Great post. I started reading and expected to find the explanation totally divorced from why I (a lacto-veg EA) am uninvolved in animal activism. But the high level explanation ‘it is unpleasant & often unrewarding moment-to-moment’ really resonated with me.
Making a community is really hard. Godspeed & good luck :)
dan.pandori
Karma: 484
Donation policy as tax policy
In percentages of pretax salary:
* 15% GiveWell
* 3% AI Safety orgs
* 1% Lightcone
Great post. Short & to the point with links to specific claims for those who want to understand more.
Your link is to a press release by Google Cloud. It’s not a financial statement and it doesn’t include expenses. Where can I “read the books” for Humanitix?
I agree that verifying “cage-free” eggs is probably harder, but it still doesn’t seem easy.
Verifiability remains very difficult even with PFG, where the ‘verifiability’ is that companies donate all of their profits.
Companies that make money on marginal sales have a large amount of discretion on what they classify as net profit. For example, as the CEO & founder you could increase your own salary. This decreases the net profit, as salaries are part of operating expenses. It doesn’t change gross profit (under most definitions), but by the time you’re discussing gross vs net profit consumers have already stopped listening. It isn’t just binary based on ownership or truthfulness of donating all profits.
Companies also have discretion about what charities to donate to, and it is hard to evaluate how effective a given charity is.
I’m also genuinely confused about whether Humanitix is profitable. This site [1] suggests that in 2021 it had ~3M in revenue, of which ~1.6M was from grants (so 1.4M left for its core services). Its total expenses were listed at 2.1 million, with a little less than 300K in grants, so 1.8M in expenses ignoring grants. 1.4M − 1.8M is a 400K yearly deficit. Businesses are totally allowed to lose money, but I think it weakens the case of Humanitix as a success story. Their more recent yearly reports have all figures redacted [2,3], and so I can’t really verify anything about what they donated vs how much they received. I’d be curious if folks could find some unredacted statements that I missed.
[1] https://www.acnc.gov.au/charity/charities/0db4989a-3aaf-e811-a961-000d3ad24182/documents/341f993f-e644-eb11-bb23-000d3ad1f9f4
[2] https://www.acnc.gov.au/charity/charities/0db4989a-3aaf-e811-a961-000d3ad24182/documents/
[3] https://acncpubfilesprodstorage.blob.core.windows.net/public/0db4989a-3aaf-e811-a961-000d3ad24182-e4b9f279-5504-4905-952f-ef0ba115c816-Financial%20Report-b67689aa-6b37-f011-8c4c-00224894978f-Humanitix_Limited_2024_Financial_Redacted.pdf
Fair point. I got confused because the examples highlighted are much more expensive than their natural competitors, and I incorrectly thought that meant the author believed them to match their description.
EX. a ‘thankyou.’ pack of 2 microfiber cloths (plus a glass cloth) is 17.95$ [1] vs Amazon sells a 12-pack for 7.99$ [2]. Similarly, ‘Good.store’ sells a coffee kit with 12 ounces of coffee and a mug for 50$ [3].
As far as I can tell no one meets the author’s bar of selling identical products at identical prices while solving real problems.
[1] https://thankyou.co/collections/cleaning-tools/products/microfibre-cloths-3pk
[2]https://www.amazon.com/Microfiber-Cleaning-Towels-Assorted-Yellow/dp/B098D79MQB/ref=sr_1_2_sspa?dib=eyJ2IjoiMSJ9.Uuyh4_VJcHdo5GS9oyjYoqhkQV4bx8B9ubn7bn49wZ0PPkZqCdSf-lei8XMFfMlEgSntrqK6MxGT_M8mTfLkws9OACyFY1vgAXfjP6ouTzRUNGk9FV_JABD40PxK9ZJ4osLmjbwf3vd9et5VHJeLXcJ0Lu1D0Lv4CCFavjDCHuc5-SwC1Cid7WJvQHVW9HjSJSbR67z4iR0wNkYu2pL9Q2sxho8kHKHCOyXYRlRywOwpaer86jQavMWNnMNXEamdU5V7GevNQlpwtqoTHRN9rzjfqTvNAQoT0HtgQxUJdpY.9ou4bL3nGy_-oU2u-zBFYHCxwYFXQIb0UC2R_kArVeA&dib_tag=se&hvadid=570596319494&hvdev=c&hvexpln=0&hvlocphy=9031923&hvnetw=g&hvocijid=11735289936918653588--&hvqmt=e&hvrand=11735289936918653588&hvtargid=aud-2443232233122%3Akwd-357431965329&hydadcr=8100_13493226&keywords=microfiber%2Bcloths%2Bamazon&mcid=9515a202da2d3465a66d8470c3bdd6b9&qid=1762833283&sr=8-2-spons&sp_csd=d2lkZ2V0TmFtZT1zcF9hdGY&th=1)
[3]https://good.store/products/coffee-bundle
No, people will not generally pay non-trivial amounts more for companies that ‘do good’.
