GiveWell, the independent effective-altruist charity evaluator, just announced a new set of charity ratings. Specifically, GiveDirectly is now the top-rated organization. (The previous top-rated organization, Against Malaria Foundation, failed to meet a milestone and so GiveWell is holding off on recommending funding until they show that they can meet the milestone.)

GiveWell evaluates charities on the basis of expected world-improving value for your money. Their philosophy focuses heavily on proven interventions - they are not likely to recommend far-future speculative causes because there’s no way to prove them. They are very thorough and demand high standards of evidence and transparency. GiveWell’s recommendations are very widely respected.

GiveDirectly just gives money to the most needy. They find the poorest households in Kenya (by looking for houses with do not have solid walls/roofs) and surprise them with $500. They have an impressively thorough body of evidence that says that this works – that people spend the money in very good ways.

Anyway, GiveDirectly is now the number one most effective donation target, as evaluated by GiveWell. This is quite surprising to me. Why? Because GiveDirectly seems to be a low bar!

Okay, wait, let me explain. Clearly, it’s not really a low bar: the money is going to verifiably poor people, and the research tells us that they spend the money in high-value ways. In fact, it’s easy to see that most charities will fall below this bar: I can easily imagine education charities, for example, where most of the money spent goes to children who would otherwise have gotten a good education in other ways.

But all of them? Essentially, GiveWell is saying “we cannot find any proven economies of scale when it comes to making the world a better place.”

Let me expand on this. Economies of scale is why companies exist. For example, it would be hard for me to make myself a cellphone from electronic parts. It would cost me hundreds or thousands of dollars just to source the parts, and then I’d have to combine them all into a working device, and it probably wouldn’t fit in my pocket. But I can get a working cellphone for fifteen dollars or less – at least two or three orders of magnitude cheaper than doing it myself. The cellphone companies make money by producing phones in the millions (or billions); manufacturing techniques allow them to make an extremely large number of phones for very little, and sell them at a profit.

So yes, I’m surprised: since companies have to make a profit to continue existing, it seems like there should be a nonprofit which leverages similar economies of scale, but which could be even cheaper (and therefore more effective) because they don’t have to make a profit.

And yet, for some reason, this doesn’t exist in the charity world. Nobody has found a charitable intervention that scales in the same way that mass production scales. Or, at least, they haven’t found one that a) scales; b) is proven to be effective; and c) isn’t already a for-profit corporation. It appears that simply giving people cash transfers beats out any kind of replicable, mass-produced charity work. There’s no mass-produced thing (a vaccine, or a drug, or a mosquito net, or a wheelbarrow) whose production scales well with charitable dollars.

Scaling: is it possible that a charitable intervention doesn’t scale as well as a company? It seems rather unlikely. The production and distribution of things like vaccines and mosquito nets don’t seem fundamentally different from all the things that are distributed for profit, like soda and shampoo. Maybe the best charity interventions are more long-term focused than most stuff people buy in stores.

Proven effectiveness: to me, this seems like a plausible problem in the charity sphere. Companies don’t have to prove anything except to themselves – if something is working, they get immediate feedback in the form of dollars and they just have to learn to scale it. But with charity, you don’t necessarily know if things are working, if they’re getting better, and so on, without trying hard to find out. It’s so much easier to just hope/believe/think that your interventions are working than to actually test…

Isn’t already a for-profit company: GiveWell doesn’t recommend for-profit ventures, I assume because world-improving profitable companies can already get plenty of investors and money elsewhere. So it raises the question – maybe all the best applications of economies-of-scale work well in a for-profit structure, and so they’re invisible to philanthropists. This actually seems also somewhat plausible, but I would expect that there were some companies which couldn’t quite make money but still would be worth funding as a nonprofit to distribute their product.

If you aren’t pretty sure that one of these is true, then it seems like there is a giant hole where effective altruists should look for promising new nonprofits to start. Something which scales, is expected to have a scientifically demonstrable positive impact, and can’t exist as a for-profit. If you could prototype your idea and gather some evidence of impact, and if you otherwise executed well, it seems quite likely that you could quickly get this organization to #1 on GiveWell’s recommendations and make a big impact.

So let’s brainstorm it. People who get grants from GiveDirectly very often buy metal roofs. Presumably the recipients have to buy them “retail” (whatever that is in Malawi) and either learn to install the roofs themselves, or pay a contractor to install the roof. But GiveDirectly already knows the houses who need metal roofs since that’s their criterion for giving them grants. If they can obtain substantial cost savings by installing them “in bulk” then shouldn’t they just go around installing roofs everywhere for free?

Well, this goes against what GiveDirectly wants to do. And maybe they’re right. How would we know? Consider how many people there would be under my scheme who don’t really want a roof, but would get one anyway if we handed it to them. Clearly, those people would prefer the cash. But we saved a lot of money mass-producing and mass-installing roofs for everyone else, so we got more roofs for the dollar than we would through GiveDirectly’s scheme, so it’s a question of whether there are more wasted roofs or whether the cost savings makes up for it.

My guess would be that there would be more wasted roofs – roofs are already mass produced and there are probably cheap contractors to help install them, so they wouldn’t get too much benefit from the economies of scale. But the real question isn’t whether this works for roofs, it’s whether it works for ANY intervention. If you can find a single intervention that works better than handing people cash, you can beat GiveDirectly.

Ok, I have two ideas for places to look: interventions which require large groups of people (collective action problems) and interventions which are against cultural norms. Both are likely to be ignored by the for-profit sphere, which is why I picked them out.