I want to raise another argument against donating to mega-charities, in favour of those who we are confident are highly cost-effective. Mega-charities, by their nature, run many programs, in many locations. In effect, they draw multiple times from the distribution of interventions by effectiveness. By the Central Limit Theorem, the average of any such samples, if not heavily conditioned on high cost-effectiveness, will be closer to the average of the underlaying distribution. The larger the sample, the closer to the average we can expect. For an individual donor who cannot reasonably choose to support only a single intervention at a mega-charity, we should expect that a donation contributes towards their average cost-effectiveness. The same goes for directed donations, which suffer from high fungibility, as GiveWell mention in this comment. In other words, it's difficult to be one of the very best on average if you are doing lots of different stuff. Even if some of the interventions you do are really effective, your average effectiveness will be dragged down by the other interventions.

Thus, from a donor perspective, mega-charities are similar to index funds, you aren't likely to go very wrong by supporting them, and the average ROI is close to the average of all relevant interventions. However, you cannot expect to get the highest ROI by supporting mega-charities. For that, you need an organisation that specialises on one, or a few interventions that comes from the very top of the distribution.

Came here to suggest this

A Norwegian translation of this post is available here.

I really like these kinds of write-ups that provide a rough calculation of the cost-effectiveness of altruistic actions! A point I would like to poke at is the conclusion that if there's a 1/3 chance of your blood being used in a (fatal) surgery, you should attribute 1/3 * your share of the blood used of a life saved. This isn't counterfactual reasoning, which I think is the best way to go about analyzing this. When deciding, on the margin, whether I should donate blood or not, I should try to figure out what the expected consequences are if I donate as compared to me not donating.

My guess is that, in developed countries, it's extremely rare that people die during surgery because of a lack of blood units available at the hospital.

Some context for this initiative and how impactful it could be. I'm heading Gi Effekivt (www.gieffektivt.no) in Norway - the inspiration for Ge Effektivt (www.geeffektivt.se) in Sweden and now Giv Effektivt in Denmark. We've been up since 2016 and fundraised NOK 32 million (~$3.5 mill) to GiveWell recommended charities so far.

Creating national EA donation platforms with localization of content and payment methods should be a no-brainer in my opinion. People seems to have a much lower barrier for donating to a registered, transparent charity in their own country. For countries with tax deduction - securing this is an important sales point as well. Creating and running a donation platform is a great movement building experience as well. It's concrete, easy to grasp the value of and many different skillsets (marketing, content, development, legal, organizing, accounting and so on) are useful, but few skills are essential as a simple website with some info and a bank account number to donate to is enough to get started.

Aren't OpenPhil? https://www.openphilanthropy.org/giving/how-to-apply-for-funding#Can_I_apply_for_a_grant

They specify that they have low expectations for unsolicited proposals, but it's possible to contact them about it.

One argument against the effectiveness from mega charities who does a bunch of different, unrelated interventions is that from the Central Limit Theorem (https://en.m.wikipedia.org/wiki/Central_limit_theorem) the average effectiveness of a large sample of interventions is apriori more likely to be close to the population mean effectiveness - that is the mean effectiveness of all relevant interventions. In other words, it's hard to be one of the very best if you are doing lots of different stuff. Even if some of the interventions you do are really effective, your average effectiveness will be dragged down by the other interventions.

Thanks for pointing out these unclear sentences. I've made some changes in this paragraph to make my point more clearly.

The first part of the sentence remains; in some views, it is not right that a giver of gifts get any privileges on other benefits. But in a pure utilitarian view, this might be the case in some sense. If one party provides a gift to another, otherwise equal party, this will create an inequality that decrease the total utility. A pure utilitarian view will demand that a redistribution of benefits should follow to restore the equal situation.

Of course, the utilitarian will not use the term "rights" or "privileges" to argue the case for a redistribution after the gift. Also, it is worth pointing out that in a utilitarian view the initial gift is immoral as it decreases total utility, but this is a bit beside the point as this gifting is an assumed fact with this argument.

This made me think of backing up online EA content. It's not that hard to automatize backing up the content on the EA Forum, the EA Hub and the websites of CEA, GiveWell and other organizations. Not all movement collapse scenarios involve loosing access to online content and communication platforms, but it may be part of both internal conflict scenarios and external shocks.

Is the EA Forum regularly backed up, Aaron?

Regarding the question about the preferred resource allocation over the next five years, I would like to see someone take a stab at estimating the current allocation of resources over the same categories. My guess is that many of the cause areas are far from these numbers and it would imply a huge shift from status quo to increase or decrease the number of people and/or money going to each cause area.

The 3.5% allocation to wild animal welfare, for example, is 35 of the most engaged EAs contributing to the cause area and the money that goes with it. Or more people and less money if we trade off the resources against each other. Currently Wild Animal Initiative, the most significant EA actor in the field, employs eight people according a document on their website, most of them part-time. Going from here to 35 people would mean large investments and need of management, coordination and operations capacity. Especially if we should interpret it to mean 35 people on average over the next five years, given the position we're starting at.

I believe similar examples could be made from biosecurity (90 people and 9% of funding) or AI in shorter-timeline scenarios (180 people and 18% of funding). I guess that meta EA, global health and development and maybe farmed animal welfare are the categories that would need to scale back funding and people involved to reach these allocation targets.

Aaron commented that the respondents answered quickly to the survey questions, and this particular question even asks for a "rough" percentage of resource allocation. This might suggest that we shouldn't look too much at the exact average numbers, but only note the order of cause areas from most to least resources allocated.

Another possibility is that the respondents answered the questions based on the assumption that one could disregard all practical issues of how to get from status quo to this preferred allocation of resources. If so, I think it would be helpful to state this clearly with the question so that all respondents and readers have this in mind.