which is very unique. So they just deserve a lot of credit, I think, for being leaders in that, and being willing to put what they’re doing to the test, and learn from that how to do better.

Luisa Rodriguez: Yeah, totally. When I first read about this, I just felt huge admiration for whoever at USAID decided they were open to this. So props to them.

I’m interested in talking about maybe just one of those programmes as an example. One of the programmes tested was a workforce training programme in Rwanda, which is a country that has very serious unemployment problems. Can you talk me through the setup of that experiment?

Paul Niehaus: Yeah, this is the first of these benchmarking studies that GiveDirectly and USAID have partnered on. And I wasn’t a principal investigator (PI) on this project, but I was involved in helping to get it set up, so I can speak a little bit to the design and the results.

The USAID programme that we’re benchmarking here is called Huguka Dukore (HD), and the goal here is around employment and livelihood for young people — not necessarily wage employment; it could also be self-employment, so there are multiple pathways through this intervention that people might be able to choose between. So in that sense, it’s a nice and thoughtful intervention, in that there’s a bit of flexibility and some recognition that not everybody’s the same, and some people should take different paths through life.

What the programme actually involved is a series of different modules that people take over the course of a year. These are things like relevant work readiness training, employability skills training, some on-the-job learning through internships, internship opportunities, some networking links to employment opportunities, and so forth. This sort of reflects USAID’s overall strategy on workforce readiness and skills training, and it also builds on a predecessor programme that USAID had run in Rwanda before. So we really think of it as being like a best-in-class reflection of our current state-of-the-art thinking on how to do this stuff and deal with this issue of youth underemployment — which, as you say, is a huge issue; not just in Rwanda, but elsewhere.

So what the study did is say that some people are going to be assigned to get that, and then some other people are going to be assigned to get cash transfers. And they varied the size of the cash transfers a bit, but the broad goal was to sort of bracket what they thought the cost of this USAID programme was going to be, so that we can say, if we were going to spend about $350 on this programme, or $350 just as a transfer, which of those would do more to achieve these goals that we’ve set?

And in terms of the outcomes they’re focusing on, it’s the stuff you’d imagine, given the goals of the programme: the primary outcomes are mostly things to do with employment and earnings, productive hours of work, that sort of thing.

Luisa Rodriguez: Yeah, cool. So talk me through the results.

Paul Niehaus: I think it’s best to go straight to the authors of the paper for this. They emphasise two points, which I think are right; I agree with their reading of the results.

One is that HD, the USAID-designed programme, did deliver real benefits on some of the things that it was designed to impact. So it did increase productive hours of work. It also increased some secondary outcomes, like productive assets and savings, and it improved subjective wellbeing. So it did do some — not all, but some — of the things that it was designed to do. We can feel confident of that statistically.

But also on every outcome that HD improved, you got bigger — in some cases significantly bigger — impacts from a cost-equivalent cash transfer. It also increased some of the other outcomes that HD did not — like monthly income, or household- and individual-level consumption, and so forth.

So to me, this is exactly an example of the sort of thinking we want to be doing. If we really put our minds to it and design a new programme to improve these things, we probably will have some impact, right? The question is not, “Is there impact or zero impact?” — it’s really more about how much, and how does that compare to this much simpler and much more streamlined approach? And so in this case, it seems like the simpler approach actually got you more of the things that you wanted.

Luisa Rodriguez: Which is especially wild in this case, because like you’ve said, the original programme was designed specifically to target underemployment, and giving cash is not. Though is cash in this case considered monthly income? Could that be responsible for that particular outcome getting bumped up?

Paul Niehaus: No, when we talk about income, it’s income net of any transfers given to people through the study. That’s in fact a not-uncommon misperception of the results from people that want to dismiss them, but that is in fact not the case. That would have been a silly own goal on the part of the researchers, and they did not do that.

Luisa Rodriguez: In any case, it is just particularly mind blowing, I think, that giving cash so unambiguously outperforms this programme targeted specifically at this problem. Do you mind saying which outcomes cash was particularly good on, relative to the programme?

Paul Niehaus: Yeah, and by the way, just two things. One is we’re talking about one specific example. As I said, GiveDirectly has now done a series of these. I think that actually in most of those you’d say that cash looks stronger on the whole, but there are cases where you’re going to get more impact on some outcomes with the conventional programming. So we’re talking about this one example to be concrete, which is great, but let’s also not over-index to it.

The other thing is that I actually think that this is not a shock to a lot of people who’ve actually worked in the aid industry and understand how the sausage gets made. They work hard and they do their best, but I think everybody understands that actually, these things are pretty hard, and there’s a lack of this kind of feedback and iteration that you’d want to have in the system. So when I’ve talked about this with people who’ve been inside for a long time, I don’t think there’s an enormous amount of surprise.

Luisa Rodriguez: They’re not shocked.

Paul Niehaus: But in terms of your question about magnitudes, it’s a little tricky in the paper, in that most of the outcomes are measured as indices or they’re measured as percentage changes. This is actually a big issue in development economics right now, which is outside of our scope for today. But I think it’s a big problem, because we don’t care about percentage changes — we care about changes relative to the cost of the programme.

But to give you one example, the cost-equivalent cash transfer raised monthly income by 99 inverse hyperbolic sine points, which you can think of as being kind of like 99%. The HD programme impacted monthly income by 28 IH points; you think of it as 28 percentage points. So that’s a sort of important metric where the difference is quite big.

Luisa Rodriguez: I guess I’m still inclined to be surprised and impressed, given that this is a best-in-class programme. It’s the best way USAID knows how to improve underemployment, and still giving cash directly is better. What is it about the way the sausage gets made that makes this not super surprising to people working on this on the ground?

Paul Niehaus: I think there are three things to highlight. One is the absence of automatic feedback that I mentioned earlier: I think it’s fundamentally difficult — when you’re doing philanthropic stuff, and you don’t have customers that can tell you this isn’t that great actually — to learn.

Two is all of the complexity that comes with being a big multinational bureaucratic organisation. And that’s obviously not unique to USAID, or to any big organisation, but there’s all sorts of stuff and baggage in terms of decision making that comes with that.

I think the third thing is that there are different categories of problems that it’s helpful to distinguish, and those are what economists would call private good versus public good problems. So me getting a job and earning a living is a very relevant problem to me, and I’m going to do what I can to do that. And I may not do it — I may face constraints that make it hard, or I may make mistakes or not understand certain things, all of that sort of thing — but we should generally have the expectation that people are going to be pretty motivated to try to figure that out on their own, at least to some extent. So that’s the sort of problem where to come in as an outsider and have a really disproportionate impact is going to be relatively hard.

Then there are problems like preventing everybody in my community from getting malaria. I have a motivation to not get sick myself, but I really don’t have a strong motivation, or perhaps strong-enough motivation, to solve everybody else’s problem for them. Or even taking it a step further, doing the innovation, the R&D, to discover a cure for malaria, a way to prevent it at scale. That’s a public good issue, where one person’s actions have much broader ramifications. So that’s a place where you’d expect, coming in as an outsider, maybe we can actually have a really disproportionate impact — because no one person on their own is going to be as motivated to solve the problem.

And so to me, these employment and livelihood-generation problems are private good problems, so I think that’s generally going to be a tough area for us to make outsized progress relative to public goods issues. And I think that’s why, when you look at the things that GiveWell has recommended over the years historically that they think do better than cash transfers, most of them have this public health, infectious disease flavour to them.

Luisa Rodriguez: So in this case of employment, can you help me understand the story for why cash is so helpful? I’m sure people do loads of different things with the cash, as we’ve already discussed. But are there narratives, at least in a few specific cases, that kind of illustrate why this is even plausible?

Paul Niehaus: Sure. And I think mostly the answer, or the honest answer, is no. That’s not what the study does; it doesn’t go into the weeds of what did people do. And there’s certainly no individual case studies or anything. This is a paper, it’s a bunch of regression tables that show you average effects. So that’s just not what it is. You see the impacts on people’s assets and their business assets and productive assets.

So in some sense, yes, mechanically we can say that people who got the transfers use them to start or expand an enterprise, and they’re earning more from that enterprise. And by the way, that’s a very common pattern across a lot of cash transfer studies: You generally don’t see too much impact on how much people are working overall — maybe slight increases, on average — but you do often see a lot of substitution away from working for somebody else and towards working for yourself.

And the reason I think that’s important to understand is that if you’re used to thinking about a rich country labour market, most people work for somebody else. And that somebody else, that corporation, provides you with the tools that you need to do your job. I get my computer from my employer, my office, things like that. In low-income countries, a lot of people work for themselves. And a lot of people would work for themselves if they were able to provide the tools they needed to do their job, which might be the motorcycle or the livestock or the sewing machine or whatever it is. But for them, in an environment where there aren’t a lot of other firms out there offering steady jobs, the big constraint is just being able to get the tools you need to work on your own account. So I think that’s a very common pattern in a lot of these cash transfer studies.

Luisa Rodriguez: Is it at all worrying that in rich countries, people are working for firms, and this is making it so that more people are working for themselves and not getting the benefits of working for firms? Or do you see this as part of a trajectory that ends up with people in these countries slowly getting wealthier? It being more possible to create firms that actually work well and employ people and give them those benefits?

Paul Niehaus: I mean, the rich country part of that question is complex, and it leads us down this path of what do we think about the gig economy and all that kind of stuff. But I think if we just think locally, for people living in or near extreme poverty today, a lot of them would love to get a steady job. That’s very clear if you ask people. A lot of people, if that were an option, would go for it. It isn’t. And so what they have is the opportunity to work on their own account if they can accumulate the tools and the capital that they need to do that.

