Hide table of contents

This article is completely outdated and does not represent my current views. A new cause selection article can be found on my blog.

This is my contribution to the EA blogging carnival on cause selection. I compare cause areas and attempt a quantitative comparison between LLIN distributions and advocacy for farmed animals. In short, I will continue to fundraise for malaria prevention but personally donate more to prioritization research in the area of animal advocacy. (Cross-posted from my blog.)

Since my selection of the cause area is highly dependent on the interventions and the implementing charities available in that cause area, this article will cover the complete stack of cause, intervention, and charity selection, which I will call CIC for short. I will also distinguish between CICs to fundraise for and such that I want to donate to directly since fundraising also requires that the CIC is marketable in my target demographic.

Preselection

Since I’m constrained in time, feel well value-aligned with our prioritization organizations, and don’t want to duplicate effort, I’m relying on GiveWell and Animal Charity Evaluators for a preselection. I’m looking forward to relying on Open Phil similarly, once the project publishes concrete recommendations. I’m also taking meta charities into account but struggle with the absence of the same preselection.

Thus my cause selection is down to NTDs (two charities), malaria, poverty, farmed animals (three charities), and various areas of EA meta activism.

Subselections

The cause areas fall into the category of those that affect mostly nonhuman animals, those that affect mostly humans, and the metacharities.

Nonhuman Animals

The Animal Charity Evaluators top charities all look great, and I have no hard and fast ideas for how to weigh them against each other. ACE’s very rough quantitative estimates are almost within the same order of magnitude, so I don’t expect the significant effort I’d have to invest to prioritize between them to be warranted. Some ideas I’ve had that might help if I wanted to do it after all were:

  1. The amount of meat consumption in the country the charity is active in, particularly of smaller animals like poultry and fish, based on the assumption that the marginal benefit of my investment may be lower in a country that is largely vegetarian already (India?).

  2. The rate of adoption of veg*nism in the country, based on the assumption that a higher rate may indicate that anticarnistic campaigns can gain good traction.

I also thought about taking the consumer price index into account, thinking that the charity can do more with my money there, but the differences between the countries that the three charities are active in were negligible.

Please note that ACE may have already done or may be working on such research.

What ACE is doing reminds me a lot of what Open Phil is doing in terms of how hard it is for them to quantify the impact of the charities. This doesn’t seem to be so much in the nature of the cause areas – as with prioritizing research or existential risk – but rather the result of the neglect of the area, so that ACE has only little and poor-quality prior research to draw on.

Imma once warned that neglectedness may be overexerted as proxy for marginal cost-effectiveness in areas where the latter is too hard to determine, but once again I see no other option. Hence ACE is also one of my candidates.

In the end Animal Equality, in addition to ACE, is ahead in terms of charities to donate to for the silly reasons that they are tax-exempt in Germany and have the highest quantitative estimate of the top charities, but I think I’ll rather make an unrestricted donation to ACE, which it can regrant if it’s somehow suddenly flooded with donations.

Mercy For Animals is ahead in terms of charities to fundraise for, because they have a fundraiser system on their website.

Humans

GiveDirectly is brilliant as certainty and scalability go, but those are important for very risk-averse donors and major foundations respectively. So long as I “only” have to allocate a few ten thousand euros, the cost-effectiveness is more important to me, and that is lower compared to the other top charities.

Evidence Action and SCI are top contenders, but I think the evidentiary basis for the effectiveness of deworming, while sizable, is more ambiguous than that for bednets, and more extreme estimates are also more likely to be wrong. So for as long as LLINs remain highly effective, I’ll rather invest in:

The Against Malaria Foundation. It seems to combine the advantages of having a solid evidentiary basis, being highly effective, and being well-run as an organization. Since it has a fundraiser system and is tax-exempt in a whole bunch of countries, it’s particularly good for fundraising for. I also hope to dedicate my master’s thesis to the topic of LLIN distributions – once I get any useful feedback on my proposal from my assessors.

GiveWell seems relatively less neglected than ACE, but that comparison doesn’t fit neatly into this section.

I’ll look into the standout charities in more detail once I either have more time or some of them have become top charities.

