Global poverty could be more cost-effective than animal advocacy (even for non-speciesists)

I agree with the general principle: non-robust estimates should be discounted, and thus areas where the evidence is less robust (e.g. animal welfare, and probably a fortiori far future) should be penalized compared to areas with more robust estimates (e.g. global poverty) in addition to any 'face value' comparison.

I also agree it is plausible that global poverty interventions may be better than interventions in more speculative fields because we are closer to selecting randomly from these wide and unknown distributions in the latter case, so even if the 'mean' EV from a global health charity is << than the mean EV of (say) a far future cause, the EV of a top Givewell charity may be higher than our best guess for best animal welfare cause, even making generous assumptions about the ancillary issues required (e.g. inter-species welfare comparison, pop ethics, etc. etc. etc.)

However, I think your illustration is probably slanted too unfavourably towards the animal welfare cause, for two reasons.

1. Due to regression to mean and bog standard measurement error, it seems likely that estimates of the best global poverty interventions will be biased high, ditto any subsequent evaluation which 'selects from the top' (e.g. Givewell, GWWC). So the actual value of best measured charity will be less than 70x greater than the actual mean.

2. I broadly agree with your remarks about the poverty of the evidence base in animal welfare interventions. However it seems slightly too conservative to discard all information from (e.g.) ACE entirely.

The DCP data does give fair evidence that the distribution for global poverty interventions is approximately log normal and I'd guess it's mean is fairly close to the 'true' population value. It is unfortunate that there is no similar work giving approximate distribution type or parameters for animal welfare/global poverty/ anything else causes. I would guess it is also lognormally distributed-ish (there or there abouts, I agree with your remarks about plausibly negative values) although with an unclear mean.

I have been working on approaches for corecting regression to the mean issues. Although the results are mathematically immature and probably mistaken (my previous attempt was posted here, which I hope to return to in time - see particularly Cotton-Barratt's remarks), I think the two qualitative takeaways are important. 1) With heavy tailed (e.g. log-normal) distributions, regression to the mean can easily knock orders of magnitude off the estimate for 'best performing' interventions, 2) regression to the mean can bite much more (namely, orders of magnitude more) off a less robust estimate than a more robust estimate.

For these reasons comparing (e.g.) top Givewell charity estimates to ACEs effectiveness estimates are probably illegitimate, as the latter's estimates will probably have much greater expected error, in part due to being a smaller org with less human and capital resources to devote to the project, and (probably more importantly) the considerably worse evidence base they have to work with. For similar reasons arguments of the form 'I did a fermi estimate with conservative assumptions, and it turns out X has a QALY yield a thousand/million/whatever times greater than Givewell top charities, therefore X is better' warrant withering scepticism.

How to go further than this likely requires distributional measures we are unlikely to get good access to save for global poverty. There is some research one could perform to get a handle on regression the mean, and potentially some analytic or simulation methods to estimate 'true' effectiveness conditioned on the error prone estimate, and I hope to attack some this work in due course. For other fields, similar to Dickens, one may conjecture different distributions and test for sensitivity, my fear is the results of these analyses will prove wholly sensitive to recondite statistical considerations.

I also agree with your remarks that global poverty causes may prove misleadingly robust given the challenge of estimating flow through effects and differential impact given varying normative assumptions. Thus 'true' EV of even the most robustly estimate global poverty cause likely has considerable error, and these errors may not be easy to characterize, and plausibly in many cases 'pinning down' the relevant variables may demand a greater sample size than can be obtained prospectively within the Earth's future light-cone. I leave as an exercise to the reader how this may undermine the method within EA for relying on data, cost effectiveness estimates, and so forth.

I think this is an imperfect comparison since finding the most effective health intervention isn't equivalent to finding the most effective animal charity; rather, finding the most effective health intervention is equivalent to finding the cheapest website to put up vegan ads. The true uncertainty over poverty relief - the way it interacts with governance systems and dictates long term development trends - necessarily needs to be included as well. And we don't know much about which poverty charities are best under that metric. I don't think it's right to imply that it is a vague or optional "far future" concern, as it's a concrete issue that can't be ignored.

Moreover, why refer to the median animal intervention as the baseline? If we want the expected value of picking a charity at random, we'll want to use the average. And if there are a few interventions which are 250x as good as the best poverty charities with most of them being ineffective, then that will strongly improve the estimate above what we would get if using the median.

Besides, we're not totally blind about animal charities and we can still make some improvements over random charity selection. I think it's reasonable to expect that we have at least a 1-in-250 chance of selecting the best animal charity. If so, then we should expect animal charity to be better than human charity on average.

Note that if this does turn out to be a close decision then the meat eater problem becomes significant again. The typical response to problems of meat consumption is that anyone who cares sufficiently about animals will donate to animal charities anyway; however, if the OP's reasoning is correct then we now have to take it seriously into consideration.

The biggest takeaway here is that animal charity research is a really good cause.

My own analysis of the raw data provided

I'm glad someone's getting something out of the two hours I spent manually copying all the DCP2 estimates :)

It looks like you're using the mean of the DCP2 estimates to estimate the expected value of funding an intervention at random. I believe the correct way to do this would be to take the reciprocals of the cost-effectiveness estimates and then take their arithmetic mean, since that tells us the actual expected value of picking an intervention at random—which is a better analogue to what we're doing with animal interventions. That gives us a minimum estimate mean of $9.65/DALY and a maximum estimate of $64/DALY. Then the difference between the mean and the best intervention is only about 10x. This suggests that we should expect picking a factory farming intervention at random to have fairly high expected value, if we think animal interventions follow a similar distribution to the interventions in DCP2.

EDIT: I would expect on priors to find that animal interventions are at least 10x more effective than global poverty interventions in general*. If animal interventions do follow a distribution like the DCP2, that suggests that blindly picking an animal intervention should have similar or higher expected value than picking the best global poverty intervention.

*For a few related reasons, including: (1) we spend hundreds of billions of dollars on global poverty but less than $20 million on factory farming; (2) most people (implicitly) discount animals by several orders of magnitude more than they should; (3) there are >10 times more factory-farmed animals than there are humans in extreme poverty.

I take issue with the statement "it depends greatly on how much you value a human compared to a nonhuman animal". Similar things are often said by EAs in discussions of animal welfare. This makes it seem as if the value one places on nonhumans is a matter of taste, rather than a claim subject to rational argument. The statement should read "it depends greatly on how much we ought to value a human compared to a nonhuman".

Imagine if EAs went around saying "it depends on how much you value an African relative to an American". Maybe there is more reasonable uncertainty about between- as opposed to within-species comparisons, but still we demand good reasons for the value we assign to different kinds of humans. This idea is at the core of Effective Altruism. We ought to do the same with non-human sentients.