A commonly voiced sentiment in the effective altruist community is that the best way to do the most good is generally to make as much money as possible, with a view toward donating to the most cost-effective charities. This is often referred to as “earning to give.” In the article To save the world, don’t get a job at a charity; go work on Wall Street William MacAskill wrote:
Top undergraduates who want to “make a difference” are encouraged to forgo the allure of Wall Street and work in the charity sector ... while researching ethical career choice, I concluded that it’s in fact better to earn a lot of money and donate a good chunk of it to the most cost-effective charities, a path that I call “earning to give.” ... In general, the charitable sector is people-rich but money-poor. Adding another person to the labor pool just isn’t as valuable as providing more money, so that more workers can be hired.
In private correspondence, MacAskill clarified that he wasn’t arguing that “earning to give” is the best way to do good, only that it’s often better than working at a given nonprofit. In a recent comment MacAskill wrote
I think there's too much emphasis on “earning to give” as the *best* option rather than as the *baseline* option
and raises a number of counter-considerations against “earning to give.” Despite this, the idea that “earning to give” is optimal has caught on in the effective altruist community, and so it’s important to discuss it.
Over the past three years, I myself have shifted from the position that “earning to give” is philanthropically optimal, to the position that it’s generally the case that one can do more good by choosing a career with high direct social value than by choosing a lucrative career with a view toward donating as much as possible.
In this post I’ll outline some arguments in favor of this view.
Responses to MacAskill’s Considerations
In the article To save the world, don’t get a job at a charity; go work on Wall Street, MacAskill gives three considerations in favor of “earning to give.” I respond to these considerations below. What I write should be read as a response to the article, rather than to MacAskill’s views.
Variance in cost-effectiveness of charities
MacAskill wrote
… charities vary tremendously in the amount of good they do with the money they receive. For example, it costs about $40,000 to train and provide a guide dog for one person, but it costs less than $25 to cure one person of sight-destroying trachoma. For the cost of improving the life of one person with blindness, you can cure 1,000 people of it…it’s unlikely that you can work for only the very best charities. In contrast, if you earn to give, you can donate anywhere, preferably to the most cost-effective charities, and change your donations as often as you like.
GiveWell has spent about five years looking for the best giving opportunities in global health, and its current #1 ranked charity is Against Malaria Foundation (AMF). GiveWell estimates that AMF saves an infant’s life for ~ $2,300, not counting other benefits. These other benefits not withstanding, AMF’s cost per DALY saved is much higher than the implied cost per DALY saved associated with the figure cited for curing sight-destroying trachoma.
GiveWell may have missed giving opportunities in global health that are much more cost-effective than AMF is, but given the amount of time, energy and attention that GiveWell spent on its search, one should have a strong prior against the possibility that one can easily find a better giving opportunity in global health. So a plausible estimate of the cost-effectiveness of donating to the best charity that delivers direct global health interventions is much lower than the above quotation suggests.
Furthermore, the phenomenon of the optimizer’s curse suggests that all charities with robust case for fairly high cost-effectiveness are closer in cost-effectiveness to AMF than explicit cost-effectiveness calculations indicate. This narrows the variance in cost-effectiveness amongst charities.
So the advantage of being able to choose a charity to support and change at any time is smaller than the above quotation suggests.
Discrepancy in earnings
MacAskill wrote:
Annual salaries in banking or investment start at $80,000 and grow to over $500,000 if you do well. A lifetime salary of over $10 million is typical. Careers in nonprofits start at about $40,000, and don’t typically exceed $100,000, even for executive directors ... By entering finance and donating 50% of your lifetime earnings, you could pay for two nonprofit workers in your place—while still living on double what you would have if you’d chosen that route.
The assumption “if you do well” is a very strong one. Only about 1% of Americans make ~$500k/year. There are some people who have a strong comparative advantage in finance, for whom “earning to give” to give may be especially compelling. But people who are able to make ~$500k/year in finance who don’t have a large comparative advantage in finance have very strong transferable skills. Such people are significantly more capable than the average non-profit worker, and can plausibly have a bigger impact than 2 or 3 such workers by working directly on something with high social value.
Replaceability
MacAskill wrote:
…“making a difference” requires doing something that wouldn’t have happened anyway…The competition for not-for-profit jobs is fierce, and if someone else takes the job instead of you, he or she likely won’t be much worse at it than you would have been. So the difference you make by taking the job is only the difference between the good you would do, and the good that the other person would have done.
I would guess that there are some highly cost-effective humanitarian interventions that are sufficiently easy to implement that the implementers are easily replaceable. I could easily imagine that this is the case for vaccination efforts.
But funding opportunities for these interventions can be thought of as “low hanging fruit.” Broad market efficiency suggests that such interventions will be funded. And indeed, GiveWell has found that straightforward immunization efforts are already largely funded, to the point that GiveWell has been unable to find giving opportunities for individual donors in this area.
This suggests that at the margin, very high value humanitarian efforts require highly skilled and highly motivated laborers.
High skilled laborers are a relatively small subset of laborers, so there are fewer people available to do these sorts of jobs than other jobs. Doing a hard, non-routine job well requires high motivation. The collection of people who are sufficiently highly motivated to do a hard job with high social value that doesn’t pay well, and who could otherwise be making much more money, largely consists of people who are trying to have a significant positive social impact.