But I think that is absolutely a big part of the development pathway, is enough people creating businesses that are successful at a meaningful enough scale to be able to generate employment for other people in their community. And that is a big challenge for us when we think about it, as development economists, because those are sort of rare events, right? So it’s relatively easy to say, here’s the average impact on a bunch of families of doing something for them. But if what you really care about is the one in 1,000 people that is unleashed to then create a business that employs a bunch of others, that’s so infrequent that the small change in the probability of that happening we’re often not able to detect in studies of the size that we’re doing.

So we can reason about that by saying, if here are the average effects we see, then presumably that is going to be helpful for whoever is the next Bill Gates or whoever it is. But it’s hard to see that directly in the data, and that is a fundamental challenge.

Luisa Rodriguez: Zooming out again to this general approach, my understanding is that basically all of the programmes that USAID has compared to cash have ended up looking like cash outperforms those programmes on most metrics. First off, is that right? Did I miss any important studies showing that the USAID programme was a big improvement on cash?

Paul Niehaus: I think that’s right. It’s obviously somewhat subjective. Different people might read things in different ways. But I think, generally speaking, that’s the takeaway that most people have.

Luisa Rodriguez: I’m curious how USAID is reacting to that. In an ideal world, they might use these results to pivot to cash transfers, or maybe to figure out how to improve their existing programmes to get better results. But part of me wonders if that’s kind of unrealistic, because at least when I imagine being a USAID employee, I would just find that extremely demoralising, and would kind of not be interested in comparing my programmes anymore, at least on some gut level. Maybe I would muster the strength to ignore that impulse. But has it changed the way they’re spending money? Are they going to stop implementing some of the programmes that look less effective than cash, and give cash instead?

Paul Niehaus: Well, first of all, I don’t think anybody should feel demoralised, because this is just fantastically valuable scientific knowledge. We’ve done the thing that’s ethically right, that’s scientifically valuable, and that positions us to do more good, and that’s great. So I think everybody should feel very proud of that. Then also, just very pragmatically, you can look at this and say, “I want to think about how to improve existing programmes to try to catch up to that cash benchmark” — and that’s an interesting and important problem. Or you could say, “I want to pivot to cash transfers,” and I could tell you that also the design and implementation of cash transfer programming is a fun and difficult and challenging and rewarding thing to do. So there’s still plenty of good work for everybody to do, regardless of how exactly you read the results.

But in terms of how things are moving forward, USAID is obviously a huge and complex organisation. So there’s no one answer to what USAID is doing. But GiveDirectly has done a bunch of work since that first project with USAID that I’m very excited about — 13 additional projects across the world in different countries, including three additional benchmarking projects — so I think it’s been and seen as a fruitful collaboration that’s going to continue.

One thing in particular that I thought was really neat is that USAID just put out an RFP in Kenya asking for proposals to address food security issues, and saying very explicitly that there was going to be a benchmarking component included in the project in order to test effectiveness, that we actually had nothing to do with. So we were surprised; we didn’t know of this until it came out. To me, that’s a good indication that people are saying this is useful, and let’s weave it into the way we normally do our work.

Luisa Rodriguez: Yeah, I feel very inspired if the thing that ends up happening is there’s a cultural shift at USAID, and all of a sudden they’re testing most of their programmes against cash, and hopefully something like if cash performs better, going for cash to achieve their goals. I’m still unclear on whether there have been any cases — and maybe it’s just too hard to say — where USAID has scaled a programme down and started implementing cash transfers instead?

Paul Niehaus: Yeah, I don’t have numbers in terms of how much is spent on these different programme types to be able to tell you that. It’s a great question. In fact, that’s something I’ve generally found difficult as I’ve studied foreign aid: there’s very little reporting in these kinds of categories that would be useful for understanding. Are we learning from the data? Things like that. That’s not an issue that’s unique to USAID. That’s true across the board.

Luisa Rodriguez: If it were the case that these kinds of aid agencies benchmarked a bunch of their programmes against cash, and cash in many cases outperformed those programmes, should these agencies basically just massively shrink and spend most of their budgets on unconditional cash transfers?

Paul Niehaus: Well, I think you should spend a pretty sizable chunk of your budget on unconditional cash transfers. What that means for headcount, I don’t know, because you also need people to design those programmes and to make sure that they’re well implemented, and that’s not a trivial problem. GiveDirectly has 800 or so people globally who work on that on a daily basis, so that’s important too. But I think that, yes, you would shift a substantial share of the budget to cash transfers, and then you’d have a part of the budget that’s allocated to how can we find things that do better than cash transfers for some of these specific things — where we think there is a strong reason for us to come in and push people and push an economy in a direction that’s a bit different from where they would go on their own.

GiveDirectly vs GiveWell’s top-recommended charities [00:35:16]

Luisa Rodriguez: Trying to find the best arguments against cash transfers being this kind of Holy Grail or benchmark, GiveWell, the charity evaluator, thinks that GiveDirectly does less good per dollar than its top recommended charities. What’s the case that people should give to GiveDirectly anyways?

Paul Niehaus: Great, yeah. And on the positive side, GiveWell also thinks that cash transfers are better than almost all things that they’ve looked at, but that’s right. I think it’s been great and very encouraging for me to see them almost set their goal as, “Let’s try to find every year the few things that we think can do better than cash transfers.” That’s very much the role that we want to be playing in the world, sort of pushing people to do that.

So if you get into the spreadsheets and the cost-effectiveness models and so forth, I think there are probably some things that I would do differently that might lead to a more favourable number for GiveDirectly. GiveWell hasn’t yet taken into account some of the indirect impacts of transfers, not just on the people who get them, but on other people in the community — which I think, based on some of the work we’ve done, can be quite large. I think there can be a significant economic stimulus effect, so that might make a degree of difference.

As I said before, I would personally, as a donor, put a significant weight on autonomy and individual freedom per se, even setting aside whatever economic value I place on the way people use that autonomy. So that would be a big one.

But I think the core thing is really that GiveWell’s question of, “If I had a modest amount of money to allocate, where could I get the highest returns?” is somewhat different from the question we’re asking, which is, “How can we improve the quality of the average aid dollar, looking at these huge budgets that are spent globally?” So a big part of the rationale for giving through GiveDirectly is yes, you get the direct impacts today on the person who received them — but we’re also part of this movement to change the sector as a whole, and to raise the bar for everything else. That’s what the USAID benchmarking projects are about, for example.

Of course it’s very difficult to attribute, but I think the conversation about cash transfers, the acceptance of them, the idea that we should think of them as a first option, has really changed in the last decade. And GiveDirectly has certainly played some role in that, although not the exclusive role.

Luisa Rodriguez: Yeah, I buy that. We’ll actually come back to spillover effects in a little bit, but before we do, I’ve also seen GiveDirectly proponents argue that GiveDirectly can absorb more funding than other top GiveWell charities, because cash is more scalable than other antipoverty interventions. Can you explain why that is?

Paul Niehaus: There are two dimensions to this: there’s scalability of implementation, and then there is scalability of relevance or of impact, if you will. I want to talk through both of those things in turn, because they’re both important, both very different.

Implementation is just “Can we do it?” For cash transfers, we know for sure the answer is yes. We can do this at enormous scale, because there are already a couple of billion people in low-income countries around the world that have received cash transfers of some sort, typically from a government programme. GiveDirectly is seen as very brazen and unorthodox and novel by Western donors, but actually, in low-income countries themselves, cash transfers are very common. It’s usually a core part of most governments’ antipoverty strategy, and they reach a lot of people. And they scaled very quickly as well during the pandemic, for example, when most countries put in place restrictions on mobility, and at the same time said we really have to get support to people quickly, so they turned in a big way to cash transfers.

There are some really cool examples of that as well that sort of illustrate scalability at speed, like the project in Togo, which the government did in conjunction with GiveDirectly, where they did everything via mobile phone without ever having to put boots on the ground to interact with people. So you have some of these really neat examples. So the first point is we could get money to essentially everybody living in extreme poverty, and that’s not necessarily true of some of the other interventions that would be harder work — slower, I think — to reach that kind of scale.

The second point is about scalability of relevance or of impact. A very concrete way to think about this is you think about some very high-impact things, like deworming medication: that’s going to have a huge impact in areas where baseline worm loads are high, so there are a lot of people that are getting sick, and it’s not going to do anything in places where worms are not a problem, where baseline worm loads are low. So it’s a great thing to use in the places where we can get these outsized returns, but it’s not a general solution to the problem of getting people up out of poverty. It’s a narrow, very-high-return thing. We should absolutely exploit it.

But if we want to think about the big problem of how do we get everybody out of extreme poverty, then I think cash transfers really are the only thing that you can expect to be relevant at that sort of scale, in the sense that we could expect to continue to see the kinds of positive impacts that you’ve seen across the whole swath of people that you want to try to reach.

Luisa Rodriguez: This is a bit of a digression, but that makes me curious about the extent to which cash transfers vary in different regions in their effectiveness. Do we see that across studies?

Paul Niehaus: I’ve been looking into this, in fact, in the last couple of weeks. I don’t think that there are well-designed, well-powered comparisons yet. I’m sure the answer is true, intuitively, just in the sense that there’s so much regional variation in what people do with their money on average, in what the opportunities are and what people want. But I don’t think that’s actually been tested. But it depends how small the categories are, right? So if I look at how many people buy a metal roof, there will be region-to-region heterogeneity in that. If I look at the broader question of to what extent do people invest in housing or in durable goods or things like that, at that level, I think we’ll expect to see more homogeneity.

Luisa Rodriguez: Getting back to potential counterarguments, my sense is that GiveWell’s view is basically that in deciding whether to donate to GiveDirectly now or wait and see what comes along, in some cases it’s better to wait, because they think they’ll be able to identify more promising things. What’s your take on that position?