Some salient counter examples:
Companies that use more humanely raised meat & eggs are not wildly successful (though I’d like them to be).
Alternative proteins are not wildly successful (though I’d like them to be).
Co-op stores generally don’t scale, and folks are more likely to shop at ordinary grocery stores & retailers.
Verifiability is also very hard. For more humane animal products, there are many competing labels. Even as a vegetarian and animal rights supporter, I have only the most rudimentary understanding of what is meant by cage-free vs free-range eggs in practice. You get about 5 words [https://www.lesswrong.com/posts/4ZvJab25tDebB8FGE/you-get-about-five-words]. Any label will get co-opted by industry, and teaching people which labels are real/meaningful is fighting an uphill battle against a superiorly funded adversary.
Status quo bias says that the largest & most successful companies are for-profit, and will stay that way in the near future.
Makes sense, thanks!
Why is it generally better for individuals to donate to 501(c)(4) organizations than to (c)(3)’s? I’m deeply ignorant in this space, so it’s a genuine question.
My super naive read is that c3’s are tax deductible (which is nice), presumably you think there is more than a 50% bonus in effectiveness of c4′s?
I don’t think that the intuition behind ‘curve fitting’ will actually get you the properties you want, at least for the formalizations I can think of.
How would you smooth out a curve that contains the St. Petersburg paradox? Simply saying to take the average of normal intuition and expected-value calculus (which you refer to as fanaticism) doesn’t help. EV calculus is claiming an infinity. I’m not aware of curve fitting approaches that give understandable curves when you mix infinite & finite values.
Plus, again, what dimensions are you even smoothing over?
I don’t get how you’re actually proposing doing curve-fitting. Like, the axes on your chart seem fake, particularly ‘moral cases’. What does 2 vs 8 ‘moral cases’ mean? What is a concrete example of a decision where you do X without curve fitting, but Y with curve fitting?
Without an actual mathematical formalization or examples, I struggle to see what your proposal looks like in practice. This seems like another downside relative to threshold deontology, where it is comparatively intuitive to see what happens before and after the threshold.
I don’t think that the study you cited here supports a 52% reduction in diarrhoea risk for these wells. The 52% quote is:
Compared with an unimproved source, provision of an improved drinking water supply on premises with higher water quality reduced diarrhoea risk by 52% (n=2; 0·48 [0·26–0·87])The wells created by Wells4Wellness do not deliver water to people’s homes, as far as I can tell. Doing so would be substantially more expensive. That study also seems to only consider water to be ‘improved’ if it is chlorinated, filtered, or solar treated. Searching for ‘well’ in that paper does not yield any hits. I could not find strong studies comparing deep well water to chlorinated water.
I think this is an interesting exercise, and it’s good to see more analyses of policy. But I don’t see this as an argument for the title ‘EAs should do more rough policy modeling’.
A post really interacting with that title would be showing how rough policy modeling is useful (ie, showing how it gets picked up by governments in actual policy, or has other positive downstream effects). Ideally, we’d get into how whether ‘rough policy modeling’ is more useful than than other similarly difficult activities EAs might do.
Best of luck! As someone who doesn’t eat eggs, I half-jokingly tell folks I can tell when cookies are vegan because they are bad. I look forward to not being able to use that heuristic!
GiveWell
For EA folks in tech, I’m still giving mock interviews. I’m bumping this into quick takes because my post is several years old, and I don’t advertise it well.
dan.pandori’s Quick takes
There are a lot of ‘lurkers’, but less than 30 folks would be involved in the yearly holiday matching thread and sheet. Every self-professed EA I talked to at Google was involved in those campaigns, so I think that covers the most involved US Googlers.
Most people donated closer to 5-10% than Jeff or Oliver’s much higher amounts, that is for sure true.
So I think both your explanations are true. There are not that many EAs at Google (although I don’t think that’s surprising), and most donate much less than they likely could. I put myself in that bucket, as I donated around 20%, but likely could have done close to twice that. Although it would be hard for me to do that in recent years, as I switched to Waymo where I can’t sell my stock.
Fully agreed that praise is an incredibly weak tool. Its all the weaker if its viewed as done with ulterior motives, like here.
RE: governments’ relationship with high tax payers. I think this is highly variable. My hometown in Iowa had a very positive relationship with the largest taxpayers/businesses/business owners. I think the Midwest has an understanding that there are many options for business owners, and so actively try to woo them.
In contrast, a large tech hub like San Francisco is much less replaceable, and so can afford to squeeze businesses much harder.