Drowning children are rare

Link post

Sto­ries such as Peter Singer’s “drown­ing child” hy­po­thet­i­cal fre­quently im­ply that there is a ma­jor fund­ing gap for health in­ter­ven­tions in poor coun­tries, such that there is a moral im­per­a­tive for peo­ple in rich-coun­tries to give a large por­tion of their in­come to char­ity. There are sim­ply not enough ex­cess deaths for these claims to be plau­si­ble.

Much of this is a restate­ment of part of my se­ries on GiveWell and the prob­lem of par­tial fund­ing, so if you read that care­fully and in de­tail, this may not be new to you, but it’s im­por­tant enough to have its own con­cise post. This post has been ed­ited af­ter its ini­tial pub­li­ca­tion for clar­ity and tone.

Peo­ple still make the fund­ing gap claim

In his 1997 es­say The Drown­ing Child and the Ex­pand­ing Cir­cle, Peter Singer laid out the ba­sic ar­gu­ment for a moral obli­ga­tion to give much more than most to, for the good of poor for­eign­ers:

To challenge my stu­dents to think about the ethics of what we owe to peo­ple in need, I ask them to imag­ine that their route to the uni­ver­sity takes them past a shal­low pond. One morn­ing, I say to them, you no­tice a child has fallen in and ap­pears to be drown­ing. To wade in and pull the child out would be easy but it will mean that you get your clothes wet and muddy, and by the time you go home and change you will have missed your first class.
I then ask the stu­dents: do you have any obli­ga­tion to res­cue the child? Unan­i­mously, the stu­dents say they do. The im­por­tance of sav­ing a child so far out­weighs the cost of get­ting one’s clothes muddy and miss­ing a class, that they re­fuse to con­sider it any kind of ex­cuse for not sav­ing the child. Does it make a differ­ence, I ask, that there are other peo­ple walk­ing past the pond who would equally be able to res­cue the child but are not do­ing so? No, the stu­dents re­ply, the fact that oth­ers are not do­ing what they ought to do is no rea­son why I should not do what I ought to do.
Once we are all clear about our obli­ga­tions to res­cue the drown­ing child in front of us, I ask: would it make any differ­ence if the child were far away, in an­other coun­try per­haps, but similarly in dan­ger of death, and equally within your means to save, at no great cost – and ab­solutely no dan­ger – to your­self? Vir­tu­ally all agree that dis­tance and na­tion­al­ity make no moral differ­ence to the situ­a­tion. I then point out that we are all in that situ­a­tion of the per­son pass­ing the shal­low pond: we can all save lives of peo­ple, both chil­dren and adults, who would oth­er­wise die, and we can do so at a very small cost to us: the cost of a new CD, a shirt or a night out at a restau­rant or con­cert, can mean the differ­ence be­tween life and death to more than one per­son some­where in the world – and over­seas aid agen­cies like Ox­fam over­come the prob­lem of act­ing at a dis­tance.

Singer no longer con­sis­tently en­dorses cost-effec­tive­ness es­ti­mates that are so low, but still en­dorses the ba­sic ar­gu­ment. Nor is this limited to him. As of 2019, GiveWell claims that its top char­i­ties can avert a death for a few thou­sand dol­lars, and the Cen­ter for Effec­tive Altru­ism claims that some­one with a typ­i­cal Amer­i­can in­come can save dozens of lives over their life­time by donat­ing 10% of their in­come to the Against Malaria Foun­da­tion, which points to GiveWell’s anal­y­sis for sup­port. (This de­spite GiveWell’s long-stand­ing dis­claimer that you shouldn’t take its ex­pected value calcu­la­tions liter­ally). The 2014 Slate Star Codex post In­finite Debt de­scribes the Giv­ing What We Can pledge as effec­tively a ne­go­ti­ated com­pro­mise be­tween the per­ceived moral im­per­a­tive to give liter­ally ev­ery­thing you can to alle­vi­ate Bot­tom­less Pits of Suffer­ing, and the un­der­stand­able de­sire to still have some nice things.

How many ex­cess deaths can de­vel­op­ing-world in­ter­ven­tions plau­si­bly avert?

Ac­cord­ing to the 2017 Global Bur­den of Disease re­port, around 10 mil­lion peo­ple die per year, globally, of “Com­mu­ni­ca­ble, ma­ter­nal, neona­tal, and nu­tri­tional dis­eases.”* This is roughly the cat­e­gory that the low cost-per-life-saved in­ter­ven­tions tar­get. If we as­sume that all of this is treat­able at cur­rent cost per life saved num­bers—the most gen­er­ous pos­si­ble as­sump­tion for the claim that there’s a fund­ing gap—then at $5,000 per life saved (sub­stan­tially higher than GiveWell’s cur­rent es­ti­mates), that would cost about $50 Billion to avert.

This is already well within the ca­pac­ity of funds available to the Gates Foun­da­tion alone, and the Open Philan­thropy Pro­ject /​ GiveWell is the main ad­vi­sor of an­other multi-billion-dol­lar foun­da­tion, Good Ven­tures. The true num­ber is al­most cer­tainly much smaller be­cause many com­mu­ni­ca­ble, ma­ter­nal, neona­tal, and nu­tri­tional dis­eases do not ad­mit of the kinds of cheap mass-ad­ministered cures that jus­tify cur­rent cost-effec­tive­ness num­bers.

