Adventures with quantitative philanthropy
Quantitative philanthropy definitely seems to be a Thing these days. In the Washington Post, Dylan Matthews is writing about GiveWell and about people who are taking high-earning jobs just so that they can give more money away; in the WSJ, Brad Reagan is writing about John Arnold and his determination “to solve some of the country’s biggest problems through data analysis and science”. (Free version here.) And Columbia University Press recently published The Robin Hood Rules for Smart Giving, a guide to the way in which philanthropies should use a framework called “relentless monetization” to guide where and how they spend their money.
Given how much I dislike philanthropy which is mainly designed to make the giver feel important, I ought to be happy about all these developments. But in fact, I’m quite conflicted.
Part of the reason is that philanthropy can not simply be reduced to dollar amounts. You can make a stab at doing so — and if you work for the Robin Hood Foundation, you will do just that. But Robin Hood is very clear that all it’s trying to do is create a way in which to compare apples with oranges, so that there’s a bit of structure under which to answer the “apple or orange?” question. If the Robin Hood quants decide that funding apples causes $4 of benefit while putting the same amount of money into oranges causes $5 of benefit, then they’ll chose the oranges over the apples. But they’re not saying that the apples actually create $4 of benefit: they’re just saying that given their mission, and the way in which they put numbers on such things, they prefer oranges to apples.
The Robin Hood people are very open about the fact that some other foundation with a different mission, or even a different set of quants at the same foundation making slightly different assumptions, could come to a very different conclusion. The quantification framework isn’t a way of replacing judgment with numbers; instead, it’s a way of helping people to be more explicit about exactly what assumptions they’re making when they choose one action over another.
This is my problem with the kind of philosophy Matthews is talking about; call it the GiveWell view. Here’s Matthews, explaining why a well-intentioned chap name Jason Trigg is working at a high-frequency trading shop:
Trigg makes money just to give it away. His logic is simple: The more he makes, the more good he can do.
He’s figured out just how to take measure of his contribution. His outlet of choice is the Against Malaria Foundation, considered one of the world’s most effective charities. It estimates that a $2,500 donation can save one life. A quantitative analyst at Trigg’s hedge fund can earn well more than $100,000 a year. By giving away half of a high finance salary, Trigg says, he can save many more lives than he could on an academic’s salary.
David Brooks does a good job of explaining how there can be something dehumanizing about such logic: there are good reasons to be suspicious of anybody who thinks that philanthropy can or should be reduced to dollar amounts or preset decision matrices. But there’s more going on here, and it’s worth unteasing the various different components of Trigg’s implicit syllogism.
First, there’s what you might call the Peter Singer imperative: “If it is within our power to prevent something bad from happening, without thereby sacrificing anything of comparable moral importance, we ought, morally, to do it.”
Second, we have the ability to turn money into saved lives. “Remember,” writes Matthews, “that giving about $2,500 can save one life from malaria.” Maybe the number is a little fuzzy: it might be $2,000 or $3,000 or even $4,000. But it is fair to say that if I spend tens of thousands of dollars a year on discretionary consumption, then there’s an opportunity cost to that spending — and the opportunity cost is dead people. They died of malaria, and if only I had given that money to the Against Malaria Foundation instead of spending it on booze and taxicabs and theater tickets, then those dead Africans would be alive today.
Third, there is no moral difference between directly saving someone’s life, on the one hand, and, on the other hand, spending an amount of money such that at the margin one extra life will have been saved. The moral worthiness of donating $2,500 to a malaria charity is exactly the same as the moral worthiness of jumping into a shallow pond to save a child from drowning. Both save one life.
Fourth, if you were walking past a shallow pond where a child was drowning, you would surely jump in to save that child.
Fifth, moral mathematics scales: if donating $2,500 and saving one life is good, then donating $5,000 and saving two lives is twice as good, and so on and so forth. If there are diminishing marginal returns on your donated dollar, then they’re small.