Metacharities

I’d love to invest in metacharities, but absent prioritization between them, I don’t know to which. I’m hoping that my expected utility auctions will alleviate that problem a little by crowdsourcing prioritization to some degree, but they’re still a thing of the future.

Some metacharities have estimated some leverage ratios, but they were always (as far as I’ve seen) the average ones of their past donations not of marginal donations, which may bias the estimate upward. (Except when the metacharity was just fundraising for operating expenses like Giving What We Can.) In some cases they didn’t take future impact into account either, which may bias the estimate downward if the metacharity had or has a lot of initial costs and is still increasingly gaining traction with its program. I’d much rather like to see a rough estimate what difference in scale my marginal donation can make a few years down the line and with what elasticity (say, $100,000 might achieve a lot, but $50,000 a lot less than half of that, because it would not be enough for a certain investment).

A focus on current leverage ratios may incentivize metachrities to defer important investments into the future, scaling up slower than they could, and thus deferring some of the alleviation of suffering that they could do into the future, allowing more dolors to accrue in total.

My intuition is that even investing in a random one of the EA meta charities I’ve looked at would be at least as good as an investment in a GiveWell top charity, so it would be fallacious to ignore the cause area just because I can’t decide between several great options, so outside small windows for investment like the GWWC fundraiser, this has been another reason for me to plan to invest in ACE in the future.

My Species Vs. Your Species

And then I have to somehow decide between humans and nonhuman animals. That’s really a bit unfair since I’m pitting one species against a whole lot of species, but at this point it seems much easier to help farmed animals than wild animals, which reduces the number of species a lot.

I attempt an estimation to compare the average ACE top charity to AMF in terms of the years of suffering that a donation averts. I’m lumping all ACE top charities together because I can’t meaningfully prioritize between them. I put the results in bold type in case you’re in a hurry.

The concept of the value of a life has been elusive to me, and when people try to estimate it by considering what impairments of health people accept for what increase in wage, the resulting estimates vary widely. The same surveys can’t be conducted among farmed animals, further complicating the picture. The multipliers that I assign intuitively vary widely, and I have trouble justifying them in ways that wouldn’t be speciesist. Hence I don’t want to rely on them for my prioritization and rather base it on well-being over time.

For clarification: Valuing well-being may in my case mean something along the lines of negative-leaning utilitarianism, though that can’t be the whole picture, and I perceive suffering as something that can be borne to some degree before it becomes increasingly unbearable, so that occasional dusts specks are perfectly bearable and don’t register for most people but torture is probably unbearable for most. I perceive whatever does the bearing like a muscle that can fatigue, recover from fatigue, and can possibly be trained (some sort of reverse hedonic adaptation). Proper philosophers have probably debated this view at length already, so I’d happily link any such debates if you know where to find them.

LLIN Distributions

GiveWell uses for its quantitative estimate of the cost-effectiveness of LLIN distributions the complete cost of the distribution including costs that AMF can leverage from partners such as Concern Universal or IMA. That is important to avoid lots of confusion about the attribution of impact. When I enter the historical average cost per net for AMF ($3.64) instead of the composite cost, the spreadsheet shows an average cost between $25 per undiscounted DALY averted and $64 per 3%-discounted DALY averted with uniform age weights. At the original composite costs, it returns cost between $44 and $111. Hence you can avert 9–40 DALYs per $1000.

However, GiveWell’s estimates only include lives saved. With 198 million cases of malaria in 2013 and 584,000 fatal ones, the nonfatal ones outnumber the fatal ones 339:1, so that the monetary and health benefits from the prevented nonfatal cases may be sizable. Plus I may value some degree of reduction of suffering over a year higher than the creation of another life year that is barely worth living.

If the YLL-to-YLD ratio is similar in the regions that AMF targets for their high malaria incidence and low insecticide resistance, then you avert one case of malaria for $6–10 (the costs per life saved for the two different net costs divided by 339). The WHO estimates that 2 million of the 198 million cases of malaria annually were severe cases, which have a disability weight of 0.21 as opposed to 0.005 of uncomplicated malaria. With a donation of $1000, you can thus avert another 100–166 cases of malaria or 0.7–1.2 DALY. This hardly changes the result above, but an estimate of the YLD averted due to saved treatment costs and prevented loss of income would be interesting.