Paul Niehaus: Broadly, go for it. That’s great. We want GiveWell to go out there and try to find the best things they can. So if they view finding things that beat cash as a good way to think about the problem, that’s what we want them to do. So I think that’s great. We’ve always seen our goal as providing a challenge function or a hurdle.

I do think, though, there is a sort of question of breadth. So if a small group of EAs who read GiveWell posts in detail decide they want to wait, that might or might not be a good strategy. If the entire world decided to just wait, and the couple hundred billion dollars a year that we spend every year on development were to just sit in a bank account and do nothing, I think that would be a tragedy. So we want to be careful about the scope of the advice, if that makes sense.

Luisa Rodriguez: Yeah, that makes tonnes of sense. It reminds me of what my colleague calls “the rule of equal and opposite advice” — where some group needs one piece of advice, and another needs the opposite. In this case, maybe there are some people who will think about this so carefully that they’ll find opportunities that beat cash, but if we’re talking to a large enough audience, a lot of them should probably be trying to fill this cash gap.

Paul Niehaus: That’s very well said. Yeah.

Cash might be able to drive economic growth [00:41:59]

Luisa Rodriguez: Let’s talk about this possibility of big positive spillovers that you mentioned. One question some people have about GiveDirectly is around whether it’s kind of a band-aid for people living in poverty — in that it helps them meet their basic needs for a few months or even years — or whether it can actually lift people and communities out of poverty indefinitely by jumpstarting local economies.

If I understand correctly, you helped run this study looking at this question by looking at what’s called the “general equilibrium effects” of cash transfers. To start, can you explain what general equilibrium effects are, and why they’re so important to answering the broader question about the value of GiveDirectly’s work?

Paul Niehaus: Yeah, but let me first just say a word about the framing of the question, because I think this notion of “lifting people up out of poverty” is very pervasive and very alluring, right? That’s obviously something we would all love to be a part of and help to do. But I want to make sure that we’re all starting from the facts — which are that, by and large, people have been getting out of extreme poverty at a pretty rapid rate on their own. There’s no doubt that rates of extreme poverty have fallen dramatically. COVID was a significant setback for sure, but setting COVID aside, the global poverty gap — the amount by which people living in poverty are below the line — has shrunk dramatically over the years.

So I think it’s better to frame the question for us as, “What can we do to accelerate that?” — as opposed to, “What can we do to lift people up out of extreme poverty?” — because the latter sort of leads to this mindset of they’re just going to stay stuck there forever unless we figure out the solution. And that’s just clearly counterfactual. That’s just not the case.

So having said that, what do we mean by “general equilibrium”? What economists mean by general equilibrium is that one person’s actions have broader ripple effects and ramifications on other people through the economy, because we live in this interconnected system. And if we do a study where we just look at what happens when one household gets money and another doesn’t, and we compare the two, we might see differences between those two households in terms of what stuff they have or how they behave or spend their time, but we aren’t picking up those broader impacts: What does that all mean for the rest of society?

With cash transfers, the most obvious example to think about, to make this very concrete, is that whenever people get cash transfers they use them to transact. So there’s somebody else that they go out to and they buy food, or they buy an asset, or maybe they even put money in a savings account — that actually rarely happens — but whatever it is that they do, there’s some counterparty to that transaction, so it’s almost mechanical that that counterparty is going to be affected at least to some extent by it. And that means if we want to think about the total consequence of the transfers, we want to take that into account.

That could also be true for other things as well. There are lots of things that we do to help people that could have these kinds of broader impacts. But I think for cash transfers it’s especially mechanical or obvious that that has to be the case, because people are going to go out and use the money.

Luisa Rodriguez: Right. The effect is clearly not just the effect on these recipients, because unless they’re all just saving the money and not spending it, that cash ends up getting transferred a bunch of times from one person to another as they make purchases. And so if you just measure the benefits to the recipient, you miss out on the benefits to the person who makes metal roofs who just got a bunch more business, and now their life has improved as well. So this study is basically not looking at individual recipients alone and is trying to see how is this community or how is this region affected if we zoom out? Am I basically getting that right?

Paul Niehaus: That’s mostly right, yeah. The only thing to adjust is that even the saving, depending on how you save, is going to affect. So if you put money into a bank account, that affects somebody else, because that money becomes available through our system of financial intermediation. It becomes capital for somebody else to borrow and invest. The one case in which you might save and it literally does nothing is if you take the money and stuff it under a mattress, which generally doesn’t seem to be what happens. But unless you do that, it’s going to have some impact somehow on somebody else.

Luisa Rodriguez: This sounds incredibly difficult to study. How do you do it? What’s the study setup?

Paul Niehaus: We do it imperfectly, but I think it’s still a significant advance on what we knew or were able to do beforehand. So this is actually a general issue in economic analysis of impacts. And we’ve made enormous progress in the last 20 years — and the Nobel Prize in 2019 for the leaders of that movement is very well deserved — but one of the things that’s very hard is that to see these system-level or economy-level impacts, it’s not entirely clear how to do it.

What we do in this study is we do an experiment, but we do an experiment at a larger scale, where the units that are getting allocated or assigned to treatment and to control are bigger. So instead of picking some households and saying these households are going to get money, and these aren’t, we’re picking some villages and saying, everybody eligible in these villages will get money, and those eligible in these other villages will not. And we’re even doing it at a slightly larger scale, because in the part of Kenya where we work, we take sublocations, which are administrative units, and say in some of these sublocations, we’re going to treat two-thirds of the villages, and in others we’re going to treat one-third of the village.

So all of this generates a lot of more aggregate variation in how many of the people around you have been treated, as well as whether or not you got treated yourself with cash transfers. The essential idea in the study is to use that spatial variation to learn something about these indirect effects.

What we do concretely, which is actually very simple, is we take each person, we draw a two-kilometre ring around them, and then we say, let’s look at your outcomes — as a function both of whether you yourself got a transfer, and also of the share of people within that two-kilometre ring that were eligible that got a transfer. And that’s it.

Luisa Rodriguez: Great. That is simple. Then you give the cash based on who the study determines gets the cash, and you wait some period of time — I imagine it’s some number of months or years?

Paul Niehaus: There’s actually a range. And that’s important, because what you’d really like to understand is not a single snapshot of what did life look like 18 months later, but how did life develop? Because we care about all of those 18 months, and we care about the next 18 months and all that. We make a little bit of progress on that in this study, in that the timing of the surveys was varied a fair bit, so we have some surveys that were run fairly shortly after the transfers went out, some that were much longer, and so we can sort of integrate across that whole time period.

And just to be very open, we should have done more of that. In fact, as we sat back and analysed the data, we were like, boy, if we had it to do over, we would do even more of that. But we have enough to be able to say, let’s integrate over this whole time period after transfers go out, and say, on average, over that time period, what sorts of impacts are we seeing on the economy in terms of people’s incomes, their standards of living, their spending, and so forth?

Luisa Rodriguez: OK, so you get this kind of full, more detailed picture of what is happening over the years after this happens.

So the headline result is that for every one dollar spent on cash transfers, there’s a 2.5 multiplier effect. Can you explain exactly what a “multiplier” means here?

Paul Niehaus: So it’s a very simple concept. All it means is if we measure GDP — essentially the aggregate output of this economy — how much did GDP go up for every dollar that we gave people? So a 1 would be people spent the money and nothing else happened. A 2.5 here means that for every dollar that we put in, economic output in the region expanded by two and a half dollars.

Luisa Rodriguez: OK. And again, I studied a bit of economics, but not loads of economics. So if I try to make it really concrete, it’s like you give someone $1, and the fact that they then spend that dollar ends up somehow generating $2.50 in the economy because of something like it enables someone to do a bit more work, which then allows them to create more goods, and those goods have value and are purchased? And over time you get basically $2.50 of extra value? Is that the basic thing?

Paul Niehaus: That’s the basic thing, yeah.

Luisa Rodriguez: OK, great. I understand what a multiplier is. Can you help me make intuitive the number 2.5? Is that really big? Is that moderate? I guess I just don’t have a good reference class for how quickly economies grow, and whether this is an impressive result or not.

Paul Niehaus: Yeah, I think there are probably two ways to think about that. One is relative to other estimates of the multiplier effects, public spending or stimulus spending. In rich countries, like in the US, for example, the range of estimates that people typically are centred around are in the 1.5–1.9 or 2 range. So this is bigger than that, but not wildly bigger than the stimulus effects that we think federal spending in the US can have if it is well designed.

The other way to think about it is just as a donor. If you’re thinking about the consequences of this, then, if you buy the analysis, you think that in this setting a dollar had 2.5 the impacts that you thought it would have had otherwise, without taking that into account.

Luisa Rodriguez: Yeah. OK, that makes sense.

Paul Niehaus: One thing to emphasise is that we’re talking here about economic output, and that’s not the same thing as people’s wellbeing, so we want to be very careful about that. In the paper, we talk about some very basic points, like if output went up but it was because a bunch of people were working longer hours and they had less time to spend with their families, there’s some tradeoffs there that we’d want to take into account. In this case, it seems like that’s actually not the case, as best we can tell. But we do want to emphasise that point that you shouldn’t think of a 2.5 GDP multiplier as being like a 2.5 wellbeing multiplier without some more careful scrutiny.

Luisa Rodriguez: Totally. So do you then have as an outcome wellbeing? Not just among recipients but somehow within other community members?

Paul Niehaus: I know, I heard the WELLBYs pitch recently, so I don’t know, maybe we should have done that. But no, we didn’t. We think about wellbeing as economists typically do, which is in terms of the equivalent income. Think about how the world has changed: What would be the equivalent change in just my income that to me would be just as good as that set of changes in the world? So we do that exercise in the paper. And the bottom line there is that, because we don’t see a lot of changes in these other things, like how many hours people are working, we do think that 2.5 is actually a pretty good figure for thinking about welfare and not just GDP. But I still want to emphasise the point that they’re not the same.