If you accept those five premises, then the conclusions rapidly and easily follow: not only should you donate the overwhelming majority of your disposable income to the Against Malaria Foundation, but you also have a moral imperative to maximize the amount of disposable income you have, just so that you can maximize the amount of money you donate and thereby the number of lives you can save.
But the fact is that when you disaggregate the logic in this way, you’ll find very few people willing to accept all five of the premises. I don’t want to wade into the realm of moral philosophy here, so let’s just concentrate on premise number two: the idea that there’s some kind of easy fungibility between dollars and saved lives. This is great as a marketing technique, and if you go through an elaborate quantitative exercise, you can end up at GiveWell’s “cost per life saved” conclusion:
Using the 2012-2013 cost per LLIN, we estimate the cost per child life saved through an AMF LLIN distribution at just under $2,300 using the marginal cost ($5.15 per LLIN) and just under $2,500 using the total cost ($5.54 per LLIN).
The Robin Hood Foundation is far from modest when it comes to its quantitative techniques, but never goes nearly as far as this. At Robin Hood, everything is based on conditionals: if you’re giving away a certain amount of money anyway, and if you have a certain mission statement, and if you make certain clearly-delineated but far-from-certain assumptions, then you will end up preferring this course of action to that one.
GiveWell, by contrast, makes a vastly bolder claim: that if you donate $2,500 to the Against Malaria Foundation, you will save a child’s life. It is entirely transparent in how it arrives at that conclusion, and is perfectly happy for people to quibble with its methodology. But by the time the claim makes it into the pages of the Washington Post, any minor caveats have fallen away, and the conclusion is taken as a simple statement of fact.
I believe that humility is of paramount importance in all philanthropy, and I worry that there’s not nearly enough of it at GiveWell. Malaria is a sexy disease: it’s probably second only to HIV/AIDS in terms of the amount of not-for-profit resources being thrown at it. Zoom out, for a minute, and look at national governments in Africa; at universities and other research institutions around the world; at national and international development banks; at huge philanthropies like the Gates Foundation or the Global Fund; at the philanthropic arms of the big pharmaceutical companies; and, yes, at smaller charities like the Against Malaria Foundation. Add it all up, and you’ll find thousands of people marshaling billions of dollars in resources. These people have been working for many years on trying to build effective strategies to minimize malaria’s footprint and, ultimately, to eradicate it altogether. Some of those strategies have been more effective than others, and the broad distribution of long-lasting insecticide-treated bed nets is undoubtedly one of the better ones. Still, here’s the thesis that GiveWell would have you believe: you can be reasonably certain that every time someone donates another $2,500 to the Against Malaria Foundation, another child’s life will be saved.
I don’t buy it. It’s not that the Against Malaria Foundation isn’t an excellent, well-run charity doing an excellent job at distributing bed nets. In and of itself, distributing bed nets is a good thing to do, and it should, on balance, be helpful the fight against malaria. The Against Malaria Foundation is a worthy cause, and if you write them a check, you’re not going to be committing any of the sins I railed against in December. What’s more, the amount of time that the Against Malaria Foundation has spent with GiveWell has surely paid huge dividends for them: thanks to the GiveWell seal of approval, they have received millions of dollars they would not have gotten otherwise.
But there’s something far too facile about putting a dollar amount on marginal lives saved. Even GiveWell says so: look at the blog post they put up in December, in which they tried to quantify the relative cost-effectiveness of their top three charities. Malaria is a lethal disease, which makes it almost too easy to try to put a dollar amount on lives saved. But attempting the same exercise for deworming is much harder, and they didn’t even bother trying for their third charity, which simply gives cash lump sums, unconditionally, to the poor. Basically, if you want to use GiveWell to help you find a charity where you can buy a saved life with a known number of dollars, you’re going to end up with a shortlist of one.
The Trigg syllogism, then, rests on an assumption which rarely obtains in the real world: that giving money to charity is something which can and will predictably save lives. Charities tend to do little to disabuse this assumption, because people who believe it are likely as a result to donate more money. But talk to anybody who actually works in development or for a nonprofit, and they’ll tell you the world is a lot messier than that. Most of the time, when we give money to charity, we can realistically only hope that our donation will make a positive difference at all; all too often, a marginal extra dollar, even when it’s given to an unambiguously good cause can cause more harm than good.