Animal Advocacy

Animal Charity Evaluator confirmed that the estimates for all interventions that comprise the quantitative estimates of their top charities rely on the same assumptions about the number of animals not eaten, market elasticity, and recidivism. Hence I adjusted the donation in the Online Ad Impact Calculator so that the resulting number of animals saved came close to the top charity average of about 4769. The result was a reduction of 1700 years of suffering for $1000. These are not yet disability adjusted.

Bailey Norwood and Brian Tomasik have discussed quantitative estimates of farmed-animal suffering. I don’t have access to Norwood’s Compassion, By the Pound, and Tomasik writes that Norwood assigned a positive value of life to chickens farmed for meat, which he contests most convincingly on the basis of their short lives and brutal deaths (Norwood’s response), but Will MacAskill writes in Doing Good Better that Norwood’s “average rating for broiler chickens is -1” anyway. (Cows, farmed for meat and milk, account for only 30 of the 1700 years.)

I created a copy of the spreadsheet and undid the merging of cows farmed for meat and dairy and that of chickens farmed for meat and eggs, and introduced a new section to the spreadsheet to multiply in Norwood’s disability weights, which I divided by ten to put them in the same range as DALYs or QALYs. I’m not sure what to call this unit, but let’s say a donation of $1000 to the average ACE top charity averts between -36 and -2210 “QALYs,” with the best estimate at -283 “QALYs.”

These units may be incorrect and only comparable to a limited degree. The estimate doesn’t include wild-caught fish, whether caught intentionally or accidentally. Neither does it include environmental or health benefits, or flow-through effects of reduced recidivism through increased social acceptance or of increased visibility of veganism.

Finally, I don’t know whether it would be speciesist to value a human QALY one or two orders of magnitude higher than a nonhuman “QALY,” and would be tempted to just split my low-volatility budget between the CICs, but since I find it much harder to rally people in support for nonhuman animals than other humans, I want to invest as much of the money that is at my sole discretion where I can’t invest fundraised money.

Over the course of the past year I’ve become less risk averse in my selection, probably through the realization that we’re already diversifying our investment portfolio by having different opinions on what the most cost-effective interventions are and prioritizing neglected ones in the hopes of earning a higher marginal utility. Hence I don’t need to worry so much about the risk attached to my own investment and can rather aim for something closer to the highest expected value. This trend may well continue, particularly once Open Phil releases recommendations.

6

0
0

Reactions

0
0

More posts like this

Comments13
Sorted by Click to highlight new comments since: Today at 7:06 AM

What ACE is doing reminds me a lot of what Open Phil is doing in terms of how hard it is for them to quantify the impact of the charities. This doesn’t seem to be so much in the nature of the cause areas – as with prioritizing research or existential risk – but rather the result of the neglect of the area, so that ACE has only little and poor-quality prior research to draw on.

This seems correct.

Evidence Action and SCI are top contenders, but I think the evidentiary basis for the effectiveness of deworming, while sizable, is more ambiguous than that for bednets, and more extreme estimates are also more likely to be wrong.

True.

GiveWell seems relatively less neglected than ACE.

I'm not sure this is correct, unless our metric is just size?

Some metacharities have estimated some leverage ratios, but they were always (as far as I’ve seen) the average ones of their past donations not of marginal donations, which may bias the estimate upward.

True.

(Except when the metacharity was just fundraising for operating expenses like Giving What We Can.)

This isn't true - it hardly takes ~$350,000 a year to cover the most basic operating costs rather than marginal activities, especially when you already have ~$320,000 in the bank and your parent organisation has got almost a million in its coffers. One indication of this is that a very large share of Giving What We Can's past donations come from the initial (volunteer-led, unfunded) creation of a website where the case for giving more and giving better was made and people could report their pledges. But few of their salaries go to basic maintenance of this, as opposed to charity evaluation and philosophy research, conferences, etc.

I don’t know whether it would be speciesist to value a human QALY one or two orders of magnitude higher than a nonhuman “QALY”

It seems that it would.

Thanks for the comments.

I'm not sure this is correct, unless our metric is just size?

