Elie Hassenfeld: [00:00:00] Hey, everyone. This is Elie Hassenfeld, GiveWell's co-founder and CEO, and today I'm sitting down with Adam Salisbury, our senior program officer focused on livelihoods, to discuss the work that he's doing to support programs that could increase people's incomes and give them the ability to purchase more of the things that they want.

As many of you know, GiveWell has supported livelihoods-focused programs in the past, but we've predominantly focused on health-related programs. Some of the reasons that we've been more focused on health is that health programs are easier to generalize from one context to another. For example, a measles vaccine works the same way whether it's administered in Nigeria or in Kenya.

Income-increasing programs can be very different, whether they're delivered in, you know, one type of context from another—for example, in a rural area versus an urban one. Income-increasing programs are also harder [00:01:00] to measure the effects of than health-related programs. With health-related programs, often what's measured is objectively measurable via a test.

So for example, in malaria, a rapid diagnostic test, or via mortality directly, where in income it's a lot harder for researchers to assess the impacts of programs. And then finally, an input into the comparison that we make between income-increasing programs and health-related programs is something that we call moral weights, where we try to assign values to the good that is achieved by averting a death or improving someone's health relative to the good that is achieved by increasing income.

And as is probably obvious, these are not questions for which there are clear true answers. Instead, these are philosophical questions about which I don't think GiveWell has any particular comparative advantage in arriving at truth. And so these moral weights also play a major role in our decision-making about how to prioritize income-increasing programs versus health-related programs.

As GiveWell has [00:02:00] grown and we have more capacity and more researchers on our team, we're in a position to expand the scope of the work that we can consider. And so with that additional capacity, we can look more at these livelihoods programs and both offer the ability to support programs to donors whose philosophical judgments, whose moral views, may be different than ours, and they might want to prioritize livelihoods programs more highly relative to health-related programs. It also gives us the opportunity to explore an area that we had not explored in as great depth in the past as we ideally would have, and gives us an opportunity to learn more about an area that could be very promising.

So today in this conversation, we're planning to do a big run-through of what Adam and this portfolio is focused on right now, and why we're focused on it. We start the conversation with the set of income-increasing programs that have more evidence behind them, and so these are programs like GiveDirectly, which delivers cash transfers to [00:03:00] people in extreme need, and poverty graduation programs, which tend to deliver some degree of training and financial support or a direct asset to people. And both of these programs have significant evidence behind them.

And then we'll move on to programs with less evidence, where the effects are somewhat more speculative. And for us, this means supporting work that could enable governments to deliver their own poverty-alleviating programs more effectively, and also research that we're supporting to help us make better decisions about where to direct our future funding, and hopefully that could influence other policymakers and actors to deliver poverty-related funding more effectively in the future.

You know, right now our team is very small that's focused on livelihoods. And so there are areas like microfinance and programs that focus on increasing agricultural production that we don't have the capacity to look into at the moment, and so we are hiring for this team. We have a job open [00:04:00] for senior livelihoods researcher, which we hope to add so that we're able to cover more ground in the near future.

And so with that, Adam, thanks for doing this, having this conversation today.

Adam Salisbury: Thanks for having me.

Elie Hassenfeld: I think we should just, you know, dive right in with some of the areas that you've been focusing on so far. And let's just begin with GiveDirectly. So maybe you could just give brief background on them and the work that you're focused on with them.

Adam Salisbury: Yeah. So you've summarized GiveDirectly's core program well at the start. So, what they do is give $1,000 unconditional cash transfers to people in poor regions of low-income countries. And there's no strings attached, so they can use the money to meet whatever needs they have. There is good evidence that when given $1,000, people use it productively.

If you look at the top two things people buy, it's food and home improvements, so food and shelter, two very basic needs. Very, very small fractions of the money is spent on things like alcohol or bad goods. That's backed up by several RCTs that were baked into GiveWell's [00:05:00] previous assessment of GiveDirectly.

About two years ago, we revisited their large lump sum program, and we updated our assessment, so we now think it's about three times as cost-effective as we did previously, and the two major updates came from newer evidence that's come to light. First, about the mortality benefits of cash transfers, so there's new studies showing that infant mortality fell in households that received a $1,000 cash transfer.

And the second is there's evidence from western Kenya that when you give $1,000 to people, that can create this virtuous spending cycle where recipients spend the money in nearby shops, nearby businesses, and that can benefit non-recipients, so people that don't receive the cash transfer. So the motivation behind these pilots were, like, we've just updated our assessment of this program. We think it's about 3X as cost-effective as we did previously, but we still don't think it's as cost-effective as what our marginal dollar would go towards. But we were wondering, are there iterations of their core program which could be even more cost effective? And so we funded three pilots to try and answer that question.

Elie Hassenfeld: [00:06:00] Cool. So tell, I mean, just what are those pilots? What are you most excited about from what we've supported there?

Adam Salisbury: Yeah, sure. So one of the pilots is happening in Malawi, and it's happening in the context where there's a large experiment happening anyway on the impact of cash transfers. The specific thing we funded was providing cash transfers to local business owners that they would receive a few months before households receive their cash transfers, and part of the rationale for that is if people know a demand shock is coming, they can maybe stock up on inventory, they can stock up on employees to best take advantage of this demand shock. Those cash transfers were disbursed last year. We should get preliminary results from that later this year. So we'll be keeping abreast of that. The second pilot . . .