So suppose that you’re a highly skilled laborer deciding whether to “earn to give” or take a job with high social value that requires high skills and motivation. If you don’t take the job with high social value, your counterfactual replacement is likely be one of the following:
1. Substantially less capable than you on account of having low skills, or low altruistic motivation.
2. A highly skilled person with high motivation, who would be doing something else with high social value if you had taken the job, and who can’t do this because they have to do the job that you would have done.
3. Nonexistent.
So the replaceability consideration carries less weight than it might seem.
Admittedly there’s a counterconsideration — broad market efficiency cuts both ways, and one could imagine that the low hanging fruit in working directly on projects with high social value is also plucked, and this counter-consideration pushes in favor of “earning to give.” I have a fairly strong intuition that “if you don’t fund it, somebody else will” is more true than “if you don’t do it, somebody else will” so that this counter-consideration is outweighed. It’s important to note that many projects of high social value are the first of their kind, and that finding somebody else to execute such a project is highly nontrivial. I think that it’s also relevant that 114 billionaires have signed the Giving Pledge, committing to giving 50+% of their wealth away in their lifetimes.
In any case, there isn’t a clear-cut, unconditional argument that favors “earning to give”: whether “earning to give” is the best option very much depends on nuanced empirical considerations rather than a general abstract argument.
Other important considerations that favor an altruistic career
There are additional important considerations that favor pursuing a career with high social value over “earning to give”:
Asymmetric implications of the existence of small probability failure modes
In Robustness of Cost-Effectiveness Estimates and Philanthropy, I described how a large collection of small probability failure modes conspires to substantially reduce the expected value of a funding opportunity. The same issue applies to choosing a narrow career goal with a view toward directly having a high positive social impact. But a worker has more capacity than a donor does to learn whether small probability failure modes prevail in practice, and can switch to a different job if he or she finds that such a failure mode prevails.
Here’s an example. Suppose that you go to medical school with a view toward the possibility of performing cleft palate surgeries in the developing world. It’s probably the case that the opportunity isn’t as promising as it seems. But if you try it, then you’ll be able to see how effective the intervention is firsthand. If it’s highly effective, then you can keep doing it. If it’s not highly effective, then you can explore other possibilities, such as
- Starting your own surgery organization.
- Switching to doing a different kind of surgery in the developing world, such as cataract removal.
- Working in a poor community in the developed world (which could have a bigger impact than working in the developing world owing to flow-through effects).
- Working for a biotech company.
- Getting involved in clinical medical research.
- Other things that haven't occurred to me.
By experimenting, one can hope to hone in on a job that has both high ostensible cost-effectiveness, and and a relatively small mass of small probability failure modes.
Altruistic careers extend beyond the nonprofit world
Even on the assumption that “earning to give” is better than working at a nonprofit, it doesn’t follow that “earning to give” optimizes social impact. There are ways to have a positive social impact in the for-profit world, in scientific research, and in the government.
Historical Precedent
For the most part, the people who have had the biggest positive impact on the world haven’t had their impact by “earning to give.”
There are a few possible exceptions, such as Bill Gates and Warren Buffett, whose philanthropic activities could be having a huge impact (though it’s hard to tell from the outside) and could well outstrip the value that they contributed through their labor. But they appear to have an unusually high ratio of wealth to direct positive impact of their work, and so appear to be unrepresentative.
Steve Jobs’ highest net worth was on the order of $10 billion, whereas Bill Gates’ highest net worth was on the order of $100 billion. I don’t think that Bill Gates contributed 10x as much as Steve Jobs to technology, and I don’t think that Jobs could have had a bigger social impact by donating than through his work (which had massive positive flow-through effects). I acknowledge that Jobs is a cherry picked example, but I think that the general principle still holds.
Mainstream consensus
Few people think that “earning to give” is the best way to make the world a better place. This could be attributable to irrationality or to low altruism, but my experience is that there are many people who care about global welfare, or just welfare within a specific cause, and many people who are highly intelligent. In light of the existence of illusory superiority, one should be wary of holding an implicit view that one knows more about how to make the world a better place than the vast majority of the population.
Steelmanning wealth maximization
It’s worth highlighting some factors that favor choosing a career with a view toward maximizing wealth in some situations:
- Comparative advantage — Some people are unusually good at making money relative to doing other things. Such people may do better to “earn to give” than to try to choose a job that has a direct positive impact (which they’re relatively bad at).
- The market mechanism — In the for-profit world, maximizing wealth is often correlated with maximizing positive social impact, and so can be used as a proxy goal for maximizing positive social impact.
- Connections and personal growth — People with high earnings are generally more capable and more knowledgeable than people in other contexts, and tend to be well connected, so positioning oneself among such people can increase one’s prospects of soaring to greater heights. Jeff Bezos started his career in finance, and later created Amazon, which has had massive positive social impact (both direct, and via flow-through effects).
- Unusual values — If one cares about causes that very few people care about, then it could be difficult to find funding for work on them, so “earning to give” could be necessary. I don’t believe this to be the case, but it’s a consideration that's been raised by others, and so is worth mentioning.
Closing summary
There are many arguments against the claim that “earning to give” is generally the best way to maximize one’s positive social impact, and I believe that choosing a job where one can do as much good as possible through one’s work is generally the best way to maximize one’s positive social impact. However, for some people in unusual situations, “earning to give” may be the best way to have a positive social impact.
Note: I formerly worked as a research analyst at GiveWell. All views expressed here are my own.
Acknowledgements: I thank Nick Beckstead, ModusPonies and Will Crouch for helpful feedback on an earlier version of this article.
Crossposted from LessWrong.