Thanks for all your work on this, and asking me for feedback on your analysis, Stan!
My recommendation for donors
Your results imply the cost-effectiveness of ASRS policy advocacy is 0.242 DALY/$ (= 32.9*0.00737). I explain below why I think this too high, but, even if not, I estimate it is only 1.61 % (= 0.242/15.0) as cost-effective as corporate campaigns for chicken welfare, such as the ones supported by The Humane League (THL). So I think donors who value 1 unit of welfare in humans as much as 1 unit of welfare in animals (i.e. who reject speciesism) had better donate to THL instead of an organisation doing ASRS policy advocacy.
I would be curious to know CEARCH’s position on animal welfare. I noted there are 0 animal welfare causes in your long list of 588. Their absence is especially surprising given the presence of causes like sporting excellence and freedom of hobby.
Points of agreement
Here are some points I agree with from the sections “Key points” and “Executive Summary” of your report:
Policy advocacy, targeted at a few key countries, is the most promising way to increase resilience to global agricultural crises. Advocacy should focus on increasing the degree to which governments respond with effective food distribution measures, continued trade, and adaptations to the agricultural sector.
[...]
Compared to other interventions addressing Global Catastrophic Risk (GCR), the evidence is unusually robust:
We know that the threat is real: volcanic cooling is confirmed by the historical and geological record.
We know this is neglected: current food resilience policy focuses on protecting farmers and consumers from price changes, regional agricultural shortfalls or from small global shocks. There is very little being done to prepare for a significant global agricultural shortfall.
We are uncertain about the effect size: we have significant uncertainty about the extent to which governments and the international community would step up to the challenge of a global agricultural shortfall. There is little evidence on the scale of the effect that a policy breakthrough would have on the human response.
[...]
We identify two main sources of downside risk. (1) Increasing resilience to nuclear winter could reduce countries’ reluctance to use nuclear weapons. (2) nuclear winter resilience efforts could be seen by other nuclear-armed states as preparation for war, thereby increasing tensions. However, these risks are unlikely to apply to broader food resilience efforts.
[...]
Our analysis suggests that uncertainties about the validity of nuclear winter theory are often overlooked, leading some authors to overstate the risk of nuclear cooling [I agree]. [...]
[...]
However, because simple adaptations and international food trade & aid are sufficient to feed everyone in many scenarios, resilient foods would only be pivotal in averting mass famine if food trade is severely inhibited, or in the most severe cooling scenarios [agreed]. We believe that these very conditions [which involve large infrastructure given thousands of nuclear detonations] would likely restrict the scaling-up of resilient foods, which require a high degree of complexity (see Human Response for justification).
Points of disagreement
I believe your cost-effectiveness should be 12.4 % (= 1/3*0.505*0.736) as large based on the adjustments below, i.e. 4.08 (= 32.9*0.124) times as cost-effective as GiveWell’s top charities, or 0.200 % (= 0.0161*0.124) as cost-effective as corporate campaigns for chicken welfare. In other words, I conclude these are 500 (= 1⁄0.00200) times as cost-effective as ASRS policy advocacy.
Remaining regression to lower cost-effectiveness (I guess this makes your cost-effectiveness 1⁄3 as large)
Your cost-effectiveness has become around 10 % as large over the course of my review, so I guess there is still some remaining regression left which would make it 1⁄3 as large. In particular:
I suspect an advocacy campaign of 1 M$ leading to a reduction in the disease burden of cooling events of 0.272 % (or 0.35 % in your case) with 24.2 % probability is optimistic. To estimate this probability, you gave a weight of 76.9 % (= 70⁄91) to an expert survey, and experts tend to overestimate the cost-effectiveness of work in their area.
Relatedly, your report was reviewed by many people working to decrease nuclear and volcanic risk, which are naturally inclined to overestimate cost-effectiveness due to selection effects (see example in the context of AI risk), but not by people sceptical of such work?
By sceptical, I roughly mean people who think the interventions is less cost-effective than GiveWell’s top charities. My adjustments imply it is 4.08 times as cost-effective as them, so I do not qualify.
Did you have the chance to contact people at GiveWell, which I do not think has funded work specifically targeting cooling events?
I believe funging decreases the cost-effectiveness of your potential grants.
All grants in this space have the potential to be fungible. Both ALLFED [Alliance to Feed the Earth in Disasters] and RCG [Observatorio de Riesgos Catastróficos Globales] conduct research alongside policy advocacy, and so even a policy-restricted grant could displace other funding towards research and other work, which we believe to be less effective on the margin. Funding other GCR-focused orgs to perform ASRS resilience work could pull them away from highly-effective work in other domains.
We do not have strong beliefs about which funding options are least fungible: those considering making an ASRS policy grant should be careful to ensure that their donations lead to a counterfactual increase in advocacy efforts.
There would be no worries about funging if you considered all the activities of the supported organisation similarly cost-effective (at the margin). Conversely, even if you do not have strong beliefs about which funding options are least fungible, you can still be especially worried about funding organisations pursuing activities that you do not consider cost-effective.
In particular, your model implies the benefits coming from resilient foods, by which I think you mean ways of increasing calorie production via new (or massively scaled up) food sectors, are negligible as a 1st approximation. Assuming the policy breakthrough only affects resilient foods (not adaptation nor trade) makes the proportion of deaths averted by it 0.805 % (= 2.84*10^-5/0.00353) as large. I am sympathetic to this view, although I would guess a higher contribution, but it implies funding organisations doing research on or advocating for new (or massively scaled up) food sectors is less promising. Importantly:
Other organisations in the space may be significantly influenced by ALLFED’s work. As yousay in the report:
“The Alliance to Feed the Earth in Disasters (ALLFED) is the only organization dedicated to food resilience in global catastrophes, and very little seems to happen in this space without some sort of ALLFED involvement”.