Elie Hassenfeld: Let me just ask you about that one first. So what kind of results would you imagine seeing from this pilot? Like what will be assessed and what will look positive here?

Adam Salisbury: So the primary outcomes would be consumption, so that would be consumption of both the recipients of household cash transfers and the recipients of these business grants. Prices would also be assessed. So [00:07:00] one, one worry someone might have if you drop a large amount of cash on an area, does that cause inflation? A trial in Kenya found that it didn't cause inflation. One argument for giving cash transfers before, like if people could stock up on inventory and can stock up on resources, inflation seems less likely because supply constraints seem less likely to bind.

Elie Hassenfeld: I see. So basically what we're testing in Malawi with GiveDirectly is we know that GiveDirectly gives these large cash transfers to households, and then people have money and they wanna go to shops and spend it.

But the shops don't know or aren't prepared for this increase in cash. And even if they knew it was coming, they might not have the wherewithal to increase their inventory of goods. But the theory is, if business owners are given cash in advance, then they can go out and purchase additional, let's say stuff, before the cash transfers come in.

Then households, when they get the money, will have more stuff available to buy. So they can get the stuff because there's a greater supply of goods, prices are less likely to go up, and because [00:08:00] the business owners invested in more inventory ahead of time, they can also earn more money when this cash comes in.

Is that like the basic rationale? And so we should see it flowing through to households being in a position to buy more, potentially because prices—well, we don't know what will happen with prices, but maybe they won't go up or won't go up as much as they otherwise would have. And then business owners should make more money because they have more supply in stock.

Adam Salisbury: That's essentially right. I think I would just emphasize, like, what these businesses look like. So most of the time, these aren't like supermarkets. These are often one-person businesses. So you can imagine if it's just one person owning like a grain mill, there are certain investments that person can make before the cash comes in that could mean they're able to service the increase in demand better. So they could maybe rent another machine, they could hire some more employees. So the question is whether doing that amplifies the consumption benefits of cash transfers while also maintaining this sort of lack of inflationary effect we saw in Kenya.

Elie Hassenfeld: Right, and I'm even imagining one of the stores like this that I visited last year when I [00:09:00] was in Malawi and it's, you know, it's like a one-room shop that is, I don't know, 10 by 5 in terms of size.

And then it just has shelves which, you know, basically cover the scope of what you would find in a normal convenience store from, you know, shelf-stable drinks and snacks to you know, basic medicine, some rice. And, you know, I imagine as a store owner if you knew that all of a sudden people in your village were going to get a lot more money and someone was offering you some additional cash to prepare, you'd get more supply and maybe even some valuable, more expensive products that people wouldn't normally be in a position to purchase, would allow you to earn more, and allow the people with the cash to improve their quality of life more directly.

Adam Salisbury: Yeah, that's right.

Elie Hassenfeld: Where did this idea come from? How did this come about, and why does this idea seem like a particularly good one to test?

Adam Salisbury: So this came about in collaboration with GiveDirectly. I should say this was before I took over the portfolio, but about 18 months ago, after this update, we approached them and said, "We're interested in, like, testing iterations to your core program that can be more cost-effective. Come back to us with your best bets as to what that could look [00:10:00] like."

They sent us back four ideas. We ended up funding three of them, and I can talk about the other two. And then the other idea was picked up by the Livelihoods Impact Fund, who are another funder that we work quite closely with.

Elie Hassenfeld: Why don't you briefly just describe the other two pilots?

Adam Salisbury: Yeah. So the other one is layering cash with the construction of rural footbridges. So a bit of context here is back in 2022, GiveWell co-funded a randomized trial of footbridge construction in Rwanda. So what effect did building footbridges have on the consumption and well-being of nearby communities?

We got some preliminary results from that last year, and it looked pretty promising. So about a 6% consumption effect, but because this spread over such a large area, it looked quite cost-effective. We made a grant to Fika, formerly known as Bridges to Prosperity, who were the implementing org in this randomized trial.

So bridges were sort of on our radar anyway. One thing we were wondering is, could bridges amplify some of these positive spillover effects of cash that were found in Kenya? And the theory of change behind that is if you read [00:11:00] this paper in Kenya, the story of these positive spillovers is really a story of demand-led economic expansion or a virtuous spending cycle.

And I think there are some interesting analogies here to what happened during COVID in the US. So the economy became depressed. The central government gave stimulus payments, so about $1,000 to most households within the US, and the idea was that would stimulate the economy. People would go out, they'd visit shops, they'd visit restaurants. The owners of the shops and restaurants would then spend that extra money, and it would sort of kickstart the economy.

For that story to take hold, people need to be able to transact with each other, so they need to be able to visit each other's shops, visit each other's businesses. If you remove an access constraint, so in this case, a literal access constraint, like the fact that people can't easily cross over a river or a very steep ravine, if you alleviate that at the same time you deliver cash, can you amplify these positive spending cycle effects again?