Of course, that’s an an­nual num­ber, not a to­tal num­ber. But if we think that there is a pre­sent, rather than a fu­ture, fund­ing gap of that size, that would have to mean that it’s within the power of the Gates Foun­da­tion alone to wipe out all fatal­ities due to com­mu­ni­ca­ble dis­eases im­me­di­ately, a cou­ple times over—in which case the progress re­ally would be per­ma­nent, or at least quite last­ing. And in­fec­tions are the ma­jor tar­get of cur­rent mass-mar­ket donor recom­men­da­tions.

Even if we as­sume no long-run di­rect effects (no re­duc­tion in in­fec­tion rates the next year, no flow-through effects, the peo­ple whose lives are saved just sit around not con­tribut­ing to their com­mu­ni­ties), a large fund­ing gap im­plies op­por­tu­ni­ties to demon­strate im­pact em­piri­cally with ex­ist­ing funds. Take the ex­am­ple of malaria alone (the tar­get of the in­ter­ven­tion speci­fi­cally men­tioned by CEA in its “dozens of lives” claim). The GBD re­port es­ti­mates 619,800 an­nual deaths—a re­duc­tion by half at $5k per life saved would only cost $3 billion per year, an an­nual out­lay that the Gates Foun­da­tion alone could sus­tain for over a decade, and Good Ven­tures could cer­tainly main­tain for a cou­ple of years on its own.

GiveWell’s stated rea­son for not both­er­ing to mon­i­tor statis­ti­cal data on out­comes (such as e.g. malaria in­ci­dence and mor­tal­ity, in the case of AMF) is that the data are too noisy. A re­duc­tion like that ought to be very no­tice­able, and there­fore ought to make filling the next year’s fund­ing gap much more ap­peal­ing to other po­ten­tial donors. (And if the in­ter­ven­tion doesn’t do what we thought, then po­ten­tial donors are less mo­ti­vated to step in—but that’s good, be­cause it doesn’t work!)

Imag­ine the world in which funds already al­lo­cated are enough to bring deaths due to com­mu­ni­ca­ble, ma­ter­nal, neona­tal, and nu­tri­tional dis­eases to zero or nearly zero even for one year. What else would be pos­si­ble? And if you think that peo­ple’s re­vealed prefer­ences cor­rectly as­sume that this is far from pos­si­ble, what speci­fi­cally does that im­ply about the cost per life saved?

What does this mean?

If the low cost-per-life-saved num­bers are mean­ingful and ac­cu­rate, then char­i­ties like the Gates Foun­da­tion and Good Ven­tures are hoard­ing money at the price of mil­lions of pre­ventable deaths. If the Gates Foun­da­tion and Good Ven­tures are be­hav­ing prop­erly be­cause they know bet­ter, then the op­por­tu­nity to save ad­di­tional lives cheaply has been greatly ex­ag­ger­ated. My former em­ployer GiveWell in par­tic­u­lar stands out, since it pub­lishes such cost-per-life-saved num­bers, and yet recom­mended to Good Ven­tures that it not fully fund GiveWell’s top char­i­ties; they were wor­ried that Good Ven­tures would be sav­ing more than their “fair share” of lives.

In ei­ther case, we’re not get­ting these es­ti­mates from a source that be­haves as though it both cared about and be­lieved them. The pro­cess that pro­moted them to your at­ten­tion is more like ad­ver­tis­ing than like sci­ence or busi­ness ac­count­ing. Ba­sic epistemic self-defense re­quires us to in­ter­pret them as mar­ket­ing copy de­signed to con­trol your be­hav­ior, not un­bi­ased es­ti­mates de­signed to im­prove the qual­ity of your de­ci­sion­mak­ing pro­cess.

We should be more skep­ti­cal, not less, of vague claims by the same par­ties to even more spec­tac­u­lar re­turns on in­vest­ment for spec­u­la­tive, hard to eval­u­ate in­ter­ven­tions, es­pe­cially ones that promise to do the op­po­site of what the ar­gu­ment jus­tify­ing the in­ter­ven­tion recom­mends.

If you give based on mass-mar­keted high-cost-effec­tive­ness rep­re­sen­ta­tions, you’re buy­ing mass-mar­keted high-cost-effec­tive­ness rep­re­sen­ta­tions, not lives saved. Do­ing a lit­tle good is bet­ter than buy­ing a sym­bolic rep­re­sen­ta­tion of a large amount of good. There’s no sub­sti­tute for de­vel­op­ing and act­ing on your own mod­els of the world.

As far as I can see, this pretty much de­stroys the generic util­i­tar­ian im­per­a­tive to live like a monk while max­i­miz­ing earn­ings, and give all your ex­cess money to the global poor or some­thing even more ur­gent. In­so­far as there’s a way to fix these prob­lems as a low-info donor, there’s already enough money. Claims to the con­trary are ei­ther ob­vi­ous non­sense, or mar­ket­ing copy by the same peo­ple who brought you the ob­vi­ous non­sense. Spend money on tak­ing care of your­self and your friends and the peo­ple around you and your com­mu­nity and try­ing spe­cific con­crete things that might have spe­cific con­crete benefits. And try to fix the un­der­ly­ing sys­tems prob­lems that got you so con­fused in the first place.

* A pre­vi­ous ver­sion of this post er­ro­neously read a decadal rate of de­cline as an an­nual rate of de­cline, which im­plied a stronger con­clu­sion than is war­ranted. Thanks to Alexan­der Gor­don-Brown to point­ing out the er­ror.