William MacAskill on Effective Altruism and Doing Good Better
What this covers
Will MacAskill, author of Doing Good Better, speaks with Russ Roberts on economics and charity. MacAskill argues that donors applying scientific scrutiny and expected-value reasoning can achieve roughly 100 times the impact per dollar compared to typical giving. He grounds this in two observations: the relationship between income and wellbeing follows a logarithmic curve, so money does far more good for the global poor than for the wealthy, and the charity sector lacks the competition that makes markets efficient—unglamorous interventions remain underfunded while donors often back well-intentioned projects that fail in practice. Roberts accepts the core mathematics but questions whether real-world effectiveness estimates are as reliable as they appear.
The conversation covers the practical and philosophical machinery of effective altruism. MacAskill walks through five evaluation questions (impact, comparative advantage, neglectedness, counterfactual impact, and success probability) and explores how they reshape what counts as high-impact work—including the case for "earning to give," where someone with rare marketable skills deliberately pursues a lucrative career to donate. He illustrates the stakes with specific cases: the Playpump, a water-pumping device that absorbed millions despite failing in the field; deworming programs that cost 50 cents per child and increase school attendance and later earnings; and cash transfers that local knowledge allows recipients to deploy more effectively than external experts could. Roberts presses back hardest on the limits of marginal thinking and simplified interventions. He argues that wellbeing depends on complex, interdependent systems—the right political regime, functioning markets, cultural conditions—that resist the kind of isolated lever-pulling that charity scorecards reward. The two also debate whether cash transfers overcome poverty traps or merely postpone deeper constraints, and whether private philanthropy is a useful complement to or problematic substitute for political action.
MacAskill argues that giving to fight global poverty can do enormously more good per dollar than typical charity if donors apply scientific scrutiny, marginal/counterfactual thinking, and expected-value reasoning; Roberts accepts the core logic but presses that complexity and local knowledge make effectiveness estimates far less reliable than they appear.
- Diminishing returns plus the log relationship between income and wellbeing imply benefiting the global poor can be ~100x more impactful than spending on oneself
- Good intentions are insufficient (Playpump) while unglamorous interventions (deworming) can be highly cost-effective
- Counterfactual and marginal reasoning change what counts as high-impact careers and donations (earning to give, the 100x multiplier, drop-in-bucket framing)
The best programs vastly outperform merely good ones, making careful selection crucial.
- Because the distribution of impact across social programs is highly skewed, the very best programs are far, far better than even merely very good ones, so it is crucial to do the most good rather than merely some good.
“if you look at the distribution of impact among different sorts of social programs, you find the fairly best social programs are far, far better than even merely very good ones”
- Unlike the stock market, where you can't pick winners because so many participants already arbitrage prices to reflect value, the charity world has no comparable efficient market—relatively few funders try to maximize good—so low-hanging fruit remains on the ground and donors can find unusually effective opportunities, even if the very best causes (e.g., polio vaccination) are already funded.
“when it comes to charity or doing good, there just isn't really the same sort of efficient market”
- When initial results suggest something is exceptionally good, you should statistically expect the effect to regress toward the mean as further evidence accumulates; the downward revision of deworming's impact in reanalysis is exactly what one should expect and does not undermine the case.
“If you have these, like, initial results that something is just exceptionally good, then you should expect to regress back towards the mean, in light of further evidence.”
- The US government will fund a program if it saves a life for about $7 million, whereas distributing bed nets through the Against Malaria Foundation can save a life in very poor countries for roughly a few thousand dollars, making global health interventions vastly more cost-effective per life saved.
“the government will fund a program if it saves a life for about $7 million. If you want to save a life in very poor countries it will cost you about a half-thousand dollars”
- Charity focused on domestic issues like US education or domestic poverty neglects that one can do far more for the very poor in poor countries, because they have so much less and the lowest-hanging improvements (bed nets, deworming) have not yet been taken.
“you can do far, far more to benefit the very poor in poor countries, simply because they have so much less than the lowest hanging fruit, in terms of improvements to people's lives till haven't been taken”
- Effective altruism asks five questions to do the most good: (1) what is the impact on wellbeing—how many benefit and by how much; (2) is this the most effective option, since the best programs are far better than merely good ones; (3) is the area neglected, invoking diminishing returns; (4) what would have happened otherwise (counterfactual impact); and (5) what are the chances of success and how good would success be (expected value).
“effective altruism is about using your time and money as effectively as possible to make the world a better place. And taking a scientific approach to doing so.”
Simple direct interventions like cash transfers, bed nets, and deworming are exceptionally cost-effective and scalable.
- Cash transfers can have lasting compounding effects rather than just temporary consumption boosts, because there is little evidence for a 'poverty trap' in which earning more fails to lift people out; the poor have genuine opportunities to improve their income through investment.
