Chris Blattman on Chickens, Cash, and Development Economics
What this covers
Chris Blattman and Russ Roberts use a simple comparison—cash versus chickens—to map a larger terrain in development economics. The conversation pivots on a concrete observation: aid agencies spend billions annually on interventions measured only by impact, never by cost, so even successful programs can be wasteful relative to cheaper alternatives. Blattman argues for structured trial and error at scale, running small experiments to swap low-return programs for better ones. But he also stakes out a provocative claim: that poverty alleviation itself, while defensible on moral grounds, matters less in the long run than a problem nobody has solved—building functioning state capabilities and human capital in the fragile states that will define global instability in 15 to 20 years.
The episode covers three distinct registers. First, micro-economics: capital constraints as the coherent theory of why people stay poor, and in-kind transfer programs like chickens, which take 15–20 years to pay for themselves despite measurable impact. Second, methodology: why randomized control trial evidence converges on poverty interventions but fractures on violence and growth, and how external validity fails when a landmark study on deworming relied on the unusual parasite burden of a single Kenyan lakeshore. Third, political economy: the tension between helping the visibly poor now and investing in growth that might help billions later; general equilibrium effects that undermine the assumption everyone can earn the same way; and the claim that aid workers' natural empathy leads them to over-invest per beneficiary, driven by seen vs unseen bias. The conversation also flags Marginalism as the key lever—if billions are already committed to aid, even small efficiency gains compound across interventions.
Blattman argues that because aid agencies are already spending billions giving poor people 'stuff,' the highest-leverage move is to run structured trial-and-error experiments (like a cash-vs-chickens horse race) to swap costly, low-return interventions for cheaper, better ones — while conceding that poverty alleviation matters less in the long run than the unsolved problem of growth and state capability in fragile states.
- Most aid programs measure impact (numerator) but ignore delivery cost (denominator), so costly in-kind transfers like chickens can take 15-20 years to pay back.
- Micro-evidence converges on a coherent theory of poverty as capital/credit constraints, making cash transfers a reliable lever, unlike growth or violence where evidence has 'cumulated into a mess.'
- Fragile, non-growing states will be the central security and development problem in 15-20 years, so understanding state capability matters more than incremental poverty interventions.
Scaling interventions without testing assumes uniform returns that market dynamics and local constraints will destroy.
- Bill Gates floated the idea that chickens are the future for Africa, envisioning roughly 30% of Africans (about 300 million people) raising chickens to escape poverty, up from the current ~5% (about 15 million people).
“he envisioned perhaps 30% of Africans. So, this would be 300 million people raising these chickens rather than the existing number, which is maybe 5% of Africans--so, maybe 15 million people”
- If a third of Africans suddenly started producing chickens and eggs, the prices of chickens and eggs would fall fast, and there are limits to how many people can consume, undermining the assumption that everyone could earn income from the same intervention.
“If a third of Africans start producing chickens and eggs, that the price of chickens and eggs are going to fall pretty fast. And there's probably limits to how many chickens and eggs people can eat.”
- Because rich and poor countries already spend on the order of $10 to $100 billion a year giving the poorest people stuff, spending roughly $15 million on a large multi-country experiment comparing interventions like cash and chickens could let aid be twice as effective by killing bad ideas and replacing them with less-bad ones on the margin.
“if we could spend $10 million dollars just to, like, tweak the direction of that, to sort of kill a bad idea and replace it with a less bad idea'--that's kind of what I want to see.”
- Rather than launching a single highly specific large program over 5 years and inevitably redesigning it midway after wasting money, governments and aid agencies should first try 5 or 10 things on a small scale, observe which succeed, ideally study them rigorously, and then scale the most successful — introducing the trial-and-error discipline that randomized testing alone has failed to bring to aid.
“Why don't you do 5 or 10 things on a small scale for the first year? You have to scale up, you have to get moving.”
Most poverty reflects capital and credit constraints; relieving them works, but long-term impact and cost matter more than initial outcomes.
- Many people remain poor not because business opportunities are unavailable but because they lack capital (cash, productive assets, tools, animals) and lack access to borrowing; when one such constraint is relieved, people often leap ahead and start businesses, moving from $1/day to several dollars a day.
“the main thing stopping them from doing this is they don't have any capital. If they had capital, they wouldn't be poor.”
- Studies of randomly giving people chickens, cows, and goats find the interventions do pay off, but it takes something like 15 or 20 years before the cumulative impact covers the program's costs.
“the studies that have looked at chickens and giving people chickens and cows and goats randomly pay off, but it takes something like 15 or 20 years before they cover the costs.”
- Most aid programs evaluate only the impact of an intervention (the numerator) while ignoring its delivery cost (the denominator); in-kind programs like chickens require expensive procurement, transport, and trainers to remote areas, so even if a chicken program has greater impact than cash, its much higher cost can make it a far worse return.
“everyone thinks about the numerator: What's the impact of this program? And nobody thinks about the denominator, which is: What's the cost of providing this program?”
- In large cash-transfer experiments, at least half of recipients do not move ahead — they start a small enterprise that fails — which is simply the nature of business and makes it hard to know whether on balance people are succeeding.
“the big cash experiments I've done, others have done--at least half the people don't really move ahead as a result of this cash. They start a small enterprise and it fails. This is what business is.”
Sustained fieldwork and local relationships produce irreplaceable understanding of politics and organizational realities that shapes effective aid strategy.
- Answering important questions about other parts of the world — whether big or small — requires spending a lot of time in those places, not just collecting data but building relationships with local academics and policymakers, because knowledge production is not an individual production function.
“to answer important questions about other parts of the world, you have to spend a lot of time in other parts of the world... it's not an individual production function.”