Elie Hassenfeld: Right. And so I think, like, an interesting thing to note both about the Malawi pilot and about the footbridges pilot is that I think often I would think about GiveDirectly, and I think many people think about [00:12:00] GiveDirectly as primarily or almost like solely benefiting the households who receive the cash.

The thing I remember most from the visits that I've made to households that have received cash via GiveDirectly is that people will often replace their thatched roof with a metal roof. That improves quality of life, and that seems like a very straightforward and obvious benefit to enable you to sleep through the night without rain coming in.

I mean, this is a really good thing to provide. And then what I think we or I had not thought as much about, and I imagine people might not, is the fact that when those households get this money and they're purchasing, they're operating in a local economy, and these two pilots both, the way you put it was to amplify the broader effects.

That if we can make it easier for people to transact and spend money, there is, you know, some material evidence that that can have a much larger effect—a positive effect—than just on the households alone.

Adam Salisbury: Yeah, I think that's right. The one thing, the one flag I'll put up now, I don't want to go down this rabbit hole, but I would say this is my major uncertainty about [00:13:00] GiveDirectly's program.

So we've got these results in Kenya which say exactly that. How generalizable are those results? Like, because there were these positive spillovers in western Kenya, would there be similar positive effects in rural Malawi, for example? I've got a few research grants I'm investigating now focused on that, which I can touch on later. But, yeah, I don't want to take us completely away from the pilots.

Elie Hassenfeld: Well, we'll come back to the research questions in a minute. But just for the sake of rounding this out, what's the third pilot that we're supporting GiveDirectly on?

Adam Salisbury: So the third pilot is in Mozambique, and it's around sort of like hyper-targeted cash transfers.

So GiveDirectly's standard program doesn't do very strict age-based targeting, so you have to be over the age of 18 to receive a cash transfer, but there's no other restrictions beyond that. It also doesn't do very granular poverty-based targeting. What they do is they will target poor regions of a country, and they'll use government surveys for that typically. But then once they've identified a poor village, they will give cash transfers to everyone in that village. What this pilot is testing is if we take very, very vulnerable young people, do they use the cash transfer [00:14:00] productively? I think there's two reasons—

Elie Hassenfeld: And before you say that, you just said "vulnerable." Like what, what do you mean by vulnerable young people?

Adam Salisbury: So I mostly mean economic vulnerability. So people in like extreme poverty, not just people that are sort of around the ultra-poor poverty line of around $3 per day, but people that are around possibly half of that. So like very, very at the very subsistence threshold.

I think there are two reasons we care about this. Like one, just in terms of our moral weights, we care very much about improving the life and living standards of those households because when you're on that level of poverty, an increase in consumption by like a dollar a day can be the difference between having enough meals to eat and not.

So in welfare terms, that seems really consequential. The other reason I'm interested in about this is because of the age angle. I think there's a reasonable hypothesis that young adults are at a very critical period in their life cycle. So between the ages of 18 and 35, that's typically when people decide whether to migrate, whether to set up a business, whether to start a family. So it's plausible to me that cash transfers timed at that specific window [00:15:00] could have outsized effects, basically

Elie Hassenfeld: One thing that I remember a lot . . . I visited GiveDirectly households in Kenya many years ago, and when I was there, it was a time when they were doing some poverty targeting within villages, giving cash to households with thatched roofs and not metal roofs. And they switched from that, like you said, to be less targeted, and one of the reasons was some of the dissatisfaction that I think they observed in villages where, you know, people who were still extremely poor by any reasonable definition weren't receiving this big cash, and there was, you know, jealousy and unhappiness that certain people in villages were getting cash and not others. And so how are they planning to address this topic in this pilot?

Adam Salisbury: Yeah, that's a great point. So that's exactly right. They used to do more granular targeting based on thatched roofs. They moved away from that because people were getting annoyed that this was just a very blunt measure of poverty, and so some legitimately poor people were being excluded.

That's a key priority for these pilots, testing social acceptability of this type of targeting mechanism. They'll be doing that mostly just through [00:16:00] interviewing community members, like before and during and after the pilot. There are two reasons where I'm maybe more optimistic that this will be more socially acceptable than their previous thatched-based targeting.

One, I think the targeting method is more sophisticated. So they're working with AtlasAI, who are a research group that produce satellite-based predictors of poverty. This is generally much more sophisticated than just measuring the material of someone's roof. So I'm hoping that there'll be less sort of exclusion errors, which is what we call them, legitimately poor people get excluded.

And then my gut says that age-based targeting may be more socially acceptable because people that survive to the age of 18, everyone passes through that cycle. But it's definitely an open question and something we're willing to test, and if it's considered not acceptable, like we wouldn't progress beyond the pilot.

Elie Hassenfeld: Okay, thanks. So that's the GiveDirectly and the pilot. It's very interesting. Let's move on to the next big area, the graduation programs, which I think also at times have been called like targeting the ultra-poor programs.

Adam Salisbury: Yeah. So I guess just to take a step back, so GiveWell's most famous for our top charities. So these are programs that deliver basic health [00:17:00] commodities. These are very cost-effective, very scalable, and very evidence-backed. I think the closest we have to those type of programs in the livelihood space are cash transfers and these things called poverty graduation programs.