“In such a small domain, this means that many of the experts in the field are either a member of, or have previously worked with, ALLFED. It also means that any new initiative in ASRS food resilience, if not run by ALLFED, would probably collaborate with them”.
Overestimation of the mortality of mild cooling events (my correction for this makes your cost-effectiveness 50.5 % as large)
I think your global mortality rate of 0.6 % for a cropland cooling in the worst 12 months of 1.5 ºC is way too high. Stoffel 2015, which you use in your volcanic winter model, suggests the 1815 eruption of Mount Tambora caused a cropland cooling of 1.05 ºC (= (0.8 + 1.3)/2).
Our [Stoffel 2015′s] tree-ring reconstructions and climate simulations are in agreement, with a mean Northern Hemisphere extra-tropical summer cooling over land of 0.8 to 1.3 ◦C for these eruptions [” in AD 1257 and 1815″]
In 1815, trade efficiency would arguably be negligible, and there would not be resilient foods, but there would still be adaptation, so I guess you would predict cropland coolings of 1.5 and 3.8 ºC then would have global mortality rates of 3 % and 5 %. Linearly extrapolating, I estimate you would predict a global mortality rate then of 2.61 % (= 0.03 + (0.05 − 0.03)/(3.8 − 1.5)*(1.05 − 1.5)), which would be 27.4 M global deaths (= 0.0261*1.05*10^9). Wikipedia’s section on the fatalities caused by the 1815 eruption of Mount Tambora suggests a much lower death toll (emphasis mine):
The number of fatalities has been estimated by various sources since the nineteenth century. Swiss botanist Heinrich Zollinger traveled to Sumbawa [Indonesian island] in 1847 and recollected witness accounts about the 1815 eruption of Tambora. In 1855, he published estimates of directly killed people at 10,100, mostly from pyroclastic flows. A further 37,825 were numbered having died from starvation on Sumbawa island.[39] On Lombok [Indonesian island], another 10,000 died from disease and hunger.[40] Petroeschevsky (1949) estimated that about 48,000 and 44,000 people were killed on Sumbawa and Lombok, respectively.[41] Several authors have used Petroeschevsky’s figures, such as Stothers (1984), who estimated 88,000 deaths in total.[29] However, Tanguy et al. (1998) considered Petroeschevsky’s figures based on untraceable sources, so developed an estimate based solely on two primary sources: Zollinger, who spent several months on Sumbawa after the eruption, and the notes of Sir Stamford Raffles,[30] Governor-General of the Dutch East Indies during the event. Tanguy pointed out that there may have been additional victims on Bali [province of Indonesia] and East Java [Java is Indonesia’s capital city] because of famine and disease, and estimated 11,000 deaths from direct volcanic action and 49,000 from post-eruption famine and epidemics.[42] Oppenheimer (2003) estimated at least 71,000 deaths,[3] and numbers as high as 117,000 have been proposed.[36]
All the estimates above refer to deaths in Indonesia, where Mount Tambora is located. Local effects are more severe than global ones, so dividing the above estimates by Indonesia’s population in 1815 would overestimate the global mortality rate. Wikipedia’s list of famines says Tambora’s eruption caused 65 k deaths in Europe (see below), i.e. 0.0304 % (= 65*10^3/(214*10^6)) of Europe’s population in 1815.
I would say the aforementioned 65 k deaths should not be fully attributed to Tambora’s eruption. From Wikipedia’s page on the Year Without the Summer, respecting the volcanic winter caused by Tambora’s eruption, they were also the result of earlier volcanic eruptions and the Napoleonic Wars, which ended in 1815:
As a result of the series of volcanic eruptions in the 1810s, crops had been poor for several years; the final blow came in 1815 with the eruption of Tambora. Europe, still recuperating from the Napoleonic Wars, suffered from widespread food shortages, resulting in its worst famine of the century.[22][23][24][25] Low temperatures and heavy rains resulted in failed harvests in Great Britain and Ireland. Famine was prevalent in north and southwest Ireland, following the failure of wheat, oat, and potato harvests. Food prices rose sharply throughout Europe.[26] With the cause of the problems unknown, hungry people demonstrated in front of grain markets and bakeries. Food riots took place in many European cities. Though riots were common during times of hunger, the food riots of 1816 and 1817 were the most violent period on the continent since the French Revolution.[23]
Between 1816 and 1819, major typhus epidemics occurred in parts of Europe, including Ireland, Italy, Switzerland, and Scotland, precipitated by the famine. More than 65,000 people died as the disease spread out of Ireland.[22][23]
I speculate only half of the 65 k deaths were caused by Tambora’s eruption, i.e. 32.5 k (= 0.5*65*10^3). However, I guess this only accounts for the effects of protein-energy malnutrition (the more visible starvation), whose mortality in 2019 was only 7.21 % (= 212*10^3/(2.94*10^6)) of that from child and maternal malnutrition. So I estimate Tambora’s eruption caused 451 k (= 32.5*10^3/0.0721) deaths in Europe (accounting for non-reported deaths), respecting a mortality rate of 0.211 % (= 451*10^3/(214*10^6)). This is 7.03 % (= 0.00211/0.03) of the 3 % I understand your model would predict.
The death rate above is for an eruption 209 years (= 2024 − 1815) ago. Poverty has been a major risk factor for famines, and the global real gross domestic product (real GDP) per capita in 2022 was 14.8 (= 16.7*10^3/(1.13*10^3)) times that in 1820, so I think Tambora’s eruption today would be way less deadly. Here are the death rate from protein-energy malnutrition (arguably proportional to the death rate from child and maternal malnutrition) and real GDP per capita in 2017-$ by country:
Eyeballing the graph above, the death rate from protein-energy malnutrition weighted by population is:
For the real GDP per capita in 1820 of 1.30 k 2017-$, 2*10^-4.