- Deep specialization — where only one or two people in the world really know what you know — lets an academic go almost anywhere and be useful by bringing a specific expertise or outside perspective, but this works best where local human capital and policy-intellectual peers exist (Colombia, Ethiopia) and is most tragic where broken systems mean almost no local has had the opportunity to develop comparable expertise (Liberia, northern Uganda).
“there's really just one or two people in the world who really knows what we know. And so I can go anywhere in the world and say something and be useful, because I've specialized so much.”
- The most important product of randomized trials and prolonged fieldwork is not the causal estimates themselves but the rich, sophisticated understanding of local politics, aid-sector dysfunction, and social and organizational realities that economists gained by spending years in places like Kenya and Ghana — knowledge that now shapes the theories, teaching, questions, and advice of the profession.
“the most important thing about randomized control trials is not the causal effect that lots of people, we've identified.”
- Intervening to fix other people's lives — whether parenting or development — carries an implicit arrogance, so it must be approached with humility and humbleness rather than the hubris usually seen, by spending a lot of time deeply understanding just a few places rather than spreading thin.
“I think you have to avoid being arrogant... I think humility and a degree of humbleness rather than the hubris you usually see is really fundamental.”
Growth alone cannot answer development's hardest questions; state capability and political change deserve more focus than economists typically give.
- Helping an African nation go from roughly $1,500 to $3,000 per capita is a comparatively tractable problem, but getting a nation to ~$20,000 per capita (middle-income status) is one nobody has a good answer to — so debates about development often conflate two very different problems.
“how do we help an African nation go from $1500 a head to $3000 a head? That's not necessarily a hard problem... how does that nation... get to $20,--- a head? Some sort of like middle-income status. And then nobody has a good answer to that.”
- Cheap data and computing tempt economists toward answerable small questions that quickly yield publishable articles, producing a misallocation of talent away from the big, hard questions about growth and state capability where more brains and money are most needed.
“there are these bright, shiny [?]s that come along; it's very appealing to get an answer that a lot of people--there's all this data and computer technology that lets us do, answer a lot of small questions”
- Growth is the wrong framing for the big development question; instead we need more people focused on understanding state capabilities (making states more effective), democratization, and politics in poor countries — a body of political-science research that, if translated into accessible terms, could be highly impactful.
“Growth is the wrong way to think about this. We don't need more people focused on economic growth. I think we need more people focused on understanding state capabilities, and democratization, and politics in these countries.”
- The micro-evidence on poverty has accumulated into a coherent, broadly consistent understanding (people are capital/credit constrained, relieving constraints helps), whereas other areas like the causes of violence or the macro effect of aid have 'cumulated into a mess' with no coherent answer — which is why Blattman is confident giving policy advice on poverty but not on reducing violence.
“we don't really have a good understanding of what reduces violence. Like the things have not accumulated to a coherent answer.”
Alleviating present poverty remains morally defensible despite benefiting future generations more, and fragile states pose lasting global security risks.
- Because most middle-income countries (China, Brazil, Russia, Vietnam) are growing toward rich-country status, in 15-20 years a residual group of about 20-30 enormously poor, unstable 'fragile states' — much of sub-Saharan Africa, parts of Central/South America and Central Asia — will sit beside a homogenous rich world, generating growing inequality, instability, and negative spillovers (piracy, extremist groups), making these states the fundamental global security and development problem.
“in 15 or 20 years there will be about 20 or 30 countries in the world that are still enormously poor and unstable, next to what are generally a relatively homogenous group of middle- and high-income countries.”
- The development community's focus on an extreme-poverty threshold of $1 or $2 a day is an artificial line that distorts a lot of policy, since the merely very poor and somewhat poor share the same capital constraints and could benefit from the same interventions.
“there's this artificial threshold of $1, $2 a day that distorts a lot of policy.”
- Even granting that focusing on growth would make future generations much better off, today's horribly poor people still exist, and depending on one's moral calculus — e.g., that we are only as good as the least among us — it remains defensible and worthwhile to spend substantial effort alleviating present poverty rather than sacrificing the current generation for future ones.
“if you think... that we're only as good as, say, the least among us; or that we have a responsibility to help the very, very least among us even if that means we wealthy people or future wealthy people who are not yet born will be substantially worse off--that's also a defensible claim.”
- The large measured effect of deworming came from a single trial on the shores of Lake Victoria in Kenya — the birthplace of humanity and of human parasites — so the unusually high parasite burden there likely inflated the impact, and the scarcity of good long-term economic-outcome trials elsewhere means deworming's effect may not generalize.
“it was on the shores of Lake Victoria in Kenya, which is the birthplace of humanity, and then not coincidentally the birthplace of human parasites. So, an impact of deworming medicine there is going to be not surprisingly quite impactful”
- Bill Gates follows only about 200-300 people on Twitter, several of them development figures, including Blattman, who tried direct-messaging him after the open letter but received no response.
“it turns out he only follows, like, 300 people; and a number of them are development people... So I even direct-messaged him on Twitter--politely... And then: Crickets.”
Aid workers' psychological bias toward visible beneficiaries drives inflated per-person spending that reduces marginal effectiveness.
- Aid workers' natural human instinct to minimize the visible failure of people in their care — people they see, care about, and feel responsible for, while ignoring the unseen people they could otherwise help — drives them to over-invest per beneficiary, doubling or tripling program costs not to make people more successful but merely to reduce their risk of failure.
“you don't see all these other people you are not helping. So, doubling or tripling or quadrupling or even further increasing the cost of a program--not to make them dramatically more successful but just to reduce their costs of failure--is really natural human instinct.”