I think they're very evidence-backed. They've all had like 30 plus RCTs on them. They're pretty scalable, and they're reasonably cost-effective. I don't think as cost-effective as the Top Charities, but like reasonably cost-effective. Poverty graduation programs entail first identifying extremely poor households within villages, so it's a more granular filtering process, and then providing these households with this wraparound program that entails giving them a productive asset, so giving them something like a sewing machine, giving them training in how to use that asset productively, and then also giving them consumption support for about 12 to 24 months so they don't have to sell the asset if they're hit with a negative income shock.

Elie Hassenfeld: Consumption support would be like small, ongoing cash transfers that enable them to have enough money to purchase food, et cetera.

Adam Salisbury: Exactly. It's typically cash transfers, but it can be like food vouchers as well. I should say as well there's no standard definition of what a [00:18:00] poverty graduation program is. Other iterations of it, I think the sort of three essentials are a large lump sum transfer, some sort of training, and then some sort of savings facilitation as well So they often set up village loans and savings associations which help community members pool their savings together.

Elie Hassenfeld: And so how does the evidence of this program, you mentioned that it has a large body of evidence, but what's the impact that these programs have and certainly the cost per impact relative to cash?

Adam Salisbury: So there's very robust evidence that these programs improve short-run consumption outcomes of recipients. So most of the evidence comes around two to three years after the program was implemented. There is some variation. In the literature, there are some programs which didn't have impact, but about the median effect is about a 10% increase two to three years later, which is in a similar ballpark to GiveDirectly's unconditional cash transfer program. At the moment, I think the way we're modeling it, so I should say our poverty graduation work is more a work in progress.

My sense is that the average program is similar for both unconditional [00:19:00] cash transfers and poverty graduation. I think there's more underlying variation underpinning these poverty grad estimates. Because I mentioned there isn't a standardized program, I think there are possibly some iterations of this program which are cheaper to deliver or deliver larger impacts that could be more cost-effective. And part of my plan this year is to see if we can identify this right tail of, like, really cost-effective graduation programs.

Elie Hassenfeld: And so what's the trade-off? It's that they do more, so they are more costly, but they potentially have more impact. Is that, like, the basic—

Adam Salisbury: So they can be. So GiveDirectly is about, it gives $1,000 to people. There's an overhead, but it costs about $1,200. Some graduation programs are more expensive than that, so around $2,000. But some programs are cheaper, around $500. There is, like, an open question as to, like, how much can you strip out of this program and still maintain those effects, and I can talk about kind of what the evidence says there.

One other big uncertainty I have is around the persistence of gains for both cash transfers and poverty graduation. I mentioned that there's very robust evidence in the short-run benefits. The long-run evidence is much thinner. There's only about three [00:20:00] studies that have tracked recipients beyond five years.

My read of the evidence is it shows slightly more persistent gains than unconditional cash transfers, but it's very much an open question, so some of the research I've funded is trying to generate more of this long-run evidence.

Elie Hassenfeld: And so basically it sounds to me like as you've come into this role leading our livelihoods work, which you've been in now for approximately—

Adam Salisbury: Five months . . .

Elie Hassenfeld: Five months, so it's still, it's still pretty early. In some sense, what we're talking about at the start of this conversation are the two big programs that have the most evidence behind them.

There's cash, and then there's the poverty graduation programs, and then the real question is how cost-effective are they overall? And how do they compare to each other? And because of the variation in location and in approach, and that's both with GiveDirectly, we're talking about potential variations and iterations on its program, but also the wide variation in poverty graduation programs.

It sounds like [00:21:00] your hypothesis, which seems reasonable, is that there should be variations in locations where each will be more cost-effective than the other. And so a major part of your work now is trying to sort through which are the underlying conditions, the program characteristics that are most worth supporting because they yield the greatest future benefits.

Adam Salisbury: Yeah, I think that's right. The way I would describe my three biggest research priorities for both interventions, so one are, are there iterations of this program, what does the most cost-effective version look like? I spoke about that with the GiveDirectly pilot, so that grant is very much in that category.

The second question I have, I just touched on, like what are the long run effects of these programs? Like, are there still effects seven, eight, nine years later? The third question, which I haven't kind of touched on, are what happens when we take these programs to scale? So when we deliver cash to like large areas of a country, what happens to prices?

When we deliver poverty graduation at scale, is there any risk that some of these small businesses start competing with each other and eroding aggregate gains? There is some preliminary evidence on both those questions, but it's a pretty nascent field, I guess. So funding more research in that space is also something I'm [00:22:00] looking at.

Elie Hassenfeld: So we're going to come back to research in a minute, but maybe first just like what's a poverty graduation program, like what's the funding or potential funding to a poverty graduation program so far that you think is most promising?

Adam Salisbury: There's a few things I'm looking at. So the largest NGO that delivers poverty graduation programs is one called BRAC. They operate in Bangladesh. They've been operating—they actually invented poverty graduation in the early noughties, so they're sort of experts on how to deliver this. We're investigating a grant to them at the moment.

There's a few shapes that grant could take. One could be just supporting direct delivery in Bangladesh. I think the one thing I'm more excited by is funding direct delivery, but in the process of doing so, layering on some answers to open research questions we have. Like, if we can randomize rollout in Bangladesh, we might be able to shed light on these questions of like what happens to prices and markets when we do graduation at scale.