For the real GDP per capita in 2022 of 17.5 k 2017-$, 1*10^-5.
So I guess the increase in death rate caused by Tambora’s eruption today would be 5 % (= 1*10^-5/(2*10^-4)) of my estimate for 1815 of 0.211 %, i.e. 0.0106 % (= 0.05*0.00211). For context, the deaths from child and maternal malnutrition in 2019 as a fraction of the global population were 0.0326 %. So I predict the increase in death rate caused by Tambora’s eruption today would correspond to 32.5 % (= 1.06*10^-4/(3.26*10^-4)) of that. This sounds reasonable:
I estimate global cooling is 54.2 % (= 1.3/2.4) of cropland cooling:
From Fig. 3a of Toon 2014, the maximum global cooling for 1.3 ºC for 5 Tg.
So Tambora’s 1.05 ºC of cropland cooling would respect a global cooling of only 0.569 ºC (= 0.542*1.05).
Global temperature often has annual variations about half as large as the above, so I would be surprised if it increased famine a lot.
You estimate cropland coolings of 1.5 and 3.8 ºC today would have global mortality rates of 0.6 % and 3.3 %. Linearly extrapolating, I estimate you would predict a global mortality rate today of 0.0717 % (= 0.006 + (0.033 − 0.006)/(3.8 − 1.5)*(1.05 − 1.5)). So, in light of the above, I would say your mortality rate for a cropland cooling of 1.05 ºC should be 14.8 % (= 1.06*10^-4/(7.17*10^-4)) as high.
I suppose the mortality rate adjustment factor increases linearly with cropland cooling, from the 14.8 % mentioned just above for 1.05 ºC, to 1 (no adjustment) for 8 ºC, which respects an injection of soot into the stratosphere of 47 Tg. From Fig. 5a of Xia 2022, this is roughly the amount of soot below which there are enough calories to feed everyone given equitable distribution, no household food waste, no inefficient consumption of animals, and no other adaptations. So I multiplied your mortality rates for a cropland cooling of:
For 8 and 14.5 ºC (the most severe cooling you considered), I used your values. My updated mortality rates imply a proportion of deaths averted by a policy breakthrough of 0.277 %, and a global mortality rate for a cooling event of 0.819 %. These make my tractability 78.6 % (= 6.72*10^-5/(8.55*10^-5)) as large as yours, and my expected burden from cooling events in 2024 64.2 % (= 7.26*10^6/(1.13*10^7)) as large as yours. So my updated mortality rates lead to a cost-effectiveness 50.5 % (= 0.786*0.642) as large as yours.
Overestimation of the persistence of the intervention (my correction for this makes your cost-effectiveness 73.6 % as large)
You estimate an annual reduction in food security from 2024 to 2100 of 1.76 %, which is the geometric mean between:
3.24 % annual reduction in the disease burden of nutritional deficiencies from 1990 to 2019.
1.45 % annual reduction in undernourishment from 2000 to 2022.
0.76 % annual increase in cereal production per capita from 1961 to 2022.
2.67 % annual reduction in extreme poverty from 1961 to 2018.
I think it is better to rely on the 1st and last of the above, and the annual reduction in the disease burden of child and maternal malnutrition from 1990 to 2019 of 2.76 % (= 1 - (295/665)^(1/29)). So I estimate an annual reduction in food security from 2024 to 2100 of 2.88 % (= (0.0324*0.0267*0.0276)^(1/3)), which makes the persistence and cost-effectiveness of the invervention 73.6 % (= 18.4/25) as large. Using the disease burden of nutritional deficiencies, and of child and maternal malnutrition makes sense to me given they are the cause and risk of the Global Burden of Disease Study (GBD) more closely connected to what you are predicting. I also agree with including extreme poverty for the 3 reasons below.
Firstly, the share of the population in extreme poverty has been a better predictor of the death rate from protein-energy malnutrition (R^2 of 0.941 for 30 points) than the share of the population that is undernourished (R^2 of 0.82 for 20 points) and cereal production per capita (R^2 of 0.297 for 30 points). I guess this would continue to hold for larger food shocks. Note the 2nd R^2 would tend to be lower if it referred to 30 points as the other 2. The graphs are below.
Secondly, poverty has been a major risk factor for famines.
Thirdly, we should expect people in extreme poverty to be especially vulnerable to increases in food prices on 1st principles. A calorie sufficient diet in low income countries in 2017 costed 0.86 2017-$/d/person, i.e. 40.0 % (= 0.86/2.15) of the maximum income of someone in extreme poverty. In contrast, someone earning e.g. 20 2017-$/h, or 160 2017-$/d for 8 h/d, can afford the same diet with just 0.538 % (= 0.86/160) of income. If food prices triple such that satisfying the caloric requirement requires 2.58 2017-$/d (= 0.86*3), someone earning 2.15 2017-$/d would only be able to afford 83.3 % (= 2.15/2.58) of the calorie sufficient diet even spending all income on it, so the person would be in trouble. In contrast, someone earning 160 2017-$/d would be able to afford the same diet with just 1.61 % (= 2.58/160) of income, so the person would be fine.
Hi Vasco. Thanks for all of your help giving feedback on the report and the modeling underpinning the CEA.
I am going to focus on the main points that you make. I hope to explain why I chose not to adopt the changes you mention in your comment and also to highlight some key weaknesses and limitations of my model.
Points I address (paraphrasing what you said):
By asking domain experts you probably got an overestimate for “probability that advocacy succeeds”. You should have also asked people in other fields.
Although you mention fungibility, you don’t account for it in cost-effectiveness estimates. You should be more explicit that fungibility undermines cost-effectiveness of grants that perform other, less effective interventions.