“there doesn't seem to be much evidence that there is such a thing as a poverty trap, in the way you'd expect”
- Because there is enormous local knowledge that outside helpers lack, giving cash to the very poor lets the normal operation of markets work: recipients can spend in whatever way is best for them, organically solving the problem of fixing only one part of the system at a time that plagues in-kind interventions like textbooks.
“There's a failure of knowledge on the part of people trying to help. Whereas if you are just giving cash to the very poor people, they have that local knowledge”
- Intestinal worms affect over a billion people; deworming costs about 50 cents per child per year with no side effects, and a landmark study found it to be one of the most cost-effective education programs because sick, lethargic children attended school more once treated, with follow-up 12 years later showing dewormed adults worked more hours and earned more.
“you can treat these worms for only about 50 cents per child... will cure them of worms for one year”
- GiveDirectly transfers cash to the poorest households (about $1,000 doubles a household's income for a year); recipients spend the majority on investment rather than consumption—commonly tin roofs that replace recurring thatch costs and yield about a 14% annual return, plus livestock—producing long-lasting compounding gains rather than a temporary boost.
“if they buy a tin roof, they effectively get a return on their investment of about 14% a year. So, way better than anything we can get by putting it in the bank”
- Whereas funding room for bed nets and deworming would be exhausted quickly (perhaps after a few billion to tens of billions of dollars, with everyone dewormed and under a net), cash transfers have enormous room to scale up almost indefinitely, which is what makes GiveDirectly especially exciting for transforming international development.
“With cash transfers you could scale that up massively, almost indefinitely”
Well-intentioned programs routinely fail because they ignore local knowledge and unintended consequences.
- The Playpump—a children's merry-go-round designed to pump water—attracted huge media praise, celebrity backing, and millions in USAID and foundation funding, but failed in practice: it required constant force, tired and injured children, was left to elderly women, broke down without maintenance, cost three times a hand pump while pumping less water, and was never requested by communities.
“unlike a normal merry-go-round which spins freely once you give it sufficient momentum, this one required constant force in order to pump clean water from the ground”
- Wanting to do good is insufficient; you must be careful and reflective about what you are actually doing if you want to make a big difference, because well-intentioned interventions can fail badly.
“It shows that good intentions aren't good enough.”
- When ICS's education programs were tested component-by-component via randomized control trials, interventions that seemed obviously beneficial—distributing textbooks, flip charts, and reducing class sizes—showed no discernible impact on educational outcomes.
“providing a textbook or providing more books would improve learning among students. But they tested it; they found no impact, in fact.”
- The 'teach a man to fish' slogan is as problematic as giving fish, illustrated by a nomadic tribe settled on a lake and taught to fish who then overfished and depleted the stock as the lake dried up; giving cash would have let them sell or use resources as best suited them via their local knowledge.
“Taught them to fish. And then there was overfishing and the fish were depleted.”
- The leftist objection that private giving crowds out superior political solutions assumes political change is easy, which is mistaken; unintended consequences are far greater for political change than for simple interventions like deworming, as the disastrous history of Communism shows that pursuing a compelling political vision can have catastrophic results.
“if you think of that as easy then you are definitely mistaken”
Global poverty differs fundamentally from domestic poverty in its scale and tractability through market-based development.
- Extreme poverty is fundamentally different from domestic poverty: the very poor abroad lack productive opportunities not because of mental illness or substance abuse (as can be the case domestically), but simply because they start with very little income and therefore have enormous opportunities to invest additional income and improve their lives.
“extreme poverty is just very different than domestic poverty”
- Much left-wing criticism of EA rests on an empirical disagreement about capitalism: critics see the economy as a fixed pie in which rich countries are rich by taking from poor countries, whereas in reality everyone started poor and some countries achieved miraculous economic success (China, India, South Korea, Taiwan) through the same market-based development, so trade like sweatshops benefits both parties rather than exploiting one.
“they see things as a fixed pie, where you've got the rich countries just being rich by taking from poor countries. And I just think that misrepresents the situation”
- For most of human history everyone lived on less than $2 a day; only in the last few hundred years have some countries escaped that condition, so the existence of prosperous mixed economies like the US and UK is essentially a miracle achieved by ordinary people, and many countries have since escaped global poverty through the same market-based development.
“The state of the world, for most of human history, is everyone living on less than $2 a day. And then it's only in the last few hundred years that some countries have been able to escape that.”
- Although sweatshops have horrific conditions, boycotting them is a mistake because for the extremely poor these jobs are actually desirable relative to the alternatives—unemployment and starvation, prostitution, back-breaking poorly paid farm labor, or scavenging in dumps—evidenced by people choosing and even migrating illegally to take them; the right response is to engage in trade to end the underlying poverty, since boycotting is counterproductive.
“these sweatshop jobs are actually really desirable places to work... compared to the alternatives”
Individual counterfactual impact matters more than absolute impact when choosing how to do good.