Luisa Rodriguez: Sure. So if you saw something like leisure time dropping dramatically. And it sounds like you didn’t measure something like stress, but that’s the kind of thing that you’d be worried about. And to the best that you’ve measured so far, it doesn’t seem like this thing is happening, where people’s lives are getting more miserable because they’re overworked. Instead, those metrics aren’t really changing, and it just seems like GDP is going up anyways.

Paul Niehaus: Right. And I like that topic, actually, because I think many people have sort of the opposite instinct. They’re like, “We want to get people living in poverty to work more” — which I think is a little weird, and I feel very uncomfortable about, but I think a lot of people feel that way, right? So it’s an interesting one.

Luisa Rodriguez: Yeah. So to make sure I understand what this really looks like in practice, do you have a sense of who is benefiting besides the recipients? Is it business owners in the communities? Other communities, other poor households, wealthy households?

Paul Niehaus: We do a bit, but not a lot. What I mean by that is the way the study was designed, GiveDirectly had decided that some households in these villages would be eligible for transfers based on indicators of their poverty status, and others would not. So we measure data for both those groups, and we can look at impacts on both of those groups. And in fact, the impacts look pretty similar on a lot of the outcomes across the two of them. That’s one of the big surprises, and maybe an indicator that these indirect effects can be very big, that we see increases in standards of living for the ineligible guys that did not get money that are of a similar size as the impact on the eligible guys.

I had that question specifically — about “Is this business owners?” — and we actually have a second paper that looks in much more detail at the targeting of the transfers. As part of that, we asked specifically, can we say anything predictably about for whom the impacts are greatest? And among those non-recipients, we really can’t, is the answer. Some of that may be because we just don’t have enough data. By the standards of an RCT, we have enormous amounts of data; by the standards of machine learning we have a trivial amount of data, and so we may just need more.

But there isn’t a big obvious pattern that jumps out, where it’s like, it’s the business owners who are seeing increased revenue. I do think that’s part of the channel, but then I think there are various ways in which that then gets passed along. The business owners who employ people pay higher wages, for example. We see that. I think because there are enough of these ways in which things percolate, it ends up being spread out pretty evenly.

Luisa Rodriguez: Cool, yeah. That is interesting. I also find it surprising that the benefits are kind of equal. One concern I’ve heard raised about GiveDirectly’s model is that there’s this worry that so much of wellbeing and welfare comes from comparing yourself to others. I’ve heard this concern that if some group of non-recipients starts noticing that there’s another group who’s gotten cash and they’re able to raise their standard of living, that first group that didn’t get anything will be, in terms of wellbeing, worse off. Because they’re no longer like, “I’ve got it pretty good” — now they’re like, “The world’s not fair. I didn’t get that transfer. And also I’m not doing relatively as well as I used to be in this community.”

Do you look for evidence of that? Do you have intuitions about it?

Paul Niehaus: Yeah, I think this is an important question. In this case, we look at impacts on the measures of subjective wellbeing of the ineligibles, and we’re not seeing negative effects there. So in that sense, and I think that probably that’s partly because economically they’re also benefiting to a great extent here.

I think this question is not really so much a question about cash transfers per se, and more a question about how we allocate whatever it is that we’re giving out, whether it’s cash transfers or food or livestock or anything: that we’re thoughtful about doing it in a way that communities perceive as being reasonable and well justified. There are good examples of studies that find that programmes that were not targeted in a thoughtful way, in a way that was seen as fair, gave rise to tension and conflict and jealousy and bitterness in communities. So absolutely that can happen.

And I think there’s also an issue in the design of experiments. If you design it in such a way that two people who live next to each other, very saliently, one of them got lucky and the other one didn’t, that’s not a great experience. And that’s one of the reasons why we’ve been pushing towards, at GiveDirectly, doing these experiments at a larger scale — so it’s not like two neighbours, but like different villages. So there are some important and difficult tradeoffs to think through there in terms of the ethics of designing experiments, which is maybe a longer conversation, but those are the issues I’d be thinking about — whether it’s cash transfers or anything else.

Luisa Rodriguez: OK, cool. Yeah, I really like this idea of randomising at the village level and not the individual level. Because it does make sense to me that you’ll get less of this problem if all of the eligible people in a given area get the benefit. Could it be the thing that’s happening is that the additional cash, people spend it, but it actually ends up driving prices up for everyone, such that people aren’t actually better off in a meaningful sense?

Paul Niehaus: Theoretically that’s entirely possible. You can imagine an extreme case where if I went and gave a bunch of money to a Robinson Crusoe little island, where there are just a couple of people, that’s going to do nothing to increase their productive capacity and so it doesn’t seem likely to do anything other than drive up prices. So that was one of the core goals of the study, to measure that. We see [statistically] significant increases on prices that are very very small. So sort of a couple of fractions of a percentage point relative to transfers that amounted to about 15% of GDP. So very small relative to the amount of money flowing into these economies, which was big. We have some stuff in the paper about why that is.

Luisa Rodriguez: Yeah, I’d love it if you could talk me through what you found there.

Paul Niehaus: So I think it’s partly because we’re working in a region that is pretty well integrated with the outside world: it’s on the main road that runs from the coast through Nairobi, inland towards Uganda, so connectivity is pretty good. But the other big factor is that we think that this is a setting in which there’s also just a lot of underutilised capacity. The prototypical thing here would be you have a little business that serves people in your community, and these guys see an average of two and a half customers a day. So that could be like a retail business, or maybe you have a grain mill where you grind people’s grain, or whatever it is.

So what that means is that you could actually do a lot more if you just had more customers, if there are more people that wanted to buy your thing — because you can certainly serve more than two and a half customers a day. I think that’s also an important factor. And that’s economically very interesting and exciting, because that issue of underutilised capacity is one that’s been hard to study, but that, especially in these lower-density rural areas, I think is actually quite important.

There have been a few other studies that have found larger price effects in very remote parts of other countries. In remote parts of Mexico, for example, you can get larger impacts on prices, so it can happen. But my sense is that from what we’ve seen as a whole, in most cases there hasn’t been too much of that price pressure.

Luisa Rodriguez: Cool. That reminds me that I was curious about another finding of the paper, which is that you somehow get this 2.5 multiplier effect, but there aren’t corresponding increases in the employment of land or capital or labour, at least as measured. How do you explain that? Is it primarily the slack capacity thing, where people have capacity that they’re not using because these economies are small? And people aren’t buying every tin roof that someone could plausibly make, because they don’t have the funds, and these are small communities? Is it that? Is it something else? Is it many things?

Paul Niehaus: Yeah, that’s right. That’s our core interpretation of the data. There are some increases in those things. There is maybe a little bit of an increase in labour hours, although it’s not significantly different from zero. There is a significant reallocation of labour hours away from wage employment, towards self-employment, as I mentioned before. And so if people are more productive in one of those things than the other, there’s going to be some increase that comes from that. If you look at some categories of capital investment, there are increases. You don’t see a lot of fixed capital investment, like people building new structures. Businesses have larger inventories, but those increases in inventories are not big relative to the increases in output that we’re seeing.

So it seems like there’s probably some of this that is just like more utilisation of labour and of capital. But a lot of it, I think, is actually more utilisation of the stuff that was already employed, but just underutilised.

Luisa Rodriguez: So the multiplier effect is 2.5x the $1. Intuitively, I have this reaction that’s like, if a lot of how you explain that isn’t that people are able to farm more land; it’s more of this thing where people weren’t working as much as they could with the resources they had before, and now they’re utilising those resources better or something because previously there was not enough demand for them to fully utilise the resources they already had.

Intuitively, I find it kind of surprising that just that slack capacity being used up is a big enough deal to create this 2.5 multiplier economic effect. But maybe slack capacity, the insufficient use of people’s time and inventories that they already have, is just such a big deal in these economies that it is intuitive to you that it could drive an effect this big?

Paul Niehaus: It’s a good question. As a point of reference, it might be helpful to start with the US macro literature on this. So there are a lot of people that try to look at when the US government spends money, how much does that stimulate the economy? There’s a range of estimates, but I think the pretty good sense of the variety that most people believe right now would be in the range of 1.5–2x for various forms of government spending. And there’s some differences. Government spends money in the US differently, a bit different from a cash transfer. But with that as a reference point, getting 2.5 in a much more rural, low-density area, where I think there probably is a lot more underutilised capacity, doesn’t feel at all super surprising.

Luisa Rodriguez: Does that maybe mean the benefits of this approach might be somewhat limited to contexts where you’ve got smaller rural communities that end up with this particular issue, and that you wouldn’t get the same benefit in a super urban area where there’s a big enough population that there’s enough demand to use up slack capacity?

Paul Niehaus: That would certainly be my intuition. I think that would be a very useful thing to test. That said, if you think that maybe the US economy is one end of a spectrum and rural Kenya is at another end of the spectrum, then maybe urban areas in Kenya, or other low-income countries, perhaps they lie somewhere in between — I think that would not be a crazy prior to have as well. So that would be my best guess if I had to guess.

I think it’s going to be very hard to test. It’s hard to imagine doing a study like the one we did, where you take the spatial approach, and then taking that to a city where I think space is much less of a constraint, things are much more dense spatially. So I don’t know if we’re going to have a good answer to that question anytime soon, but if I had to guess, I would bound it with those two data points.

Luisa Rodriguez: So how long do you expect this benefit to last? Do you expect it to peter out, or could it drive kind of indefinite growth?

Paul Niehaus: I thought that the estimated effects would fade out over time — not so much because I think the economic impacts are going to fade out, but because I think that the treatment effects are going to keep spreading out through the economy. And so over time, the control group is going to get more and more treated, and eventually there won’t really be a meaningful distinction anymore between the two.