You overestimate the mortality effects of mild cooling events: if we apply your model to the 1815 eruption, we get higher mortality rates than actually occurred.
Poverty is a strong predictor of famine mortality. If your persistence estimate relied only on poverty & malnutrition burden indicators, the full-term benefits of policy advocacy would be significantly lower
By asking domain experts you probably got an overestimate for “probability that advocacy succeeds”. You should have also asked people in other fields.
I agree that domain experts are likely to overestimate the probability of successful policy advocacy in their space.
In my defence, only two of the seven experts I consulted for estimates worked in food resilience specifically. The geometric mean of their estimates was 30%; only slightly higher than the group average of 24%. The other experts would be best classified as GCR experts (so still likely to be overly optimistic)
The difficulty is that people who are not domain experts are (by definition) not well-informed. I don’t think people at GiveWell will have an accurate understanding of prospects for ASRS resilience policy advocacy. Especially because this is a very small field.
Although you mention fungibility, you don’t account for it in cost-effectiveness estimates. You should be more explicit that fungibility undermines cost-effectiveness of grants that perform other, less effective interventions.
I think that in ideal circumstances, fungibility should be accounted for in cost-effectiveness analysis. But since it depends on the organization receiving the funding, I decided not to do quantitative estimates of fungibility effects in this report. Maybe we will do so when we evaluate specific grants in this area.
I agree that funding to orgs who only do one highly cost-effective thing is generally less fungible.
You overestimate the mortality effects of mild cooling events: if we apply your model to the 1815 eruption, we get higher mortality rates than actually occurred.
I love this analysis, thanks for doing it.
First, let me say that yes, my model is very sensitive to mortality estimates in mild cooling scenarios. My estimate may be too high, but I believe there are compelling reasons not to be confident of this.
To illustrate my model for mortality in a mild cooling event (1-2.65 degrees cooling):
2% probability of 6% mortality (no adaptation, no food trade)
18% probability of 3% mortality (no food trade)
80% probability of no mortality
This gives an average of approximately 0.6% mortality
My counterarguments are as follows:
On a broad level, I think that a ‘panic’ scenario could include countries banning food exports to secure domestic supplies. This would be catastrophic for some food-importing countries even in normal climate conditions. The same cannot be said for the world of 1815, where almost all food was consumed locally and very few people lived far from agricultural areas.
I think the comparison with 1815 is well worth doing. However, there are a number of reasons why the validity of the comparison is limited:
A global 1% mortality event in 1815 may not have even been noticed. We have to patch together estimates of famine mortality in 1815 because there was almost no systematic documentation at the time. People were not aware of any global phenomenon, so nobody was trying to “join up the dots” and tally the full famine impact that year. Especially if much of the effects were felt in South Asia, East Asia or Africa, it is possible that a major-yet-distributed famine could have gone unnoticed
Relatively few people in the world of 1815 relied on food imports, as mentioned above[1]. Breakdown in international food trade is the main famine mechanism in modern-day agricultural catastrophes, but it barely applies to the agrarian economies of 1815.
Local famine effects may not have been worse in Indonesia. Undoubtedly, the effects of ash blanketing crops would have been worse near the eruption. But stratospheric soot quickly circulates around the world. Furthermore, Indonesia has a warm climate and would not have been at risk from completely failed harvests through unseasonal frost. Multiple annual harvests are common in the tropics; in higher latitudes, a failed crop leaves farmers without food for almost a year.
Cooling damage is highly superlinear. The Pinatubo eruption of 1991 caused 0.5 degrees of cooling and is not associated with important declines in agricultural productivity. Thus we might expect the expected burden of an 1815-level cooling event (0.8 to 1.3 degrees cooling) to be far lower than a 1-2.65 degree cooling event[2].
Poverty is a strong predictor of famine mortality. If your persistence estimate relied only on poverty & malnutrition burden indicators, the full-term benefits of policy advocacy would be significantly lower.
To push back:
A world with no extreme poverty and no nutritional diseases would still be vulnerable to global agricultural catastrophes. The advantage of my “grain production per capita” metric is that it has no such edge-case problems: more grain is always good for food security.
The famine/poverty relationship is strong in normal times. There are a number of reasons that it may become less strong in an agricultural catastrophe
Many of the global poor are subsistence farmers in warm countries. This demographic will be close to food supplies and away from the risk of frosts etc.. They may be better-positioned to survive than their middle-class compatriots in cities.
People at the bottom of the pile are the most at-risk of post-catastrophe famine, regardless of whether they are in absolute poverty. A wealthier world would simply have higher food prices, leaving the poorest without enough to eat.
As described above, the main cause of famine in my model is the breakdown of food trade. It is not clear that progress against poverty and nutritional diseases is an indication that populations are less dependent on food trade. If anything, development is probably associated with increased reliance on trade.
A counterargument could be that Western Europe appears to have had particularly bad summer cooling in 1816 - as well as better record-keeping than much of the world—and their famines were not so bad. On the other hand, spring cooling may be more important, as late frosts can ruin harvests of wheat, potatoes etc.
Thanks for the clarifying comment, Stan! I strongly upvoted it.
As a preliminary note, I think it makes a lot of sense to give feedback on analyses like yours privately, but I wonder whether it is worth for me to invest significant time in writing comments like mine above. It seems that they are often downvoted, and that I can sometimes tell before hand when this is going to be case. So, to the extent karma is a good proxy for what people value, I wonder whether I am just spending signicant time on doing something which has little value. In this particular case, I am still guessing it was worth it because, even if it had negligible value to the public, it was still relevant for my own cause prioritisation (and making it public had little cost).