- The good you do as a doctor is not the lives you save but the difference between the world with you as a doctor and the world where someone else takes your place doing similar work; because a replacement would perform similar life-saving care, your true counterfactual impact is much smaller than it appears.
“the impact you are having as a doctor is just the difference between the way the world is given that you've become a doctor, and what it would have been like had someone else been in your shoes”
- Because wellbeing varies roughly with the log of income (any percentage increase in income yields the same wellbeing gain), and a typical American is about 40 times richer than the poorest billion while money goes about 2.5 times further overseas, halving your income could double the income of about 100 very poor people, meaning you can benefit others roughly 100 times as much as you can benefit yourself.
“if I were to just halve my income, I could double the income of a hundred people in very poor countries”
- For perhaps 15% of people, the best way to do good is 'earning to give'—deliberately taking a lucrative career to donate—because of comparative advantage (some have valuable but charity-irrelevant skills), the ability to target the most effective charities, money's flexibility to switch as evidence changes (unlike labor), and the value of building business skills before moving into the nonprofit sector.
“earning to give is the idea of deliberately taking a lucrative career in order that you can do good through your ability to donate rather than through your direct contribution of your labor”
- When deciding how much to give, you should ask what is the best thing to do given how everyone else is behaving, not 'what if everyone did this'; the fact that an action won't solve the whole problem is irrelevant—what matters is not the size of the bucket but how big a drop you add, and saving hundreds of lives over a lifetime is a massive drop.
“It's not the size of the bucket that matters. It's how big a drop in that bucket you are making.”
Aid spending has demonstrably improved global health and poverty outcomes despite structural constraints.
- Even assuming aid did no good at all except eradicating smallpox—a disease that killed 300 million people and whose eradication in 1973 saved 60-120 million lives (more than world peace would have)—dividing total aid spending by lives saved yields a cost per life of about $70,000, roughly 100 times better than US willingness to pay.
“supposing aid did no good at all, except insofar as it eradicated smallpox, a disease that killed 300 million people before we eradicated it in 1973”
- Over the last 50-60 years aid has cost comparatively little—about $20 per person per year—while life expectancy rose by a third and child mortality fell dramatically, and the share of people in extreme poverty is half what it was a few decades ago, giving grounds for optimism even though aid cannot claim credit for most of this progress.
“we've spent what's comparatively a tiny amount of money, like $20 per person, per year... And life expectancy has increased by a third.”
- Global health is so clearly effective that even the chief aid skeptic, Bill Easterly, who otherwise lambasts aid in The White Man's Burden, concedes that global health is awesome and has changed millions of lives for the better, with documented progress on measles, polio, malaria, and tuberculosis attributable in part to public funding like the WHO and the Gates Foundation.
“even chief among the aid skeptics, Bill Easterly... spends a couple of pages on global health; and he's like, 'Yeah, yeah, global health is awesome'”
Earning to give and platform advocacy are legitimate paths to impact beyond direct charity work.
- Doing good directly is only one path; one's ability to advocate from a position of standing is another, illustrated by Martin Luther King who did not work for a charity but used his platform to advance an enormously important cause—suggesting Roberts could do good by using EconTalk as a platform to promote important ideas.
“Martin Luther King didn't work for charity but he had standing that led him to advocate for this incredibly important cause”
- The simplest way to get into the habit of doing good is to start donating regularly—the amount initially matters little—since it builds the habit of making a big impact, after which one can scale up to a 10% commitment via 'try-out giving' that gradually increases.
“the simplest thing to get yourself in the mood of doing good I think is to start donating regularly. To begin with it doesn't really matter how much that is. But it gets you in the habit”
Evaluation requires judgment calls and sanity checks alongside quantitative data and RCT evidence.
- Beyond measurable RCT-backed interventions, you can do good by pursuing speculative, hard-to-quantify activities like systemic policy change; by estimating how good success would be and multiplying by even a low probability of success, the expected value can sometimes exceed that of more certain but smaller-magnitude benefits.
“you multiply that low probability by that very large amount of value, and at least sometimes that means it will be even better to pursue this more uncertain path over something that would be more concrete and more certain”
- Evaluating development programs is not a hard science like physics; even with evidence, judgment calls are unavoidable, so one should use sanity checks alongside data when comparing interventions like deworming versus bed nets.
“this isn't a hard science like physics. It does mean that there's going to have to be judgment calls along the way.”
- The Google-hosted effective altruism conference drew about 450 people, more than three times the attendance of the prior year's event, indicating rapid growth in the movement's reach.
“About 450 people were at that conference.”
- If you saw a chance to run into a burning building to save a child you'd feel like a hero; by using your time and money wisely you have the opportunity to save lives roughly every year or few years, an exciting opportunity to have a huge impact.
“If you were someone who could run into a burning building and save the child's life, you'd feel like a hero... You have the opportunity to do that every year, every few years”