Luisa Rodriguez: It’s like people will keep buying other things, services, goods — and at some point people will buy other things from the control villages, and then those control villages will start getting the benefits that originally the treatment villages were meant to be getting. And that’ll just happen more and more, because economies are complicated, and over time, as that happens, everyone will kind of equalise out.

Paul Niehaus: That’s exactly it. And we are picking that up to some extent, because we are looking at people in those controlled villages and saying, “Was there an impact on this person as a function of how many of their neighbours were treated?” But the ability of that design statistically to pick this up, I expect it to sort of fade out over time.

Things that are pushing against that are that there may be important forms of increasing returns here. If one village gets off to a good start, that attracts in and crowds in other people and other capital. So it really puts that village or that community on a permanently or long-term somewhat different trajectory. So that is also conceptually a possibility. I’m not a PI on this part of the study, but the team that’s doing followup work on this, I think they’re going to have results out fairly soon. Verbally what they are telling me is that they’re seeing some really interesting results and continued differences, which is a bit different from what I would have predicted for the reasons we described. But we have to wait and let them tell that story when they’re ready.

Luisa Rodriguez: Fair enough. Can you tell the difference between something like the effect being kind of diffused across treatment and control groups, and the effect just disappearing because whatever good thing that’s happening dissipates and you actually just lose it?

Paul Niehaus: I don’t think so. I know people would love for there to be an answer to that problem. I think it’s a fundamental issue. The essence of the experimental idea is for there to be somebody who’s unaffected that we can use as a counterfactual, and say, “That’s what life would have been like for Paul if this treatment hadn’t happened.” And if you believe that the world is very interconnected, and that over time things ripple out and affect everybody, then that’s just not true.

Luisa Rodriguez: I understand the intuitive story for how these effects keep spreading and it gets harder to measure, but we think they’re still there. What’s the intuitive story for how they just disappear? Economically disappear. Disappear in a meaningful sense, not just your ability to measure them stops.

Paul Niehaus: I think it would be, first, to some extent, people have spent some of the money on things that improve their life today, but that don’t have long-term impacts. And I think that’s fine; I support them doing that, but that would be one part. The second is that the things they have invested in, the impacts of those things fade out over time. The assets they’ve purchased deteriorate, and the additional income they’ve made from them or the additional benefits they’ve earned from them, for whatever reason, they’ve chosen not to reinvest in maintaining them or replacing them or things like that. It’d be things like that.

Luisa Rodriguez: So overall, this just seems like a really exciting study. It suggests that you might actually be able to help very unproductive, isolated economies, where lots of people are living in poverty, kind of jumpstart economic growth through just giving people cash. I have the intuition that a lot of people will find this sounds too good to be true. Do you buy it on a very gut level?

Paul Niehaus: Yeah. I don’t think it’s out of line with, as we said before, studies that have been done in other contexts — although it’s the first time we’ve been able to do this experimentally, and that is a big advance and increases my confidence in the results. I think it’s also in line with general thinking about a lot of other stuff that happens during the process of growth. The role of demand of purchasing power in stimulating parts of the economy is a pretty classic idea. It’s not something that we’ve been able to study experimentally with RCTs, but it’s a core part of the way we think about the process of development, that in some cases that may be important.

Luisa Rodriguez: Cool. And if you had to give a best guess for why, if you replicate the study in another context and you don’t see a multiplier nearly as big, what’s your best guess for why that might be?

Paul Niehaus: I think we could totally find a place where in fact capacity was very highly utilised, and so you get a bit more than a dollar per dollar benefit, but not a lot more. I think that would be totally plausible.

Luisa Rodriguez: OK, so we have some reason to think that cash transfers can lead to some GDP growth, and that it’s nontrivial, but if you’re trying to do as much as possible to increase economic growth in a poor country per dollar spent, do you think cash transfers would end up being among the most cost-effective ways to do that?

Paul Niehaus: I think they’d be a big part of the mix. There are certainly things that you need from the public sector. You need investment in infrastructure — roads and things like that — for which you need public intervention. And we’ve actually had some people who’ve received GiveDirectly transfers who have pooled resources to build a road. So there’s some local capacity, I think, for communities to actually create public goods like that on a small scale. Nobody’s saying there should be none of that. But I think there’s also an enormous amount of private investment that’s needed — and the evidence says cash transfers can be a very good way both to finance that investment and also to motivate it, by creating demand where there wasn’t a lot of demand before.

Luisa Rodriguez: Sure. It seems like a lot of the stories that we have about low-income countries developing and becoming higher-income countries has to do with shifting out of subsistence agriculture and into higher-productivity sectors like manufacturing and services — often because people have migrated to more urban areas, or at least into larger-scale, more professional and productive agriculture. Do we have any evidence that cash transfers contribute to that kind of urbanisation process?

Paul Niehaus: Certainly to the structural change that you mentioned. So if you look at the results from the GE [general equilibrium] study that we’ve been talking about, for example, the economy to begin with is going to be majority agricultural. The expansion that we see in response to the transfer is entirely concentrated in retail and small-scale manufacturing. So it’s exactly that pattern that you see — which, again, I think is partly investment-led, but largely demand-led in this setting.

So we haven’t been seeing a lot of people moving to the cities, which was something that I thought could happen. If anything — and I think we’ll see more on this as some of the longer-term followups from these studies come out — you’re seeing people moving into these communities from other communities, because they’re becoming larger hubs of economic activity. So I think there is a bit of that spatial agglomeration that’s part of the process, but maybe not quite in the way that we expected. We’re getting densification, but in a different place than perhaps we expected.

Luisa Rodriguez: OK, that makes sense.

Fraud and theft of GiveDirectly funds [01:09:48]

Luisa Rodriguez: Let’s push on to another topic. GiveDirectly has written publicly a few times about fraud and theft of GiveDirectly funds. In 2021, there was a blog post about fraud and theft across programmes globally, and I think in total that added up to something like $250,000 for the year. But then also, just this year, there is a single case of major theft of just under $1 million in the Democratic Republic of the Congo.

So these are two pretty different cases, where the $250,000 figure reflects a bunch of very small, isolated incidents of fraud and theft across different programmes, and then this DRC case is this big single theft case. I want to talk about both of them, but let’s start with the smaller, more common cases first. Can you start by talking me through just a typical case of fraud or theft?

Paul Niehaus: Sure. There’s a variety. Fraud and theft are a normal part of business in what we do, and something that we have to be always dealing with, always thinking about, always staying a step ahead of. So there’s a variety of different things. Which, by the way, is pretty similar to other businesses I’ve started in payments where I’ve worked: to lose a percent or so to chargebacks on debit cards that people use that turn out to be fraudulent would be pretty normal; that’s a cost of doing business. So we have to think about it as a cost of doing business: we have to control it, keep it low, and think about tradeoffs.

The sorts of things that we would typically worry about might be somebody going into a community, working for us to enrol recipients, and enrolling friends and family members instead of the kinds of people that they’re looking for. Or somebody going into a community, enrolling the people that they’re supposed to enrol, but also asking them to then kick back some of the money to them — and maybe having a story around that, in terms of, “You’re getting this because of me” or “You’re not going to get it if you don’t agree to share the benefits with me.” Those would be pretty typical day-to-day risks that we worry about and police.

Luisa Rodriguez: OK, and just to get a sense of the magnitude of how common that is, can you put the $250,000 figure in context? What percent of the overall funds GiveDirectly intends to deliver are lost through this theft, bribery, or imposter kind of thing?

Paul Niehaus: So for that year, 2021, when we did that recap, I think that represented 10–20 bips or so of what we moved. And that’s a number that you always want it to be zero, but at the same time, if you say it’s going to be zero, the problem with that is that then people don’t discover it and report it. So there’s a sense in which that’s a number that I’m happy with, if that makes sense. Although I realise that may sound a little weird.

Luisa Rodriguez: I can see the logic. I want to ask more about whether that might actually be the right number or what even is the right number, if not zero. But first, can you say a bit about how you work to prevent it or investigate it? And I guess in some cases you’re able to rectify it by recovering the funds?

Paul Niehaus: Yeah, I can say a bit. I’m also actually not going to go into an enormous amount of detail — because part of the point is to always stay a step ahead, and some of the things we do we intentionally don’t want to disclose.

But as with most of these things, there’s a core principle of checks and balances and separation of powers. So the person who goes out and has that initial interaction with the recipient is not the same person who then subsequently goes back and backchecks with that recipient. And that’s different from the person who’s available by phone to talk to a recipient if something isn’t working or if they’re having problems, or who’s placing outbound calls to the recipient. And that’s also separate from our own internal audit team, which we have a function to do that.

So there’s a separation-of-powers principle: there’s multiple people looking at any one of these things, so that there is scrutiny from different angles. And there are certainly things you can do with data and with automation to look for irregularities and patterns that are powerful as well. But those are core principles.

Luisa Rodriguez: Oh, cool. Is there an example of the data-pattern thing that doesn’t risk someone figuring it out and working around it that you can share?

Paul Niehaus: I think that one will hold off on, for that reason. Suffice it to say that there’s a lot of transaction data generated, and that we look at it carefully.

Luisa Rodriguez: Fair enough. OK, so I am interested in this question of how much you should accept in terms of how much you’re willing to lose for this particular reason. Two hundred fifty thousand dollars, I don’t know, it seems high. It also doesn’t seem that high. How do you think about whether to invest more in preventing this from happening more often or invest less, because maybe those final cases of fraud are actually just extremely hard to notice, and you shouldn’t bother?