For what is worth, I was already aware of the arguments you mentioned, and directionally agree with all the points you make. I just think their effect is not as strong as you do, so I maintain my adjustments are warranted.
In any case:
[...] even if not, I estimate it [ASRS policy advocacy] is only 1.61 % (= 0.242/15.0) as cost-effective as corporate campaigns for chicken welfare, such as the ones supported by The Humane League (THL). So I think donors who value 1 unit of welfare in humans as much as 1 unit of welfare in animals (i.e. who reject speciesism) had better donate to THL instead of an organisation doing ASRS policy advocacy.
I would be curious to know CEARCH’s position on animal welfare. I noted there are 0 animal welfare causes in your long list of 588. Their absence is especially surprising given the presence of causes like sporting excellence and freedom of hobby.
Could you comment on the 2 points above?
A global 1% mortality event in 1815 may not have even been noticed.
Agreed, so I adjusted for underreporting in my calculations. I considered the actual mortality to be 13.9 (= 1⁄0.0721) times as high as the reported one.
Cooling damage is highly superlinear.
Agreed, so I adjusted less strongly your mortality rates for more severe coolings:
I wonder whether it is worth for me to invest significant time in writing comments like mine above. It seems that they are often downvoted, and that I can sometimes tell before hand when this is going to be case. So, to the extent karma is a good proxy for what people value, I wonder whether I am just spending signicant time on doing something which has little value.
I am sad to see your comment getting downvotes as I do think it contributes a lot of value to the discussion.
I can guess why you might be getting them. You often respond to cause-prio posts with “what about corporate campaigns for chicken welfare?”, and many people now probably switch off and downvote when they see this. Maybe just keep the chicken comparison to one line and link to your original post/comment?
Also, you comment is 3200 words long—about 3x longer than the actual post. I think a 200-word summary-of-the-comment with bullet points would be really useful for readers who have only read this post and are unable to pick up the finer points of your modeling critique.
On animal welfare
I think that if you adopt RP’s moral weight estimates and reject speciesism, it is almost inevitable that the most cost-effective interventions to improve wellbeing will be animal welfare interventions.
My understanding is that CEARCH is not against evaluating animal welfare interventions in principle, but in practice we are not doing so while comparisons between human and animal welfare remain so shaky. Our research direction is also partly driven by the value of information, ie. how much resources we can plausibly redirect and the impact this will have. Maybe this is too deterministic of me, but I feel that banging the drum about corporate chicken campaigns will only open so many wallets.
Thanks for all your work on this, and asking me for feedback on your analysis, Stan!
My recommendation for donors
Your results imply the cost-effectiveness of ASRS policy advocacy is 0.242 DALY/$ (= 32.9*0.00737). I explain below why I think this too high, but, even if not, I estimate it is only 1.61 % (= 0.242/15.0) as cost-effective as corporate campaigns for chicken welfare, such as the ones supported by The Humane League (THL). So I think donors who value 1 unit of welfare in humans as much as 1 unit of welfare in animals (i.e. who reject speciesism) had better donate to THL instead of an organisation doing ASRS policy advocacy.
I would be curious to know CEARCH’s position on animal welfare. I noted there are 0 animal welfare causes in your long list of 588. Their absence is especially surprising given the presence of causes like sporting excellence and freedom of hobby.
Points of agreement
Here are some points I agree with from the sections “Key points” and “Executive Summary” of your report:
Points of disagreement
I believe your cost-effectiveness should be 12.4 % (= 1/3*0.505*0.736) as large based on the adjustments below, i.e. 4.08 (= 32.9*0.124) times as cost-effective as GiveWell’s top charities, or 0.200 % (= 0.0161*0.124) as cost-effective as corporate campaigns for chicken welfare. In other words, I conclude these are 500 (= 1⁄0.00200) times as cost-effective as ASRS policy advocacy.
Remaining regression to lower cost-effectiveness (I guess this makes your cost-effectiveness 1⁄3 as large)
Your cost-effectiveness has become around 10 % as large over the course of my review, so I guess there is still some remaining regression left which would make it 1⁄3 as large. In particular:
I suspect an advocacy campaign of 1 M$ leading to a reduction in the disease burden of cooling events of 0.272 % (or 0.35 % in your case) with 24.2 % probability is optimistic. To estimate this probability, you gave a weight of 76.9 % (= 70⁄91) to an expert survey, and experts tend to overestimate the cost-effectiveness of work in their area.
Relatedly, your report was reviewed by many people working to decrease nuclear and volcanic risk, which are naturally inclined to overestimate cost-effectiveness due to selection effects (see example in the context of AI risk), but not by people sceptical of such work?
By sceptical, I roughly mean people who think the interventions is less cost-effective than GiveWell’s top charities. My adjustments imply it is 4.08 times as cost-effective as them, so I do not qualify.
Did you have the chance to contact people at GiveWell, which I do not think has funded work specifically targeting cooling events?
I believe funging decreases the cost-effectiveness of your potential grants.
On the last point, you say in the report that:
There would be no worries about funging if you considered all the activities of the supported organisation similarly cost-effective (at the margin). Conversely, even if you do not have strong beliefs about which funding options are least fungible, you can still be especially worried about funding organisations pursuing activities that you do not consider cost-effective.
In particular, your model implies the benefits coming from resilient foods, by which I think you mean ways of increasing calorie production via new (or massively scaled up) food sectors, are negligible as a 1st approximation. Assuming the policy breakthrough only affects resilient foods (not adaptation nor trade) makes the proportion of deaths averted by it 0.805 % (= 2.84*10^-5/0.00353) as large. I am sympathetic to this view, although I would guess a higher contribution, but it implies funding organisations doing research on or advocating for new (or massively scaled up) food sectors is less promising. Importantly:
A significant fraction of ALLFED’s work has been research on new (or massively scaled up) food sectors, like greenhouse crop production, lignocellulosic sugar, methane single cell protein and seaweed.