Size, in terms of staff, would be one measure. Another is their budget: GiveWell’s revenue and expenses were in the area of $2–3 million in 2014; Animal Charity Evaluators’ is about one order of magnitude lower.

This isn't true - it hardly takes ~$350,000 a year to cover the most basic operating costs rather than marginal activities, especially when you already have ~$320,000 in the bank and your parent organisation has got almost a million in its coffers.

Maybe someone from Giving What We Can knows more? The Fundraising Prospectus 2015 (page 16–17) talks about fundraising £150,000 for the rest of 2015 and 2016 with a stretch goal to add another six months. The Prospectus also only estimates the average cost-effectiveness of past donations (and I think they argue somewhere that it’s the result of their ongoing work to a substantial degree and not just very early seed funding or activities) and doesn’t say anything about the cost-effectiveness of donations that would allow them to scale up and expand beyond their current activities in the future. So either they really only fundraised for operating expenses or what they fundraised beyond that isn’t covered by their cost-effectiveness calculation.

Size, in terms of staff, would be one measure. Another is their budget: GiveWell’s revenue and expenses were in the area of $2–3 million in 2014; Animal Charity Evaluators’ is about one order of magnitude lower.

Those are both surely factors to take into account in evaluating how neglected something is, but other factors include how important and tractable the work (or the marginal work is). All of this lets you calculate out the value of marginal dollars.

The Fundraising Prospectus 2015 (page 16–17) talks about fundraising £150,000 for the rest of 2015 and 2016 with a stretch goal to add another six months.

Page 17 gives £223,765 as the budget for 2015, which is very roughly $350,000 (depending on how the UK Pound's swung).

I think they argue somewhere that it’s the result of their ongoing work to a substantial degree and not just very early seed funding or activities

I didn't find this convincing but I don't want to press that point: the more basic point is that $350,000 isn't necessary to cover the most basic operating costs, and some substantial portion of it does indeed go to fund additional marginal activities.

So either they really only fundraised for operating expenses or what they fundraised beyond that isn’t covered by their cost-effectiveness calculation.

You're exactly right; I'm simply saying it's the latter. It's not clear that we disagree!

All of this lets you calculate out the value of marginal dollars.

Yes, I’m optimistic about that in both cases. GiveWell is farther along thanks to so much more prior research and thanks to eight years of operation and scaling up, but then they’re also tackling a superset of what ACE is tackling, so I don’t know if I can compare them on these grounds. Hence I only looked at neglectedness, not the actual marginal cost-effectiveness.

Page 17 gives £223,765 as the budget for 2015, which is very roughly $350,000 (depending on how the UK Pound's swung).

Oh, thanks, I see now where the number came from!

I'm simply saying it's the latter.

Hmm, okay. If that’s the case, then I’m less satisfied with the calculations in the Prospectus than I used to be.

I am very late to this discussion, but the quote from William MacAskill that the “average rating for broiler chickens is -1” does not accurately represent Norwood's views.

Norwood gives his estimates on animal welfare for one breeder chicken (-4) and one market chicken (+3) for chicken meat (page 229). Norwood and Lusk also report how many breeder chickens (0.001804675) and how many market chickens (0.259740260) are associated with 1lb of chicken meat (page 233). Given the numbers provided, the weighted average of animal welfare for 1lb of chicken meat is:

(-4 0.001804675) + (3 0.259740260) = 0.77

So, based on Norwood's estimates, eating chicken meat increases animal welfare.

I don't know whether I agree with this, but that is the implication of Norwood's estimates, and that is clearly Norwood's intent. In fact, Norwood and Lusk write on page 241 that "Because Bailey [Norwood] believes chickens are overall happy animals, fewer chickens means less happiness in the world."

Avi

I forgot to follow-up in this earlier, but William MacAskill has issued a correction in his errata:

http://www.effectivealtruism.com/errata/

Yep! Thanks for pointing that out. I’ve read the book at this point, and by “average rating” Will seemingly means an unweighted average, which is a bit silly when one pound of chicken is associated with 0.001804675 breeder and 0.259740260 nonbreeder individuals, so 144 to 1 should be the weights, so that the breeder animal can be just about ignored in the calculation. Then again it’s also silly for Norwood not to take the slaughter into account for animals that live for 42 days before being deported to the slaughterhouse.

