In response to my writing that you should pick one top charity and just give your money there, that you shouldn't try to avoid risk by diversifying, a friend wrote:

I was thinking about risk-aversion in charitable giving. I'm familiar with arguments (I think we've talked about them) on about how risk-aversion in individuals doesn't make sense in giving (and relatedly, why it makes sense to pick your favorite organization), but I was thinking about whether society at large should be risk-averse in some sense.

If diversification is good on the whole, and people's choices are correlated, then maybe that leads to situations where individuals should diversify, or introduce a random element in their choice.

They're right in part: if we identify the best charity and then everyone donates only to them, very quickly they'll have more money than they can use effectively. This isn't how to maximize good for your dollar. Given that charity advice and evaluation organizations are not that widely heeded, this isn't currently a huge problem, as someone like givewell saying that village reach is the best charity they know of doesn't actually make that many people go give to them. Ideally, as "smart giving" expands, though, this will become something which we need to worry about.

The biggest argument for diversity in giving is the room for more funding aspect, well explained in this givewell blog post. The idea is that charities only have a certain amount of additional money that they would be able to spend as effectively or more effectively than the money they currently have, and if you give too much money even to the best charity it's not that effective. If you (individually or collectively) are giving enough money to exhaust a charity's room for more funding, than you need to be giving to multiple charities.

While the solutions above of individually diversifying or choosing partially at random would work, they have downsides. [1] Two other solutions come to mind. Givewell's answer to this is that you should give to them "for regranting only", and then they will give to what they belive to be the best charity that still has room for more funding. Another is what we currently do, which is give to oxfam and then because oxfam is quite large, it can allocate money internally so that it's running many projects, each with the money that it needs.

[1] If you give to multiple organizations then they need to spend more staff time and money on donation processing so your donation is worth less. While this is much more true with $20 donations than $2000 donations, even with large donations, this still applies. You want to be giving to the most effective charity, and if you give to multiple ones, some of your money is not being spent as effectively as it could be. Advocating that people be partly random in giving their money is the same as advocating giving to multiple organizations in the abstract.