Other organisations in the space may be significantly influenced by ALLFED’s work. As you say in the report:
“The Alliance to Feed the Earth in Disasters (ALLFED) is the only organization dedicated to food resilience in global catastrophes, and very little seems to happen in this space without some sort of ALLFED involvement”.
“In such a small domain, this means that many of the experts in the field are either a member of, or have previously worked with, ALLFED. It also means that any new initiative in ASRS food resilience, if not run by ALLFED, would probably collaborate with them”.
Overestimation of the mortality of mild cooling events (my correction for this makes your cost-effectiveness 50.5 % as large)
I think your global mortality rate of 0.6 % for a cropland cooling in the worst 12 months of 1.5 ºC is way too high. Stoffel 2015, which you use in your volcanic winter model, suggests the 1815 eruption of Mount Tambora caused a cropland cooling of 1.05 ºC (= (0.8 + 1.3)/2).
In 1815, trade efficiency would arguably be negligible, and there would not be resilient foods, but there would still be adaptation, so I guess you would predict cropland coolings of 1.5 and 3.8 ºC then would have global mortality rates of 3 % and 5 %. Linearly extrapolating, I estimate you would predict a global mortality rate then of 2.61 % (= 0.03 + (0.05 − 0.03)/(3.8 − 1.5)*(1.05 − 1.5)), which would be 27.4 M global deaths (= 0.0261*1.05*10^9). Wikipedia’s section on the fatalities caused by the 1815 eruption of Mount Tambora suggests a much lower death toll (emphasis mine):
All the estimates above refer to deaths in Indonesia, where Mount Tambora is located. Local effects are more severe than global ones, so dividing the above estimates by Indonesia’s population in 1815 would overestimate the global mortality rate. Wikipedia’s list of famines says Tambora’s eruption caused 65 k deaths in Europe (see below), i.e. 0.0304 % (= 65*10^3/(214*10^6)) of Europe’s population in 1815.
I would say the aforementioned 65 k deaths should not be fully attributed to Tambora’s eruption. From Wikipedia’s page on the Year Without the Summer, respecting the volcanic winter caused by Tambora’s eruption, they were also the result of earlier volcanic eruptions and the Napoleonic Wars, which ended in 1815:
I speculate only half of the 65 k deaths were caused by Tambora’s eruption, i.e. 32.5 k (= 0.5*65*10^3). However, I guess this only accounts for the effects of protein-energy malnutrition (the more visible starvation), whose mortality in 2019 was only 7.21 % (= 212*10^3/(2.94*10^6)) of that from child and maternal malnutrition. So I estimate Tambora’s eruption caused 451 k (= 32.5*10^3/0.0721) deaths in Europe (accounting for non-reported deaths), respecting a mortality rate of 0.211 % (= 451*10^3/(214*10^6)). This is 7.03 % (= 0.00211/0.03) of the 3 % I understand your model would predict.
The death rate above is for an eruption 209 years (= 2024 − 1815) ago. Poverty has been a major risk factor for famines, and the global real gross domestic product (real GDP) per capita in 2022 was 14.8 (= 16.7*10^3/(1.13*10^3)) times that in 1820, so I think Tambora’s eruption today would be way less deadly. Here are the death rate from protein-energy malnutrition (arguably proportional to the death rate from child and maternal malnutrition) and real GDP per capita in 2017-$ by country:
Eyeballing the graph above, the death rate from protein-energy malnutrition weighted by population is:
For the real GDP per capita in 1820 of 1.30 k 2017-$, 2*10^-4.
For the real GDP per capita in 2022 of 17.5 k 2017-$, 1*10^-5.
So I guess the increase in death rate caused by Tambora’s eruption today would be 5 % (= 1*10^-5/(2*10^-4)) of my estimate for 1815 of 0.211 %, i.e. 0.0106 % (= 0.05*0.00211). For context, the deaths from child and maternal malnutrition in 2019 as a fraction of the global population were 0.0326 %. So I predict the increase in death rate caused by Tambora’s eruption today would correspond to 32.5 % (= 1.06*10^-4/(3.26*10^-4)) of that. This sounds reasonable:
I estimate global cooling is 54.2 % (= 1.3/2.4) of cropland cooling:
From Fig. 1a of Xia 2022, the maximum cropland cooling is 2.4 ºC for 5 Tg.
From Fig. 3a of Toon 2014, the maximum global cooling for 1.3 ºC for 5 Tg.
So Tambora’s 1.05 ºC of cropland cooling would respect a global cooling of only 0.569 ºC (= 0.542*1.05).
Global temperature often has annual variations about half as large as the above, so I would be surprised if it increased famine a lot.
You estimate cropland coolings of 1.5 and 3.8 ºC today would have global mortality rates of 0.6 % and 3.3 %. Linearly extrapolating, I estimate you would predict a global mortality rate today of 0.0717 % (= 0.006 + (0.033 − 0.006)/(3.8 − 1.5)*(1.05 − 1.5)). So, in light of the above, I would say your mortality rate for a cropland cooling of 1.05 ºC should be 14.8 % (= 1.06*10^-4/(7.17*10^-4)) as high.
I suppose the mortality rate adjustment factor increases linearly with cropland cooling, from the 14.8 % mentioned just above for 1.05 ºC, to 1 (no adjustment) for 8 ºC, which respects an injection of soot into the stratosphere of 47 Tg. From Fig. 5a of Xia 2022, this is roughly the amount of soot below which there are enough calories to feed everyone given equitable distribution, no household food waste, no inefficient consumption of animals, and no other adaptations. So I multiplied your mortality rates for a cropland cooling of:
1.5 ºC by 20.3 % (= 0.148 + (1 − 0.148)/(8 − 1.05)*(1.5 − 1.05)).