Paul Niehaus: It’s all case by case, as you’d imagine. To give you one concrete example, one thing that we did in the early days as a way of checking that we were finding the kinds of households we wanted was we were using satellite imagery. The idea was we were working at that time in places where having a thatched roof was a pretty good indicator of being poor, and so we were trying to target households that had a thatched roof. And we said that this is great, because we can see that from the satellite imagery. So if we get GPS coordinates from our field teams, then we can go back and plug those in and look at the images, and see if there is indeed a thatched roof there. And if there isn’t, then that can trigger an audit, so we’ll follow up and double-check on that.

So that seemed like a cool thing to do. And in fact, I think it was a lovely talking point when we were giving pitches about GiveDirectly. It just turns out that the false positive rate there is extremely high, so we were triggering a lot of backchecks and audits which, when we went out, we almost never found anything untoward. And so eventually we said that this sounds cool, but actually the numbers are bad, so we cut it. But I think a lot of the work looks like a series of case-by-case incremental tests and adjustments, as you figure out which things are actually catching things at a high rate and which are not.

Luisa Rodriguez: So it sounds like the overall picture is: there is fraud and theft. It amounts to about $250,000 a year, or at least it did in 2021, and it turns out that’s actually not something terrible. That’s about what you’d expect, given the amount of resources that make sense to invest in making sure it doesn’t happen more.

Let’s talk about what happened in the Democratic Republic of the Congo, which is GiveDirectly’s biggest fraud case to date. Can you give just the basic outline?

Paul Niehaus: Sure. And let me first just say we’ve written very publicly about this, and so I’ll talk about what we think we’ve learned from it, how we think we should interpret it in the big picture. But it is absolutely gut-wrenching to lose that much money. It’s something where we feel like we failed here, and that we owe an apology to folks involved: to the recipients, and to our partners in this project. So we’ve done that and done that very publicly. And I think that’s important.

In terms of what it means for the mission and the model overall, we feel like there are things we have to learn and adjust. It’s going to be less than a percent of all the money that we delivered in 2022. So we accept that this is a chess game that never ends, that we’re still playing it, and that we have to make adjustments. But fundamentally, I don’t think it shakes our confidence in our ability to keep doing what we do.

So with that having been said, what happened specifically in the DRC is — and as you can imagine, there are multiple layers to this — but the first and fundamental thing is that we have a control procedure that we usually impose, which says that when we give a SIM card to recipients, which is what they need to then be able to start receiving transfers, they need to go and register that with a mobile money agent themselves. That’s an important piece of our control process. And we made an exception to that in the DRC, because the DRC is a tough place to work, and we thought this would have meant long travel and potentially some risks for recipients. So we decided to give our field staff permission to register those SIM cards themselves in the name of the recipients and then distribute them.

We’re balancing risk and return there, in terms of thinking about how this is going to impact recipients and what the risk would be. And the core lesson from this is going to be that we got that wrong, and have to change that this time around. The issue this created was that in this case, some of our staff were able to register SIMs in the name of recipients, but then keep them, and instead give recipients other SIMs that were useless because they were not registered in our system to receive transfers. And so with those SIMs in hand, staff are then able to go and collude with mobile money agents to withdraw cash from the accounts that belong to the recipients and get it out themselves.

There are then multiple other accountability layers in the system that could have caught this — and that did eventually catch it, but that took too long, in part because the people who were stealing money at that point of sale were able to recruit accomplices in those other layers. So on net this went on for about four months, from the end of August 2022 to January of 2023, before we caught it. The design question for us now is, of course we want to catch it sooner than that if something like this ever happened again.

Luisa Rodriguez: So it was something like the people who are registering the SIMs realised that at some point maybe someone on an auditing team would notice that, maybe because people would complain, “I’m not getting the money that I was supposed to be getting.” But the people doing theft were able to recruit people on the auditing team basically, because they could give them a share of the money, I guess. And because of that, those additional checks didn’t work. Is that kind of the picture? Are there details I’m missing?

Paul Niehaus: Yes, that’s right. They didn’t work soon enough or as soon as we would like them. They did work eventually. There are multiple types of audits. So there is the internal audit team that I mentioned. There are in-person checks. There is a call centre placing outbound calls. There are also people there who are receiving calls. And so if a recipient calls in and says, “I was expecting money and it hasn’t shown up yet. What’s wrong?,” that should trigger a flag. Those things all just took too long to trigger because folks had been able to recruit accomplices in some of those other teams.

Luisa Rodriguez: Got it. How was it finally resolved? Who figured it out?

Paul Niehaus: For some safeguarding reasons, I can’t get too much into the details of what’s happened and what is happening. But what I can say is that eventually we did hear about it. There’s since been wide-ranging turnover — some because people’s contracts have just expired, but in some cases because we’ve let folks go and have referred some of them to the authorities for investigation, prosecution.

And then there’s a bunch of process stuff that we’re going to be doing differently. First and most important, of course, is not allowing this registration exception for SIM cards, or at least not unless there are additional controls in place. Also to improve the firewalling between the different parts of the organisation, to make it harder for people to identify and build a relationship with the people that are holding them accountable. And then third, there’s some stuff again that I mentioned that we can do in terms of automated data checks so that this stuff becomes visible to anyone, even if you’re not in the DRC, quickly, if something’s not happening.

Luisa Rodriguez: Very cool. Is there any part of you that wonders if it just doesn’t make sense to deliver cash in regions that are unsafe enough that you can’t have the processes that mean you can guarantee that the cash is going to get where it’s meant to go?

Paul Niehaus: I think that it’s possible that you could reach that conclusion for some parts of the world. I don’t feel that we’re anywhere close to that here, in the sense that I think essentially we had this core tradeoff to judge — Is it better to ask recipients to make these trips themselves or for us to streamline that for them? — and we’ve learned that we got that one wrong, and we should do it the other way. And I think we’ll do that.

I think that’s part of learning the optimal cost mix, which includes both the costs that we bear and also costs that are imposed on recipients as part of the process. We’ll keep learning as we go. I think part of the context here is the big picture that increasingly the world’s extreme poor do live in places that are fragile and conflict-afflicted. So learning how to work in places like the DRC — even if there are some mistakes made along the way and some pain in that process — is actually very important to addressing the problem of extreme poverty.

Luisa Rodriguez: Yeah, that makes sense to me. I do just want to say I was very impressed and appreciative and grateful, as someone who cares about GiveDirectly and has donated, that the team basically spoke as publicly as it did about this case. So I do think that is great. Are there any other lessons that you feel like come out of it, besides some of the specific things you’ve mentioned? I guess it’s not a new lesson, you’ve got this commitment to transparency and you’ve basically kind of delivered on that.

Paul Niehaus: Thank you. I really appreciate that. Of course, it is, as I said, gut-wrenching for us to have misjudged one of these things and got it wrong, and to have to fix it. And at the same time, I’d say that people have been, as you have been, incredibly supportive, and said this is actually in fact the way we need to talk about things in order to all learn and get better together. So I think that’s been great.

And that’s something that I’ve often told people over the years, and that I think still remains true: every time in our history when something has gone wrong and we’ve had to make a decision about how much do we talk about it and how open to be, I think that we’ve really been rewarded in the long run for being upfront and honest with people about it. And so I’d say that is encouragement to other folks that are starting out.

Luisa Rodriguez: Yeah, nice.

Universal basic income studies [01:22:33]

Luisa Rodriguez: Let’s push on to another topic. Another study that GiveDirectly is running is on universal basic income. I think there are studies being run in Kenya, Malawi, Liberia — and I think the one in Kenya, if I remember correctly, is actually the longest-running UBI study to date. I’m excited to hear at least some of the basic details, though the study is still in progress, so you don’t have the long-term results that you’ll eventually have. But just to start, can you explain the basic idea behind universal basic income?

Paul Niehaus: Sure. I think the way people usually define the idea is to say UBI is about giving everybody enough to meet some basic standard of living.

That’s actually a little bit incomplete without saying a bit more about where the money is going to come from. It turns out that there are some people who like that idea if what we mean is that we’re going to finance it with additional taxes, and there are some people who like it if what we mean is we’re going to cut some existing programmes that they don’t like so much. So superficially, it might seem like we all like the idea; in fact, we have very different things in mind. So I think that’s important to say, that you can’t describe UBI without saying where you’re going to get the money for it from.

The other thing I’d emphasise — and this relates to the point you just made — is that UBI is both the idea that everybody at any given point in time is getting enough to meet their basic needs, but it’s also this idea that that’s going to continue, so that at all points in the future of my life I could anticipate getting enough to meet my own basic needs. I think that’s very important, and we can get into this, but that’s one of the reasons why having this study in Kenya — which is going to be the longest-running, or longest commitment to providing people that basic standard of living — is very important.

Luisa Rodriguez: Cool. I do want to get into that. Before we do, I wanted to talk about just some of the common objections.

I think the most common ones I’ve heard are that it might disincentivise work among recipients. Is that something you’re worried about? It sounds like it’s not something you’ve seen in other programmes, but maybe it is the kind of thing that you might worry about when there is that long-term commitment?

Paul Niehaus: Right. So I think “disincentivised” is, in fact, not quite the right concept — in the sense that there are programmes where your eligibility for benefits tapers out as you get better off. Like the EITC [earned income tax credit] in the US, for example: there’s a phase out where if you’re earning above a certain level, you no longer get it, so there’s a very mechanical disincentive to earn more there. And that’s not what we’re talking about with UBI, because the whole idea is that it is unconditional on anything. It’s like, no matter what, you’re going to get this money.

I think what people actually have in mind here is not an incentive per se, but more that maybe you’re just less motivated if some of your basic needs are already met to go out and earn more — so it’s more of an impact that income or wealth has on your personal motivation, which is a somewhat different thing.