The purpose of GiveWell is to try to identify the charities where that doesn’t happen — the charities where your donation will do the most good and the least harm. Certain quantitatively-minded types consider that a very worthy cause, and I’d be one of them too, if I believed that GiveWell could really do that.
Instead, however, I think that GiveWell faces three very big problems. The first is model risk: although GiveWell is very open about the models that it’s using, there’s no particular reason to believe that they are robust; obviously, their recommendations are no better than the models used to generate them. The second is that GiveWell imposes a substantial burden on the large number of charities it investigates and doesn’t recommend; that burden carries a real cost.
Finally, and most importantly, there’s GiveWell’s built-in bias towards relatively small-scale, replicable and quantifiable interventions. (Robin Hood has the same bias, for the same reasons.) Such actions are an important part of the philanthropic universe, but the Robin Hood and GiveWell types have a tendency to become evangelical about the way in which virtually all charities should adopt such a framework, and that, I think, is a very bad idea, for reasons that Rob Reich explored in his Boston Review essay about foundations.
Reich explains that things like bed nets are public goods, best provided by the state. And that philanthropies are best placed to do something else entirely: they “can operate on a longer time horizon than can businesses in the marketplace and elected officials in public institutions,” he says, “taking risks in social policy experimentation and innovation that we should not routinely expect to see in the commercial or state sector.” He continues:
When it comes to the ongoing work of experimentation, foundations have a structural advantage over market and state institutions: a longer time horizon. Once more, the lack of accountability may be a surprising advantage. Commercial entities in the marketplace do not have an incentive structure that systematically rewards high-risk, long time horizon experimentation; they need to show quarterly results. Similarly, public officials in a democracy do not have an incentive structure that rewards high-risk, long time horizon experimentation; they need to show results quickly from the expenditure of public dollars in order to get re-elected. In contrast, foundations are not subject to earnings reports, impatient investors or stockholders, or short-term election cycles.
Foundations, answerable only to the diverse preferences and ideas of their donors, with a protected endowment permitted to exist in perpetuity, may be uniquely situated to engage in the sort of high-risk, long-run policy innovation and experimentation that is healthy in a democratic society.
It seems to me that the philosophy of GiveWell and Robin Hood runs directly counter to this philosophy: they require precisely the short-term results which are required also by the marketplace and by the state. As such, they are deliberately disarming philanthropies of one of their key structural advantages.
Both GiveWell and Robin Hood were founded by marketplace-friendly hedge-fund types — but then again, so was the Laura and John Arnold Foundation which falls squarely into Reich’s sweet spot, funding the kind of ambitious, risky, long-term projects with none of the predictability that GiveWell and Robin Hood require. The Arnolds — like the Gateses — take a scientific, quantitative view of the world. But that doesn’t mean they take a narrowly quantitative approach to philanthropy itself. Bill Gates has more than enough money to give a bed net to every living child in Malawi — but his ambition is greater than that. And while I have my quibbles with the way the Gates Foundation is run, I do applaud its ambition and the way in which it tries to do the kind of things that only a monster-size philanthropy could ever even attempt.
Jason Trigg, in this light, is not a moral hero equivalent to a man who has saved dozens of children from drowning; instead, he starts to look more like someone contracting a third party to provide a service which is probably best provided at the state level in the first place. There’s nothing bad about what he’s doing — far from it. But I do think that what he does is limited along a number of axes, and that if he spent less time working at high-frequency trading, and more time out in the field learning about the way that non-profit develop organizations work in practice, then he might develop a richer, more nuanced, and probably humbler view of what his financial contributions can and can’t achieve in reality.
Update: GiveWell responds, pointing to a 2011 post in which they made most of the points I’m making here, and in which they said that they were going to move away from narrowly quantitative estimates and towards a more holistic view of where money can be best spent.