3.8 ºC by 48.5 % (= 0.148 + (1 − 0.148)/(8 − 1.05)*(3.8 − 1.05)).
5.5 ºC by 69.4 % (= 0.148 + (1 − 0.148)/(8 − 1.05)*(5.5 − 1.05)).
7 ºC by 87.7 % (= 0.148 + (1 − 0.148)/(8 − 1.05)*(7 − 1.05)).
For 8 and 14.5 ºC (the most severe cooling you considered), I used your values. My updated mortality rates imply a proportion of deaths averted by a policy breakthrough of 0.277 %, and a global mortality rate for a cooling event of 0.819 %. These make my tractability 78.6 % (= 6.72*10^-5/(8.55*10^-5)) as large as yours, and my expected burden from cooling events in 2024 64.2 % (= 7.26*10^6/(1.13*10^7)) as large as yours. So my updated mortality rates lead to a cost-effectiveness 50.5 % (= 0.786*0.642) as large as yours.
Overestimation of the persistence of the intervention (my correction for this makes your cost-effectiveness 73.6 % as large)
You estimate an annual reduction in food security from 2024 to 2100 of 1.76 %, which is the geometric mean between:
3.24 % annual reduction in the disease burden of nutritional deficiencies from 1990 to 2019.
1.45 % annual reduction in undernourishment from 2000 to 2022.
0.76 % annual increase in cereal production per capita from 1961 to 2022.
2.67 % annual reduction in extreme poverty from 1961 to 2018.
I think it is better to rely on the 1st and last of the above, and the annual reduction in the disease burden of child and maternal malnutrition from 1990 to 2019 of 2.76 % (= 1 - (295/665)^(1/29)). So I estimate an annual reduction in food security from 2024 to 2100 of 2.88 % (= (0.0324*0.0267*0.0276)^(1/3)), which makes the persistence and cost-effectiveness of the invervention 73.6 % (= 18.4/25) as large. Using the disease burden of nutritional deficiencies, and of child and maternal malnutrition makes sense to me given they are the cause and risk of the Global Burden of Disease Study (GBD) more closely connected to what you are predicting. I also agree with including extreme poverty for the 3 reasons below.
Firstly, the share of the population in extreme poverty has been a better predictor of the death rate from protein-energy malnutrition (R^2 of 0.941 for 30 points) than the share of the population that is undernourished (R^2 of 0.82 for 20 points) and cereal production per capita (R^2 of 0.297 for 30 points). I guess this would continue to hold for larger food shocks. Note the 2nd R^2 would tend to be lower if it referred to 30 points as the other 2. The graphs are below.
Secondly, poverty has been a major risk factor for famines.
Thirdly, we should expect people in extreme poverty to be especially vulnerable to increases in food prices on 1st principles. A calorie sufficient diet in low income countries in 2017 costed 0.86 2017-$/d/person, i.e. 40.0 % (= 0.86/2.15) of the maximum income of someone in extreme poverty. In contrast, someone earning e.g. 20 2017-$/h, or 160 2017-$/d for 8 h/d, can afford the same diet with just 0.538 % (= 0.86/160) of income. If food prices triple such that satisfying the caloric requirement requires 2.58 2017-$/d (= 0.86*3), someone earning 2.15 2017-$/d would only be able to afford 83.3 % (= 2.15/2.58) of the calorie sufficient diet even spending all income on it, so the person would be in trouble. In contrast, someone earning 160 2017-$/d would be able to afford the same diet with just 1.61 % (= 2.58/160) of income, so the person would be fine.
Hi Vasco. Thanks for all of your help giving feedback on the report and the modeling underpinning the CEA.
I am going to focus on the main points that you make. I hope to explain why I chose not to adopt the changes you mention in your comment and also to highlight some key weaknesses and limitations of my model.
Points I address (paraphrasing what you said):
By asking domain experts you probably got an overestimate for “probability that advocacy succeeds”. You should have also asked people in other fields.
Although you mention fungibility, you don’t account for it in cost-effectiveness estimates. You should be more explicit that fungibility undermines cost-effectiveness of grants that perform other, less effective interventions.
You overestimate the mortality effects of mild cooling events: if we apply your model to the 1815 eruption, we get higher mortality rates than actually occurred.
Poverty is a strong predictor of famine mortality. If your persistence estimate relied only on poverty & malnutrition burden indicators, the full-term benefits of policy advocacy would be significantly lower
By asking domain experts you probably got an overestimate for “probability that advocacy succeeds”. You should have also asked people in other fields.
I agree that domain experts are likely to overestimate the probability of successful policy advocacy in their space.
In my defence, only two of the seven experts I consulted for estimates worked in food resilience specifically. The geometric mean of their estimates was 30%; only slightly higher than the group average of 24%. The other experts would be best classified as GCR experts (so still likely to be overly optimistic)
The difficulty is that people who are not domain experts are (by definition) not well-informed. I don’t think people at GiveWell will have an accurate understanding of prospects for ASRS resilience policy advocacy. Especially because this is a very small field.
Although you mention fungibility, you don’t account for it in cost-effectiveness estimates. You should be more explicit that fungibility undermines cost-effectiveness of grants that perform other, less effective interventions.
I think that in ideal circumstances, fungibility should be accounted for in cost-effectiveness analysis. But since it depends on the organization receiving the funding, I decided not to do quantitative estimates of fungibility effects in this report. Maybe we will do so when we evaluate specific grants in this area.
I agree that funding to orgs who only do one highly cost-effective thing is generally less fungible.
You overestimate the mortality effects of mild cooling events: if we apply your model to the 1815 eruption, we get higher mortality rates than actually occurred.