So that's in the NGO category. One direction of travel that I think is important to flag with both poverty graduation and cash transfers is that governments in low- and middle-income countries are increasingly delivering these programs. So I'll throw a statistic at you. [00:23:00] Back in 2015, about 100% of people globally that were reached by these poverty graduation programs were reached by NGOs.

In 2025, that figure was about 35%, and about 65% was government programs. So there are two opportunities we're looking at about whether there's a role for GiveWell to support the government rollout of these programs. One of them is a technical assistance opportunity in India. So the Indian government is actively trying to scale up this program right now.

The question we're asking themselves is, can we fund advisors from BRAC, who are the sort of leading experts, The Nudge, who are an India-based NGO, and J-PAL, who are a leading evaluation partner, can we fund them to try and make sure this rollout goes as well as it can do? That's one category which I'm happy to talk about, and the other category is, can GiveWell work with an organization like the World Bank to scale government programs directly?

So there's a specific program in the Sahel, so in West Africa, which I'm looking at, which is attractive for us for similar reasons to I discussed in Mozambique, like Sahel [00:24:00] is home to some very, very poor people, very vulnerable. We care a lot about improving the welfare of those people because it can often be the difference between getting enough calories to eat and not.

Elie Hassenfeld: So I want to talk more about the supporting governments. But before we move on from the key programs, like one thing that you said, but I think it's just worth checking and emphasizing, is I think in both in the cash pilots and also in the graduation programs, both program delivery and research, experimentation, and learning are critical components.

And so with everything we're supporting we both anticipate significant improvements in the welfare of the people who are benefited by the program, but also significant learnings that inform ideally GiveWell's future giving and hopefully, optimistically, others' future giving in this space as well. Is that right? How would you describe that?

Adam Salisbury: That's right. And I think another way I would explain why this is a priority for me is just taking a step back and looking at the livelihood space in general and what evidence generation has looked like over the last 30 years. [00:25:00]

To be honest, I think it's got a checkered track record of striking the right balance between investing in program rollout and investing in learning. I think there are some examples of programs like cash transfers, like poverty graduation, like PROGRESA, which was a social welfare program in Mexico, where there was a lot of research which happened at the same time as rollout.

So both governments, implementers, and researchers learnt a lot about the optimal size of cash transfers, the optimal conditionality criteria in the case of Mexico, the optimal assets in the case of poverty graduation. I think there's also been interventions that I'd put in the livelihood space, like the Millennium Village Projects, for example, which were a program which delivered farming interventions with health interventions with, like, infrastructure interventions in villages.

I think those programs under-invested in research. There was no randomized evaluation attached to that. It's quite hard to say exactly what that project accomplished. Part of my motivation in this role is to make sure that newer interventions are more, that we strike the right balance basically between generating research but also scaling things along the [00:26:00] way.

Elie Hassenfeld: Yeah, one of the things that excites me about this research, and really like GiveWell's in some ways, you know, unique role, is that, you know, we're obviously an organization that's very interested in research and learning so that we know what to fund. And because of that, we have this unique position where we're excited to fund very rigorous research, you know, essentially like academic or academic-style research, but it's directly connected to programs we can support and funding decisions that we will make.

And, you know, I think that people may not realize that that is often very rare, that the funders and practitioners of research are not directly connected to programs that could be scaled up. And in this case, the research that GiveWell is supporting via your livelihoods work, will feed directly into future decisions that you make about where to allocate money and which programs to scale up.

Adam Salisbury: I think that's right. Now the evidence to like policy or decision pipeline is just very, very tied at GiveWell, because it's the same people consuming the research are the same people making the decisions.

Elie Hassenfeld: Yeah. [00:27:00] Well let's just talk a little bit about the research now. So we're going to go beyond just the research on specific programs, but the big research questions you have and what support we've provided.

Maybe we'll start with, you know, this big question that I guess you didn't want to go into before, but how the effects of GiveDirectly's program, which we've largely gathered from its long-term work in Kenya, might generalize to other locations and other scales. So just tell us a little bit more about what's on your mind there and what you're doing.

Adam Salisbury: Yeah, sure. I'll come at this from, like, the highest possible perspective and then drill into the more specifics. So I think, that when I joined GiveWell, I think my prior would've been that effects wash out more at scale in the livelihood space. And part of why I thought that is that there are—it's not that hard to find promising treatment effects in the randomized literature in livelihoods.

So finding programs that cause a 10 percent increase in consumption, like, there's lots of those in that category. But if you zoom out and look at aggregate growth statistics, so if you look at median consumption per day, for example, those trajectories in places that have received a lot of livelihoods funding like Uganda, [00:28:00] Malawi, those have remained much flatter than the trajectories in the health space, for example.

So in the health space, we had randomized trials in the '80s and '90s that showed large mortality reductions from vaccines, from bed nets, from antiretrovirals. Those programs were scaled up in the noughties and in '10s, and then we did see, like, pretty big mortality reductions in both kids and adults.

Something that sort of rocked my priors a bit was this GiveDirectly update two years ago. So there was a big RCT in Kenya which delivered cash to people, and the implication of that study was that actually general equilibrium effects could point in the positive direction. So for every one dollar of cash delivered in this context, around two dollars fifty of aggregate economic gains were generated.

