I’ve seen many people claim that splitting your donations (that is, donating some money to one charity and some money to another charity, instead of donating all your money to one charity) is irrational because you should just donate to the thing that you think is the best at using money on the margin.
However, I am confused about this. It seems to me that there are more complicated issues that this simple argument has not grappled with.
Imagine a toy problem where there are 100 effective altruists, each of whom are going to donate $100. There are two effective charities, the Against Paperclips Foundation and the Foundation to Cure Rare Diseases In Cute Puppies. The Against Paperclips Foundation creates +1.0000001 utility per dollar, while the Foundation To Cure Rare Diseases In Cute Puppies creates +1 utility per dollar (the puppies are really cute). The Against Paperclips Foundation has a room for more funding (RFMF) of $9000, while the Foundation To Cure Rare Diseases In Cute Puppies has a RFMF of $1000; if the RFMF is filled, further donations produce zero utility.
In theory, the effective altruists could fill the Against Paperclips Foundation’s room for more funding and then switch to the Foundation To Cure Rare Diseases In Cute Puppies. But there are going to be coordination issues. For example, many EAs donate near the end of the year, so the Against Paperclips Foundation’s RFMF might be filled all at once before they have a chance to announce that their RFMF is full. Some people are not going to read the EA Forum and find out that the Against Paperclips Foundation’s RFMF is full. Some people are going to forget to switch their automated donations.
In short, if everyone follows the “do whatever seems the best” rule, the Against Paperclips Foundation is going to receive extra money that does far less good than the Foundation To Cure Rare Diseases In Cute Puppies.
Conversely, if everyone donates 90% to the Against Paperclips Foundation and 10% to the Foundation To Cure Rare Diseases In Cute Puppies, there is a tiny tiny utility cost, because the Against Paperclips Foundation is slightly better than the Foundation To Cure Rare Diseases In Cute Puppies. But it also totally solves the coordination problem! No one has to communicate with anyone else; no one has to learn whether the RFMF is full; no one has to remember to donate to a different place. The right outcome happens automatically.
Further, “room for more funding” is a bit of an oversimplification. The Against Paperclips Foundation’s most pressing needs might be more important than the Foundation To Cure Rare Diseases In Cute Puppies’s most pressing needs, but that doesn’t mean that any money the Against Paperclips Foundation can use productively is more important as the Foundation To Cure Rare Diseases In Cute Puppies’s most pressing needs. The first paperclip researcher might be more important than curing a sad puppy of its fatal disease, but that doesn’t mean getting them new office furniture that will make research slightly more productive is more important than even the least expensive puppy to treat.
For this reason, I expect at the size the effective altruism movement is now (likely more than a hundred million non-OpenPhil dollars moved), the optimal outcome would be charitable donations spread across many charities. However, if everyone donates to the single best charity, this is unlikely to happen.
When I’ve brought up this argument to other effective altruists, they’ve said that without donation splitting you’re still going to get people donating to many different charities, because people have different values and worldviews. But it seems very unlikely to me that people having different values and worldviews would result in the money being allocated the way I would prefer. Indeed, many such people would donate in ways other people find useless or counterproductive (as a person who doesn’t think animals matter might think about effective animal advocacy). If someone donates to Machine Doggo Research Evaluators, it still doesn’t solve the problem of how to allocate money between the Against Paperclips Foundation and the Foundation to Cure Rare Diseases In Cute Puppies.
Second issue: I have often found myself in a situation where I’m not certain which of two charities is the best to donate to. They both seem like they will have very robustly positive effects, but my uncertainty is high enough that I’m genuinely not sure which of them has the highest expected value.
The sources of uncertainty are often difficult to resolve. For example, building an effective animal advocacy movement seems likely to save animals from factory farming, but it is extraordinarily difficult to say how many animals would be saved by, say, a new translation of Peter Singer’s Animal Liberation. In some cases I donate to a particular grantmaker– for example, through the EA Funds– because I expect their judgment to be better than mine; I can guess ‘this grantmaker will make better decisions than I will’ without having any idea how many animals this grantmaker will save.
Similarly, in many cases figuring out how good something is would involve solving very difficult philosophical problems. If a new translation of Animal Liberation inspires a person who would otherwise not have engaged in advocacy to become an advocate, and they save ten million animals, how many of those animals do I, a donor, get credit for? Or consider comparing across cause areas– how many animals spared from a factory farm is equivalent to a 20% lifetime increase in income for a person in the developing world? These are difficult problems to solve, and while it would be nice to solve them, I am going to donate this year and not after several decades of philosophical reflection.
In situations of sufficiently high uncertainty, my 95% confidence interval for two charities can often overlap. In general, in these situations, I wind up donation-splitting.
What am I missing?