Key Takeaways

• Optimizing your giving's effect on "EA's portfolio” implies you should fund the causes your value system thinks are most underfunded by EA's largest allocators (e.g. Open Phil and SFF).
• These causes aren't necessarily your value system's most preferred causes. ("Preferred" = the ones you'd allocate the plurality of EA's resources to.)
• For the typical EA, this would likely imply donating more to animal welfare, which is currently heavily underfunded under the typical EA's value system.
• Opportunities Open Phil is exiting from, including invertebrates, digital minds, and wild animals, may be especially impactful.

Alice's Investing Dilemma: A Thought Experiment

Alice is a conservative investor who prefers the risk-adjusted return of a portfolio of 70% stocks and 30% bonds. Along with 9 others, Alice has been allocated $1M to split between stocks and bonds however she sees fit. The combined $10M portfolio will be held for 10 years, and its profits or losses will be split equally among the 10 portfolio managers. The other 9 portfolio managers tell Alice they're planning to go with 100% stocks.

Alice's preferred asset is stocks (in the sense that if she could control the whole combined portfolio, she'd allocate the majority to stocks). However, the underallocated asset (by Alice's risk-adjusted return preference) is bonds. In this case, Alice best realizes her preferences by allocating her entire $1M to bonds! This holds even though Alice prefers stocks to bonds.

In Charity, We Should Optimize The Portfolio of Everyone's Actions

In Alice's investing dilemma, the premise that's doing the work is that Alice wants to optimize the combined portfolio instead of her particular $1M share.

In the case of effective giving, we typically focus on our giving's direct impact, but not on how it fits into the portfolio of the net effect of everyone's actions. But optimizing the portfolio of everyone's actions seems to directly follow from EA principles:

• The recipient of charity doesn't care who's giving it, so it would seem like a bias to be focused on the part of "the portfolio of everyone's actions" that is your actions rather than the whole (or any other particular part).
• Reducing funging to ensure counterfactual impact is already one way of reasoning about the effect of your giving on the portfolio of everyone's actions. This proposal simply extends that idea to also optimize for your value system's objectives.

There are many legitimate reasons to not overemphasize optimizing the portfolio of everyone's actions, such as many people's concerns about personally making a difference. However, I think we should put more thought into optimizing the portfolio of everyone's actions than we currently do.

Theoretical Implications

• You should prefer funding the causes your value system thinks are the world's most underallocated.
• These causes are not necessarily your value system's most preferred causes! ("Preferred" = The ones which you'd allocate the plurality of the world's resources to!)

"The Portfolio of Everyone's Actions" vs "EA's Portfolio"

In theory, this post argues that you should be optimizing the portfolio of the net effect of anything anyone will ever do (under your value system). But that's obviously intractable!

To deduce practical recommendations, one can assume that non-EA-aligned actions have negligible net effect relative to EA-aligned actions. In that case, "optimizing the portfolio of everyone's actions" reduces to optimizing the portfolio of EA's resource allocation. If you think this assumption is generally accurate, we can deduce some practical recommendations.

Practical Recommendations

Many EAs split their personal donations between cause areas including global health, animal welfare, and longtermism. If you're optimizing EA's portfolio, you probably shouldn't do this. Instead, you should identify which cause area your value system says EA most underfunds, and only donate there.

I personally believe longtermist interventions have the highest expected value, and I would allocate the plurality of EA resources to them if I could. But due to risk aversion, I think a substantial portion of our resources should go towards reducing near-term suffering, which animal welfare interventions do most cost-effectively. Since my value system says animal welfare is more underfunded than longtermism, when optimizing EA’s portfolio, it seems best for me to donate only to animal welfare. This holds even though longtermism is my preferred cause area.

For other value systems, the implications could be completely different! If Bob doesn't care about animals, but would want to split EA’s resources between 30% global health and 70% longtermism, optimizing EA’s portfolio by Bob’s value system means he should donate only to longtermist interventions.

EA's Current Resource Allocations

Knowing EA's current resource allocations would be helpful if you think this post's recommendations have merit. The most complete and up-to-date reference I know of is Tyler Maule's from November 2023:

The consensus of EA leaders and the EA community is that global health is overfunded. If global health is excluded, Tyler's aggregation gives:

Some other potentially helpful aggregations:

1. Open Phil grants by cause area by Hamish McDoodles (updated daily)
2. Resource allocations by cause area by Ben Todd (as of 2019)

If anyone is interested in maintaining a more complete and up-to-date aggregation, that could be impactful. The EA community could use that as a canonical resource to better target EA's most underallocated causes.