We spent a lot of last year kicking the tires of this paper, so in great credit to the authors, they made all the code and data publicly available. So we hired an external consultant called David Roodman to, like, really kick the tires of it. He thought the results stood up. So my main priority now is asking myself to what extent would these [00:29:00] results generalize to other contexts in Africa or other contexts where GiveDirectly works.

The context of the trial was in western Kenya, and it wasn't that rural. Like, it was fairly near a major highway corridor which connects Uganda and Kenya. It also wasn't that agricultural. Like lots of people there had small businesses in, um, like retail and services and food manufacturing, for example. If you read that paper, it had a lot of these conditions that made it especially primed for positive general equilibrium effects.

There was a lot of underutilized capacity in the economy, lots of machines lying around not being fully utilized, lots of excess labor in the manufacturing sector. Under those conditions, our economic theory says that, like, increases in output without inflation are very possible. The big question I have now is, like, how generalizable are those conditions to other parts of Africa?

There are two things we're looking at now to try and build that evidence. One is we're supporting a trial in Malawi, which is trying to assess the effects cash transfers have on local consumption [00:30:00] and prices in a more remote context, and one that's more agricultural as well.

So that's just testing like a very different, yeah, macroeconomic or local economic environment. We're also in very early discussion with GiveDirectly about whether we should fund a third trial as well. We're scoping this at the moment, but one shape that could take is specifically varying one of the drivers that ex-ante we think could drive these spillover effects.

So one of them is, like, market connectedness, for example. For this demand side story to hold, people need to be able to spend money in each other's businesses. Therefore, market connectedness should be like a mediator of these effects. Can we run a trial where that's deliberately varied and see what effect that has on these, like, general equilibrium effects and spillovers?

Elie Hassenfeld: So that makes sense that there's this finding from Kenya, and the question is how does it generalize to other locations. And then how are you thinking about this other question of, you know, level of, of scale of GiveDirectly giving? In a world where, you know, instead of giving $1,000 to some small percentage of [00:31:00] households in a location, they're giving that much to a much larger portion of households. How do you think about that question and what are you trying to do about it, if anything?

Adam Salisbury: Yeah. So we're having active discussions with both GiveDirectly and Coefficient Giving about that at the moment. So when GiveDirectly gives cash transfers to people, they have a very intense recruitment process.

So they will physically visit these villages, they will manually enroll people, and they will check people's identification cards to make sure that the person they want to give cash to is indeed that person. That's great because I think fraud rates are very low. Like, they publish these on the website, but it's, like, well below 1%.

One question I have is how scalable is that model? Like, it's not trivial to, like, massively scale that up because you need to hire people, you need to, like, physically get to these villages, which is, yeah, not a trivial problem to solve. One question we're discussing is like, is there a more scalable recruitment model?

Like, are there ways we could use digital tools to enroll people remotely? And like, what would the trade-offs be if we did that? Like, would there be, like, an increased risk of fraud? Would there be an increased risk of exclusion where, like, legitimate people are just excluded because they don't have phones or, [00:32:00] yeah, there's a host of considerations, but those are active discussions.

Elie Hassenfeld: Right. And I guess there's other challenges that come with scale, whether it's additional political risk or even changes in our estimate of the inflationary effect of more cash that all become material at, at larger scale.

Adam Salisbury: Yeah, I think that's right. And the Malawi study is trying to get at that. But one of the things I'm working on with them is are there ways we can get quicker feedback loops? So randomized trials are the gold standard of getting rigorous causal estimates. They can take many years to run.

Part of what I've been discussing with them is, like, are there tighter feedback loops we could do? So rather than have a control group, do things like AB testing and then try and use remote sensing, so like mobile wallet transaction data or maybe satellite imagery to try and pick up these yield effects or consumption effects from, from afar.

Elie Hassenfeld: Right. So I know another area where you're interested and provided some support is in education. So maybe just talk a little bit about education, what the big question for you is there and what the research is intended to help decide. [00:33:00]

Adam Salisbury: Yeah. I think we've got maybe two or three big questions.

Like, one is just what are the returns to education? So, if you improve people's skills in childhood, how does that translate to later life wages, but also, like, quality of life and other outcomes? So one interesting thing is that randomized trials became really popular in development economics about 20 years ago, and there were lots of RCTs that were run on education, so children that were about 7 to 11 years old at that time.

Fifteen, twenty years later, a lot of those children are now young adults and they've entered the labor market. They've made potential migration decisions and marriage decisions. We've funded some research to try and track down as many of those kids as possible and try and measure, like, what the later life outcomes are.

Elie Hassenfeld: Because these are kids who already went through a randomized controlled trial. They had some education intervention. We have the two groups, and then the question is, can you find them 20 years later and determine what effect it had on their earning ability in early adulthood?

Adam Salisbury: Yeah, exactly. And, like, some of these interventions, it was things like teaching at the right level.

So you structure kids based on their abilities rather than their [00:34:00] age. They showed pretty promising effects in the short run, so increases in foundational numeracy, foundational literacy. The question I have is, like, does that translate into wage gains later or just, yeah, broad quality of life gains?