That’s also very important because I think — and I think the data also say — that those sorts of income effects are actually probably very different in different contexts. So in low-income countries in particular, people are extremely poor — so getting somebody from below the poverty line to $2.15 a day is by no means going to make them feel content with their life, or as if there’s nothing else that they wish they could have. And on top of that, one of the barriers for many of them to work is just access to the capital, to the tools they need. And so there’s this additional channel where having access to some money might actually enable me to invest in ways that would make it worth working more.

So what we’ve actually seen in the data on most cash transfer programmes in low-income countries has been either not much change in how much people work, or a bit of an increase — which is contrary, I think, to what a lot of people expected or were worried about.

Luisa Rodriguez: Cool. Yeah, I do feel persuaded in particular about this. If you’re taking someone just slightly above the poverty line, that feels pretty different to giving them some high monthly allowance that means they can not only meet all of their basic needs, but have all the luxuries they want. So yeah, I can see how someone just meeting their basic needs would not necessarily be discouraged from doing other types of productive work.

Before we move on and talk more about the study, I’m curious if you have a guess at what the best objection to UBI is?

Paul Niehaus: I think it depends a bit on where we’re talking about. In rich countries, if you do the math on something like UBI, it’s very expensive. And I think that in rich countries we have the administrative machinery to target benefits to people who are disabled or who have hit something that comes as a shock — like health insurance, things like that — in ways that poorer countries have less capacity to do. So if you do this sort of technocratic math, it’s not as clear to me that in some of the richer countries this would be the best way to spend a dollar to help people living in extreme poverty.

In poorer countries, it may be that some degree of targeting or means testing or something like that is a good idea, but the capacity to do that is more limited. So I think there’s a stronger case for, maybe it’s not universal everywhere, but in large regions, for example, everybody getting some degree of basic income. Something like that.

But the other thing to emphasise is that I don’t think that UBI is fundamentally a technocratic idea, right? It’s not like someone sat down and wrote out the optimisation problem of how can we do the most good for the world, and UBI popped out as the solution to that, with a given budget. It’s more like this would be a different politics and a different ethics of what we think a just society might look like, and something that people might be willing to get behind and therefore to spend or to give more than they would otherwise. So in some sense, I think that’s the real question about UBI, and it’s not one that experimental evidence of impact is going to directly answer — although it could contribute to some extent.

Luisa Rodriguez: Right. So that politics thing, the idea there is basically that currently we’re not thinking of these basic needs as a universal right the way we think of other things. Like, it seems like most people in most countries agree that no one should be able to physically harm you: that’s a right you have. And here I guess another example is that some countries think healthcare is a universal right, others don’t. But UBI is basically seeing if people can get behind the idea that people have the basic right to have their basic needs met, and the way of operationalising that is giving people enough resources to get at least those very basic needs met. Is that the basic idea? Am I getting that right?

Paul Niehaus: That’s it. Look at how political communication works, right? Nobody gets up and says, “Great news! I have this complicated plan. We’ve really thought it through carefully. It’s got these five different parts. Healthcare is going to work this way. And all this stuff, this is a great vision for what a fair society is going to look like.” It just doesn’t work that way. But potentially you could say, “I have this vision, which is that everybody should get enough to meet their basic needs,” and people might support that and be willing to get behind that. So the idea that this might be a politically viable narrative — even if it’s not dollar-for-dollar the absolutely ideal, optimal way to allocate a given budget — I think that’s very much an important part of the question about UBI.

Luisa Rodriguez: What do we know about that political viability at the moment?

Paul Niehaus: A bit. I think the best I’ve seen on this is a book I like, called Give a Man a Fish, which is about the new redistributive politics, particularly in sub-Saharan Africa and South Africa in particular. I think it’s still nascent. There are people that look at it and say, “Yeah, this seems like something that might get some traction.” And there are people who say, I mean, I don’t know, Andrew Yang is probably going to be peak UBI; it’s not going to get any higher than that. So we’ll see.

Luisa Rodriguez: Let’s talk about specific experiments GiveDirectly is doing. I’ve mostly heard about UBI proposals in relatively wealthy countries, and that already sounds pretty great to me. But in this experiment, people are able to go from below the poverty line — perhaps not having enough food to feed themselves — to having enough money to meet their basic needs for 12 years, which just sounds really wonderful to me. How much exactly are recipients getting?

Paul Niehaus: There are three different arms in this study, so three different groups. The core, which we’re calling the long-term arm, are the people who get that commitment for 12 years. They’re getting 75 cents US dollars nominal per day in monthly instalments for 12 years. The way to think about that is 75 nominal cents; that works out to about $1.90 at purchasing power parity — so adjusting for differences in prices between the US and Kenya — so that’s almost the $2.15 a day poverty line that we currently think about and talk about. It’s not quite, but it’s going to get everybody to that, because people were not starting from zero; they’re starting from some low number. So it’s going to be enough to get everybody over that extreme poverty line for 12 years.

Then there are two other arms, and those are there primarily as reference points or as comparisons. In one, people are getting the same amount of money as in that long-term arm, but for just two years. Part of what’s interesting about that is it means when we go back and do surveys before those two years are up, those people have received the same amount of money as the people in that longer-term arm, but their expectations of the future are different. That’s important because that lets us learn something about do those expectations of the future matter for the way you behave today?

Luisa Rodriguez: Right. Can you talk about the hypothesis for why that might matter, those expectations?

Paul Niehaus: You could probably imagine them. I mean, think about the way we all make our plans and think about the future. We contemplate whether to take risks where it may matter, how much of a safety net is there for us in the future. We decide whether to jump at an opportunity now or maybe wait to see if something better is going to come along. Those are the sorts of decisions that you might make differently if you envision that, for some very long period of time, I have this safety net there. We don’t know, but those are the kinds of things we want to test.

Luisa Rodriguez: Yeah, it reminds me, I was recently talking to a friend. I grew up in the US, but I live in the UK. And my friend thinks it’s just kind of absurd that I would ever hesitate before calling an ambulance. But this is just something I’m very trained to do, because in the US an ambulance ride might cost thousands and thousands of dollars. I guess there’s just a similar thing in this kind of financial stability sense, what you might do with your career. Or a bunch of life plans I can imagine making very differently if I knew that at the end of the day, things were going to be roughly fine. They are going to bottom out not at horrendous; they’re going to bottom out at tolerable.

Paul Niehaus: I think that actually academic economists find this idea fairly easy to comprehend, because it’s very like tenure, right? You get tenure and you’re like, “Great, unless I do something awful, I’m going to have a very nice job for the foreseeable future. So I can afford to take some risks and try some things that are a bit more out there” and that sort of thing.

Luisa Rodriguez: Yeah. Is there anything besides risk taking that it might influence?

Paul Niehaus: I think personally that another big one is if you’re staying in your community, and they’re all getting those transfers as well, is there’s this business opportunity that everybody here is going to have all this purchasing power for a long time. That’s kind of a hard thing to tease apart in the data, but my basic economic intuition says there’s a big opportunity there, and some people are going to jump at it.

Luisa Rodriguez: Right. Some people might start businesses knowing that there’s going to be more demand for goods than there was before because no one had enough money to buy things.

Paul Niehaus: That’s right. There is a third very simple thing to keep in mind, which is that, in principle, it’s possible that some people are able to get more of that money now. So if you have an annuity, for example, there are markets where you can sell your annuity and get a lump sum now. Here we think those financial markets are probably not going to work very well, and this is not an easy thing to securitise, because it’s not a standard thing: it’s a very unusual thing to be getting UBI as part of this study. But there may also be some ways in which people are able to essentially borrow against that future stream of earnings, so that could also matter in terms of what they’re able to do now.

Luisa Rodriguez: OK, just to make sure I understand, you can imagine something super informal, where some person knows they have this income that’s going to come in indefinitely. Maybe there aren’t exactly the formal structures for them to get a loan, but maybe they know someone who knows and believes that they’re going to get that money, and so they might be able to borrow from that person. And that ability to borrow could make a big difference to their plans?

Paul Niehaus: Exactly. And now we’re getting a little bit ahead of ourselves. In the data we’re not seeing an enormous amount of that. We don’t see that people in the long-term arm are holding more debt, either formally or informally, but it’s another consideration to keep in mind.

So there’s one more arm, in which people get the same amount of money as they do in this short-term arm that we’ve discussed, but they get it all at once, in one big lump sum.

So why is that interesting? It’s because if you look at everything that we know about the financial lives of people living in or near extreme poverty, putting together lump sums is very difficult. It’s difficult to do it by borrowing, because credit markets don’t work very well. It’s difficult to do it by saving, because there are lots of pressures on your saving — urgent needs, people asking for help — and savings vehicles just aren’t very good. Some people don’t even have access to an account that bears a positive interest rate on it. So for all those reasons, we think that it might be more impactful for people to get a chunk of money all at once than to get it spread out over the course of two years. That’s what that last arm lets us examine.

Luisa Rodriguez: Cool. So the thinking there is that there are things in life that cost a lot more than these people might be able to save. Maybe paying for some long-term education, or buying some bigger asset that they might be able to make productive. So the thinking there is that UBI might be pretty good, but maybe it doesn’t beat a single lump sum that would allow them to make some of those bigger purchases?

Paul Niehaus: That’s right. And to kind of presage a little bit what we’ll talk about, when GiveDirectly has done studies, we’ve gone to people and said, “If there’s a given amount of money that you could get, how would you like to structure it? Would you rather get it as 12 monthly payments over the course of the next year, or one big lump sum right now?,” overwhelmingly people say they’d like it in one or two or maybe three big lumps. And that’s really interesting, because that is starkly different from the way most cash transfer programmes are structured today. Most of them are these streams of smaller payments.

Luisa Rodriguez: Right. And I don’t know that you’ve done any surveys to find out, but do you suspect that the reason for that is basically most people feel like they’ve got some kind of big-ticket item they’d like to buy, or would like that option?