I love this analysis, thanks for doing it.
First, let me say that yes, my model is very sensitive to mortality estimates in mild cooling scenarios. My estimate may be too high, but I believe there are compelling reasons not to be confident of this.
To illustrate my model for mortality in a mild cooling event (1-2.65 degrees cooling):
2% probability of 6% mortality (no adaptation, no food trade)
18% probability of 3% mortality (no food trade)
80% probability of no mortality
This gives an average of approximately 0.6% mortality
My counterarguments are as follows:
On a broad level, I think that a ‘panic’ scenario could include countries banning food exports to secure domestic supplies. This would be catastrophic for some food-importing countries even in normal climate conditions. The same cannot be said for the world of 1815, where almost all food was consumed locally and very few people lived far from agricultural areas.
I think the comparison with 1815 is well worth doing. However, there are a number of reasons why the validity of the comparison is limited:
A global 1% mortality event in 1815 may not have even been noticed. We have to patch together estimates of famine mortality in 1815 because there was almost no systematic documentation at the time. People were not aware of any global phenomenon, so nobody was trying to “join up the dots” and tally the full famine impact that year. Especially if much of the effects were felt in South Asia, East Asia or Africa, it is possible that a major-yet-distributed famine could have gone unnoticed
Relatively few people in the world of 1815 relied on food imports, as mentioned above[1]. Breakdown in international food trade is the main famine mechanism in modern-day agricultural catastrophes, but it barely applies to the agrarian economies of 1815.
Local famine effects may not have been worse in Indonesia. Undoubtedly, the effects of ash blanketing crops would have been worse near the eruption. But stratospheric soot quickly circulates around the world. Furthermore, Indonesia has a warm climate and would not have been at risk from completely failed harvests through unseasonal frost. Multiple annual harvests are common in the tropics; in higher latitudes, a failed crop leaves farmers without food for almost a year.
Cooling damage is highly superlinear. The Pinatubo eruption of 1991 caused 0.5 degrees of cooling and is not associated with important declines in agricultural productivity. Thus we might expect the expected burden of an 1815-level cooling event (0.8 to 1.3 degrees cooling) to be far lower than a 1-2.65 degree cooling event[2].
Poverty is a strong predictor of famine mortality. If your persistence estimate relied only on poverty & malnutrition burden indicators, the full-term benefits of policy advocacy would be significantly lower.
To push back:
A world with no extreme poverty and no nutritional diseases would still be vulnerable to global agricultural catastrophes. The advantage of my “grain production per capita” metric is that it has no such edge-case problems: more grain is always good for food security.
The famine/poverty relationship is strong in normal times. There are a number of reasons that it may become less strong in an agricultural catastrophe
Many of the global poor are subsistence farmers in warm countries. This demographic will be close to food supplies and away from the risk of frosts etc.. They may be better-positioned to survive than their middle-class compatriots in cities.
People at the bottom of the pile are the most at-risk of post-catastrophe famine, regardless of whether they are in absolute poverty. A wealthier world would simply have higher food prices, leaving the poorest without enough to eat.
As described above, the main cause of famine in my model is the breakdown of food trade. It is not clear that progress against poverty and nutritional diseases is an indication that populations are less dependent on food trade. If anything, development is probably associated with increased reliance on trade.
Thanks again for the detailed feedback!!
Admittedly, this would have made some people more vulnerable as it was difficult to relieve famine-stricken areas.
A counterargument could be that Western Europe appears to have had particularly bad summer cooling in 1816 - as well as better record-keeping than much of the world—and their famines were not so bad. On the other hand, spring cooling may be more important, as late frosts can ruin harvests of wheat, potatoes etc.
Thanks for the clarifying comment, Stan! I strongly upvoted it.
As a preliminary note, I think it makes a lot of sense to give feedback on analyses like yours privately, but I wonder whether it is worth for me to invest significant time in writing comments like mine above. It seems that they are often downvoted, and that I can sometimes tell before hand when this is going to be case. So, to the extent karma is a good proxy for what people value, I wonder whether I am just spending signicant time on doing something which has little value. In this particular case, I am still guessing it was worth it because, even if it had negligible value to the public, it was still relevant for my own cause prioritisation (and making it public had little cost).
For what is worth, I was already aware of the arguments you mentioned, and directionally agree with all the points you make. I just think their effect is not as strong as you do, so I maintain my adjustments are warranted.
In any case:
Could you comment on the 2 points above?
Agreed, so I adjusted for underreporting in my calculations. I considered the actual mortality to be 13.9 (= 1⁄0.0721) times as high as the reported one.
Agreed, so I adjusted less strongly your mortality rates for more severe coolings:
I am sad to see your comment getting downvotes as I do think it contributes a lot of value to the discussion.
I can guess why you might be getting them. You often respond to cause-prio posts with “what about corporate campaigns for chicken welfare?”, and many people now probably switch off and downvote when they see this. Maybe just keep the chicken comparison to one line and link to your original post/comment?
Also, you comment is 3200 words long—about 3x longer than the actual post. I think a 200-word summary-of-the-comment with bullet points would be really useful for readers who have only read this post and are unable to pick up the finer points of your modeling critique.
On animal welfare
I think that if you adopt RP’s moral weight estimates and reject speciesism, it is almost inevitable that the most cost-effective interventions to improve wellbeing will be animal welfare interventions.
My understanding is that CEARCH is not against evaluating animal welfare interventions in principle, but in practice we are not doing so while comparisons between human and animal welfare remain so shaky. Our research direction is also partly driven by the value of information, ie. how much resources we can plausibly redirect and the impact this will have. Maybe this is too deterministic of me, but I feel that banging the drum about corporate chicken campaigns will only open so many wallets.
Thanks for the feedback on the votes and animal welfare comparison!