Like, are they more satisfied with their life? Do they have better health outcomes, et cetera? That's one big question. That's, I think it's going to take a while for those studies to come online. I think there's more we could do just with the information we have available now.

So the last time GiveWell really revisited education as a cause area was, I think, maybe five years ago. There's been more research that's been published since then. I would like us to take another look at it. At the moment, capacity is the main reason we're not, but the plan is not necessarily to wait until these long-run follow-ups are done before revisiting this.

Elie Hassenfeld: How do you think about the generalizability of those results to, you know, a future that could be different? And the future could be different on, on one hand, just what people are doing in the labor market could be very different. What education is necessary could be very different. But even more simply, in, you know, a country that could experience significant [00:35:00] economic change in the next 20 years. The results of, I don't know, seven-year-olds from the year 2002 to 2022 could be very different than seven-year-olds in 2030 to 2050.

I don't know. Is that an important question? How do you think about that kind of consideration in what you're supporting?

Adam Salisbury: It's definitely an important question. Like, if you think about the returns to education are ultimately prices, so they're dependent on the demand for skilled labor and the supply for skilled labor.

I don't think there's a strong reason to expect that to stay the same across time. I think it is a major question and it's a major limitation of RCTs. So I think RCTs are very strong on something that we call internal validity, so they can yield unbiased treatment effect estimates.

But they're, they can be weak on generalizability. Like, to what extent will this be relevant to country X in year Y? Another grant I've made is to researchers at the University of Oxford, which is using more up-to-date data sets and more comprehensive data sets. So we're trying to do it across every country in Africa.

Combine those with non-experimental techniques to try and estimate the labor market returns to education. [00:36:00] I think those studies are less strong on the internal validity front. They're more likely to be biased because we don't have randomized treatment. But external validity, they score, they score more strongly at.

So I think, like, those two investments complement each other quite nicely. But I think, I don't know, just we're gonna have massive questions even after both these studies are completed.

Elie Hassenfeld: Right. Okay, we're talking a lot about what we've done. We talked about GiveDirectly, poverty graduation programs, research that you're supporting.

I want to talk a little bit about the areas that you haven't prioritized so far, but people may know about. And so these are programs like microfinance, agricultural support, and others. Let's just start with microfinance. First just say what it is and then how you're thinking about it relative to the other programs you're prioritizing.

Adam Salisbury: Sure. So microfinance is giving small loans to people and businesses which are typically frozen out of the traditional financial system. So they're too poor, they don't have enough collateral for formal banks to offer them loans. It was developed in the '80s—'70s and '80s—and the original versions of it were premised on this group-based liability model where groups of [00:37:00] people in rural communities were bandied together, and they'd be held jointly responsible for these loans.

And microfinance organizations would lend them money. They would invest it in whatever projects they thought they wanted to invest it in, and then they would have to repay the loan back to the microfinance institution. There was a wave of randomized trials in the Northeast which tested what are the effects of microfinance on things like income, consumption, business profits. The general impression I have of those early RCTs was results were generally quite underwhelming.

So if you look at the average treatment effects, the effects on business profits was quite small. The effects on income and consumption was quite small. There's been some more recent research which I think paints more nuance to that picture and does lend itself to some open questions for me that I'm hoping we'll get to later in the year.

The first nuance is that underneath that aggregate result, there's a lot of what we call heterogeneity, so variation in effects. So there's a whole mass of people where microfinance doesn't look like it's [00:38:00] producing transformational effects or like it's probably helping them solve cash flow problems, but it's not lending itself to productive investments.

But there is a small subset of entrepreneurs that it looks like can use microfinance really productively. They can invest in productive businesses, grow their businesses, hire more people. So a question that raises for me is like, can we target those right tail of, they're called gung-ho entrepreneurs in the literature?

The second strand is there's been just newer products that appear to have initially more promising results. So these newer products emphasize more flexible loan repayments. So I think an issue identified in the first wave is that people weren't using the loans for productive investment.

A lot of it was for, like, consumption smoothing, uh, and partly because there were very strict repayment schedules, so people didn't want to take risks on, like, large investment products. The two newer products I'm aware of, one makes the repayment schedule more flexible. The other specifically ties the loan to a productive asset, um, so a rickshaw or a sewing machine.

There's been some RCTs of those type of models, and those show initially more promising results on [00:39:00] average. The big question I have here, and this is part of what I focus on, is, um, sort of why is philanthropy needed? So I should mention that there was very high repayment rates in these RCTs as well.

If repayment rates are above 95 percent and people can use these loans productively, like, why don't microfinance institutions innovate themselves? Like, why don't they commercialize these products? I think there could be good reasons for that to do with transaction costs. Like, it can be really costly to transact with people. But, that's the question I would dig into if and when we get to this.

Elie Hassenfeld: Got it. And then how about agricultural-oriented work, which is another major area of focus for funders?

Adam Salisbury: Yeah, I think part of the reason for that is the majority of people that are in extreme poverty are subsistence farmers.

So this is just why this is a very, both an area GiveWell cares about but the other development funders care about as well. If we could find something that could reliably improve smallholder production or consumption by 10 percent, that would be, like, a massive deal in welfare terms. I would say if I look at the really long historic track record, [00:40:00] we haven't found what that is yet.