Paul Niehaus: Absolutely. And if you ask people, it’s straightforward, and it’s things like what you just described. It’s, “This is when school fees are due,” or “I want to purchase this,” or “I’m going to build a house,” whatever it is.

Luisa Rodriguez: It feels like another instance of this general principle, which is like GiveDirectly has this money, and the default should maybe be let people decide what they want to do with the money, and don’t impose constraints on them — for example, spreading it out across 12 years. What’s the best reason that spreading it out might actually be much better?

Paul Niehaus: I think that there are some people who might value that if they feel like, “It’s hard for me to save and budget and make sure that I consistently am able to meet basic needs.” There’s been some interesting recent research on the problem that people face during the lean season — the time of year that’s before the harvest, which is when cash runs shortest for most people. This is something that happens every year, and you can anticipate that it’s going to happen every year, and yet still sometimes people don’t have enough money or just enough food set aside to budget for that, and it can be helpful to kind of help them plan for it. So if you’re somebody who knows that’s something you have a hard time with, I could totally imagine that it’s helpful to say, “Let’s just make sure that through that part of the year, money shows up every month.”

Luisa Rodriguez: Yeah. I can imagine some sceptic thinking that the distributed payments might help people be more financially responsible. And this is not me implying that people in poorer countries are more financially irresponsible — I think I, for example, would have a harder time making financial decisions I endorsed if I got all of my salary for the year at one time. Is that something that you can investigate?

Paul Niehaus: I hear what you’re saying, and I love this point, because I think it both illustrates the insight and the hubris that we sometimes bring to these issues. I think it is exactly right that people living in poverty sometimes have a hard time sticking to a plan, or using money in the ways that their long-term self would like to. And they say that, but what they say is actually the exact opposite of what we just said, which is, “The hard thing is, if I have a little bit of money every month, there’s all these pressures of people asking for help. I’m going to be tempted to just use it on something fun today. So it’s going to be hard for me to ever save up and make the investment or buy the big thing that I wanted to. So I actually really need the lump sum.”

So this intuition — that the problem is that if you get a lump sum you’re going to blow it — is exactly the opposite of the way people are thinking about their own self-control problems. And it’d be silly not to tap into that. They understand it, I think, much better than we do.

Luisa Rodriguez: Yeah, I buy that. Cool. That just makes lots of sense to me. Should we talk about the results? At least the preliminary results?

Paul Niehaus: Yeah, so we haven’t put out a full set of preliminary results yet, so I’m not at liberty to talk in much depth about any of that. We did put out an initial set of results during the pandemic, and the philosophy there was simply to let policymakers know what we knew at that time that we thought might be relevant for policymaking during the early days of COVID. That was actually quite exciting, in the sense that I think there’s some basically positive results in there that contributed to some other countries in sub-Saharan Africa deciding to ramp up their cash transfer programming in response to COVID.

Luisa Rodriguez: Very cool. So one set of results was promising enough that other countries started doing transfers during the pandemic. Can you say what was so exciting about those results?

Paul Niehaus: You know, it’s fairly straightforward stuff. It’s saying that people are less likely to be experiencing food insecurity, unable to eat enough, during the pandemic. So I don’t think it’s anything deeply surprising, but there were some questions in the early days of the pandemic about if we’re in lockdown and people can’t even go out to buy food, is it even going to matter if they get cash transfers? Things like that. I think it helped a little bit to address some of that stuff.

The other thing that was interesting is that if you look at the people who had been getting transfers before the pandemic started, they see less of a drop in their food security, but more of a drop in their income — because they had in fact started businesses, and been earning more money before the pandemic. So you see this dual function of the transfer: they’re stimulating investment — which in one sense makes you more vulnerable to a big economic shock, because your business might lose all of its customers — but they’re also providing you with this cushion, where you may lose a lot of revenue, but you’re still able to put food on the table.

Luisa Rodriguez: Moving on to the broader results, outside of the pandemic, what are the basic things we’re seeing so far?

Paul Niehaus: We haven’t put this out, so I’m only going to talk in generalities about things that I’d say we all agree are pretty fair big-picture takeaways at this point.

Three things I think stand out. One is just that you see this big expansion in local economic activity, analogous to what we talked about earlier in the context of the general equilibrium study. The second is that because I think people care so much about work and labour supply, and how hard are people working is this disincentivising effort — although as I said, I think that’s not quite the right phrasing — but again, we’re not seeing significant changes in how many hours people are working. We are seeing some changes in how they’re working, in that they are moving away from working for other people and towards working for themselves — so more self-employment — and that self-employment is mainly going to be in retail and non-agricultural self-employment.

And then the third thing, which I think is really interesting, is that certainly on some of these aggregate measures of economic activity we are seeing pretty big differences — in some cases significant — between that long-term arm where people in the community expect to get money for 12 years, and the short-term arm where they’re only getting it for two. Even though at the time we measured these outcomes, they had both received the same amount of money. So their payment streams up until now, at the time that we measure outcomes, have looked identical. The difference is that in one group they expect them to continue longer — and in that group, we see these much larger effects on economic activity.

So we’re still kind of digging through and thinking about what we can say with some confidence about why exactly that is. And we’ve talked earlier about some of the possible mechanisms, but the very practical reasons it’s important is because there are many of these UBI pilots and evaluations now being run around the world — which is cool, and I think we can learn a lot from them, but most of which are very short term, right? A year or two-year commitment. So what this result says is that, at least in our context, we’d miss something important if we weren’t able to look further out than that.

Or maybe another way to say it is that I think all of these are going to be very informative about what people do with incremental money. So if that’s the main thing that you’re worried about with respect to UBI, or cash transfers in general, this is going to be really helpful data for you. But if what you’re wondering is, “Do people live their lives differently if they know that there is this safety net there for them in the future?,” we don’t learn about that unless we actually make that long-term commitment.

Luisa Rodriguez: OK, so the experiment in Kenya is meant to last a total of 12 years. When those 12 years are up, what kinds of outcomes do you hope to see for recipients?

Paul Niehaus: I don’t know that I’d say there’s something specifically I’m hoping for. I’m kind of doing this because I’m curious, and I think that we’ll learn things that will matter for decision making.

One of the things that I’m curious about is this idea that with a long-term transfer, you don’t need to commit to a course of action right away, because you have time to wait. And I don’t know if that’s going to be true, but it could be that there’s a little bit of difference between long term and short term, or long term and lump sum, initially, but then over time, you really start to see people say, “Here’s the opportunity I’ve been waiting for. Now is the time when my family’s in good shape, and I can afford to move to the city,” or whatever it is, right? I feel like that’s one of the most important hypotheses that we’re going to be testing as we continue to track people over the longer run.

Luisa Rodriguez: Cool. When can we expect an update?

Paul Niehaus: It’s always “soon” with academic papers. We have a responsibility to be really sure that everything’s right, and that’s the thing we’re going to prioritise. But that said, I’m thinking that “soon” is this year, if all goes well.

Luisa Rodriguez: Well, we look forward to hearing that update, and we’ll link to it whenever that comes out.

Skyjo [01:44:43]

Luisa Rodriguez: OK, we’ve got time for just one final question. So your research focuses on accelerating the end of extreme poverty. Is there a topic you’re into without clear social benefits, that’s kind of purely just intellectual curiosity?

Paul Niehaus: I think the most useless thing I can think of, the thing that is like there’s no plausible way in which it could make the world better, is I play kind of a niche card game called Skyjo with my friends. My summer project has been to try to teach an AI to play it well enough that I can pick up new strategies from the reinforcement learning algorithm so that I can beat my friends. And that’s entirely zero sum. Whatever I gain, they’re going to lose.

Luisa Rodriguez: That’s awesome. What does that entail for you? How are you basically training it, and how do you learn from it?

Paul Niehaus: It’s reading up on reinforcement learning, understanding it. It’s cool stuff, and it’s actually very similar to the modelling techniques that economists use to solve dynamic optimisation problems. It’s learning some coding. I’ve been blown away at how good ChatGPT is as a teacher and a tutor, so I’m able to pick up new coding things so much faster. I’ve been really impressed by that.

Luisa Rodriguez: Very cool. Have you been able to use any of the strategies yet?

Paul Niehaus: I have so far only built totally naive strategies that play randomly, so I’ve learned nothing to date. I have some unresolved questions about the size of the game and how much computational resources I’m going to need to be able to actually do this, and whether it is sort of the thing that one can actually do as a summer project or whether I need to go out and raise a round of funding if I want to do this. So I’m still trying to calculate that, but it’s been fun.

Luisa Rodriguez: Cool. I love that. Well, my guest today has been Paul Niehaus. Thanks so much for coming on.

Paul Niehaus: Thank you so much for having me. Thank you all for listening, and to the extent this has been compelling and you want to support us, GiveDirectly.org and sharing the evidence on what we’re doing I think is the other key thing to do.

Luisa’s outro [01:46:37]

Luisa Rodriguez: If you want to learn more about the most effective ways to alleviate poverty, I’d recommend going back to listen to some of our past episodes:

• #153 – Elie Hassenfeld on two big-picture critiques of GiveWell’s approach, and six lessons from their recent work
• #129 – Dr James Tibenderana on the state of the art in malaria control and elimination
• #124 – Karen Levy on fads and misaligned incentives in global development, and scaling deworming to reach hundreds of millions
• #105 – Alexander Berger on improving global health and wellbeing in clear and direct ways

All right, The 80,000 Hours Podcast is produced and edited by Keiran Harris.

The audio engineering team is led by Ben Cordell, with mastering and technical editing by Dominic Armstrong and Milo McGuire.

Additional content editing by me and Katy Moore, who also puts together full transcripts and an extensive collection of links to learn more — those are available on our site.

Thanks for joining, talk to you again soon.