Whether Norwood’s conclusion follows is also dependent on whether one adhere’s to classic hedonistic utilitarianism or any other form of utilitarianism, e.g., preference utilitarianism. They touch on that in the context of the repugnant conclusion.

Norwood says he did take slaughter into account, but that he didn't think it affected animal welfare much relative to other factors. See his discussion with Simon Knutsson:

http://simonknutsson.com/files/exchange-bailey-norwood-2013-07-29-for-publication.docx

FWIW, if I had a choice whether to live as a broiler chicken and be slaughtered after 42 days or not exist at all, I think I'd prefer not to exist at all. OTOH, I think I'd prefer to live 2 years or so as a beef cow and then be slaughtered rather than not exist at all. So I feel a lot more comfortable eating beef than chicken. But I'm not completely comfortable because I'm not fully confident in this preference, and because I think "animal rights" arguments have some merit independent of utilitarian considerations.

Avi

Great conversation! I hadn’t seen that yet. Simon doesn’t ask about slaughter in the case of chickens again, unfortunately, because it seems like Bailey replied with other animals in mind than chickens farmed for meat. But his reply with regard to transport probably transfers to slaughter: “Yes, I did take into account the number of times they were transported relatively to their lifespan. It is hard for me to say exactly how much it influenced the scores, because they are just my judgment calls, but it is something I tried to account for, and something I definitely thought of.”

I thought I had read something to the effect that he didn’t factor in slaughter in the book itself, but I can’t find it anymore. What I may be remembering is Bailey’s comment on Brain’s old blog: “I do not make any adjustments for the length of the life for any animal. I did at first, but I found the difficulties and ambiguities unsatisfying.” I understood this as meaning that the chickens would receive the same score even if they lived for years, but Bailey may’ve meant something else.

And agreed. When I consider what life I would prefer to not living, I arrive at similar results, with considerably less uncertainty about my preference in the first than in the second case.

“I think ‘animal rights’ arguments have some merit independent of utilitarian considerations”: I tend to consider those from a two-level utilitarian perspective. For example, a society that is alert to speciesism will also be better equipped to avoid discrimination of sentient algorithms, what I’ve heard called substratism. Hard and fast, broad and imperfect principles can make it easier for many people to see violations.

Then consistency is also important to me, so when I’m comfortable with beef I also ought to be comfortable with a company serving a cannibal customer base by raising children in isolation, providing them with most things they desire including libraries for intellectual stimulation and so on, and then harvesting them at age 16–20 or so for their meat. I don’t think I would be entirely comfortable with that and rather prefer such a company to not exist, maybe because of the years of life stolen from them which to keep they already formed a preference. Not sure.

I don’t have access to Norwood’s Compassion, By the Pound,

You can download a copy here. :)

Oh, awesome! Thinking, Fast and Slow will have to wait.

Great analysis! I like the way you laid out your reasoning--it was easy for me to understand why you came to the conclusions you did.

You explained that you expect meta-charities like Giving What We Can to have a high impact, but you didn't say you planned on donating to them. Why not?

I was a bit confused by your conclusion--are you saying you think you should give to ACE-recommended charities, or to ACE itself? Do you believe a donation to ACE itself is more effective than donations to ACE top charities?

I’m counting ACE as metacharity, and I’m planning to donate to ACE itself. I wish I could present a good reason why I think that a (marginal) donation to ACE is even more effective than a donation to its top charities, but it’s in the nature of the beast, the startup, that there’s no solid track record of the effectiveness of its program, and that what track record there is surely still understates its potential.

I think that in the area of farmed animal advocacy:

  1. a lot more research into program cost-effectiveness is needed,
  2. greater awareness for the importance of this research is needed, and
  3. charities are still unsure whether an ACE recommendation is worth the time they’d have to invest in it.

My hope is that once ACE becomes notable enough that 3 becomes a nonissue for most charities, this will encourage 2 and lead to 1. Thus the top charities will become even better and the overall effectiveness waterline in the cause area will also rise as less effective charities at least become aware in which direction they should optimize. All the while donors will put more trust into ACE’s recommendations and thus more money into its top charities, leading to a virtuous circle.