There's been lots of attempts. So there were the Millennium Village programs in the noughties, which I've touched on briefly. There was the Alliance for the Green Revolution in Africa. Generally speaking, those programs seem to just fall well far short of their objectives. So it's a bit of a, I wouldn't call it a graveyard of, like, failed opportunities, but there's a lot of stuff which doesn't seem to have worked very well.

The more positive spin is that there are some recent trials of newer programs which do show much more promising effects. So one of those that I'm looking into right now is called Raising The Village. These guys provide free high-yielding seeds, agronomy training. They set up savings facilitation vehicles similar to poverty graduation programs, and they provide this package of interventions for everyone in the village at the same time.

There was a randomized trial of their program, which is a working paper at the moment, but it shows pretty encouraging results, so 10% consumption gains three years later. A big question for me now is how would those results generalize. And more precisely, the original trial was [00:41:00] done during the COVID pandemic, so it was from 2020 to 2022.

I think there are several reasons why effects could be different in a more business-as-usual setting. And then the other question I have is have they been able to maintain quality at scale? I think things like farmer training, I don't think it makes sense to think of as a homogenous category, because I think the devil is in the detail.

Like, you can have some really good farmer training and really bad farmer training. I also think historically, that type of intervention, which is very, like, labor-intensive, it requires, like, hiring good people, it requires training them well and monitoring them well, that's very, very hard to scale. I'm curious whether RTV has sort of cracked that problem. So we're scoping new research projects with them at the moment to try and shed light on both of these questions.

Elie Hassenfeld: And so do you think it's a fair summary to say that there are a lot of areas that we would be extremely interested in doing more on in the future to evaluate and investigate? Right now, the livelihoods team is extremely small, and so our aim is to grow that team, grow capacity, and we're hiring for a senior livelihoods researcher right now that can enable us to do [00:42:00] more.

And with increased capacity, we'd be in a position to cover much more ground, which is what we're intending to do in, as soon as we can.

Adam Salisbury: Yeah, that's right. So yeah, put a plug out there, we're actively hiring at the moment. I think my vision in the medium term is to start having other people own some of these verticals we've spoken about.

So education's an example where, like, that could easily be its own portfolio in its own right. Agriculture as well. So we're hiring at the moment. But lots of exciting research happening in the meantime as well.

Elie Hassenfeld: Okay. So Adam, let's just close with this question of, you know, right now everything we've talked about is focused on a certain scale of giving.

I think you expect to give away about $20 million in your portfolio in 2026. How does this program area scale? Do you think it can scale? What do you think in GiveWell terminology is the room for more funding?
You know, how much you could imagine absorbing and what you might be supporting at, at larger scale. How have you thought about that? How much have you thought about that question?

Adam Salisbury: So this isn't something I thought about a ton until about two months ago, but it's been much more top of mind for me.

If I had to scale the portfolio [00:43:00] to, like, a substantial degree, I think cash transfers and poverty graduation would be my two most solid bets at the moment, in terms of just, like, evidence-backed, scalable. With that being said, I've only been on the job about five months, so I think it's possible that there are whole new cause areas I've not looked at yet, which could absorb a lot of funding.

One example of that is humanitarian response, which is giving food or cash or commodities to people in the aftermath of a disaster or during a protracted conflict setting. We're doing, like, very high-level landscaping at the moment. I think it's possible this cause area sits more with nutrition at the moment because, I think the line between what's, like, a nutrition program and what's a livelihoods program gets especially blurred in humanitarian contexts.

But the reason we're looking at it is that there have been some pretty large funding cuts to the humanitarian space at the moment, and so there's just high need. What we're looking at is that, are there really impactful investment opportunities in this space that GiveWell could help fill?

Elie Hassenfeld: Well, great. Thanks so much, Adam. This was great, and appreciate, you know, everything that you shared and for taking the time to go through it with us.

Adam Salisbury: Cool. Thanks, Elie.

Elie Hassenfeld: [00:44:00] Hey everyone, it's Elie again. I hope you enjoyed what was really a whirlwind tour through this area that's of increased focus to us right now. You know, this is an area where, as you heard, we're really building up our knowledge and aiming to build our team so we can do even more.

We're hiring, and so if you're interested in the senior livelihoods researcher role, you can see that on our jobs page. I think one of the most important and interesting things to me here is that I think this area really illustrates how GiveWell is, you know, a relatively rare institution in our space because we're both a funder of programs, and in that we consume research to help us decide what to do, but now we're also in a position, due to the support of donors, to be a funder of research.

And so we are very much able to look at an area like livelihoods where there has been a lot of research, but then say what research questions are most material to our future decisions, and then provide the funding that will affect the decisions that we make in the future. And of course, that [00:45:00] research can influence our own funding, which we're very, you know, grateful is fairly substantial at this point.

But of course, it's still small relative to what some of the largest governments do in their poverty alleviation programs. And so, you know, I'm also very excited about this work Adam described, where we're considering ways to support more effective government delivery of poverty alleviation programs. And, you know, it's one more way in which GiveWell's work, our funding, and the research that we support can extend our impact even further.

As always, thank you so much for your support and for your interest in our work. We appreciate it.