Thank you for this post on a very important topic! And thank you for the kind words on my Mortality Cost of Carbon paper.
I think that, at least from the perspective of using my paper, the analysis is actually much simpler than what you do above. Instead of using the 83 million cumulative 2020-2100 excess deaths, use the mortality cost of carbon itself: i.e. the number of lives saved per ton of carbon dioxide reduced, which is provided by the paper. So instead of the equation you show above, the equation now becomes:
Marginal Cost Per Life Saved = (Marginal Cost Per Ton CO_2 reduced)/(The Mortality Cost of Carbon)
Future pessimistic - $5.50 per tonne, so $23,400 to avert 4,255 tons*
Future realistic - $0.29 per tonne, so $1,234 to avert 4,255 tons
Future optimistic - $0.03 per tonnes, so $127 to avert 4,255 tons
The issue with your original equation above is that you are implicitly assuming linearity, i.e. assuming that the marginal cost of saving a life from marginally reducing emissions is equivalent to the average cost of saving a life if we were to reduce emissions all the way to zero. However, one of the findings of the Mortality Cost of Carbon paper is that the system is actually nonlinear and highly convex, so the number of lives saved from marginally reducing emissions is actually much greater than the average number of lives saved that you would get per ton if you were to reduce all the way to zero (see figure 4). This is all a fancy way to say that there are diminishing marginal returns in terms of saving lives from reducing carbon dioxide on a planetary scale. So to determine the marginal impact of reducing emissions, use the marginal estimates provided by the paper (the mortality cost of carbon).
And of course, as you mention above, the mortality cost of carbon is just the projected number of excess deaths from 2020-2100 caused by marginal emissions due to temperature-related mortality - i.e. the net effect of more hot days (bad for mortality) and fewer cold days (good for mortality). It leaves out potentially important climate-mortality pathways such as the effect of climate change on infectious disease, civil and interstate war, food supply, flooding, as well as the co-benefit from less air pollution. Despite these limitations, Louis was still finding that these projections were cost-competitive with Givewell's top recommendations.
*Note that Louis was using the 2020 Working Paper version of the Mortality Cost of Carbon, which included all mortality sources from the 2014 WHO paper (one of three papers used to construct the mortality damage function, which Andrew also mentions in the post), whereas the 2021 published version of the paper in Nature Communications used just the temperature-related mortality estimates from the 2014 WHO report. This ends up leading to a slight difference in the mortality cost of carbon estimate, from 1/4255 in the working paper version to 1/4434 in the published paper version. Recalculating Louis's analysis with the published paper version numbers yields:
Future pessimistic - $5.50 per tonne, so $24,387 to avert 4,434 tons*
Future realistic - $0.29 per tonne, so $1,286 to avert 4,434 tons
Future optimistic - $0.03 per tonnes, so $133 to avert 4,434 tons
Thank you for commenting! I felt like I was relying on your paper without fully understanding it. I'm afraid that much of my post is just an attempt to reinterpret your work.
It's encouraging that Founder's Pledge thinks they can get such a low price on carbon! Interventions at that price really might be effective.
One major question I had about your paper is; what's the breakdown of harms between direct deaths, economic harms, and other losses (like non-fatal hunger)? When the WHO estimates 250,000 deaths each year from 2030-2050, should I interpret a multiplier on that to account for other harms like productivity lost from sickness?
I knew I was making a bad linearity assumption, but I think I might have underestimated how much error it was introducing. If I use my naive model to match your work, I get 50,000 tons/death, which is an order of magnitude off from your estimate. Is that because of an improper linearity assumption?
No worries! I'm glad you found the paper useful and interesting!
The mortality cost of carbon is just the number of excess deaths from temperature-related mortality in units of excess deaths from emitting one metric ton of CO_2. So it's just excess deaths and nothing else. The social cost of carbon is the full monetized value of all climate impacts from emitting one ton of CO_2, which includes the monetized value of those excess deaths in addition to other sources of climate damages. You can see that before the model accounted for temperature-related mortality, the social cost of carbon was $37, but after accounting for temperature-related mortality, it is $258. However, note my caveat from the conclusion: "It is important to note that recent literature has identified other shortcomings in the DICE model including other issues with the climate-economy damage function and the climate module. Besides adding the effect of climate change on mortality and subsequent feedbacks, DICE-EMR takes the rest of the DICE model as given without updating other factors."
It's hard for me to determine how much the different simplifying assumptions from your back-of-the-envelope formula are affecting your estimate. The linearity assumption is certainly causing a big difference because the system is highly convex. Also, the DICE-EMR model has the DICE climate model built into it that can show the climatic effect of changes in emissions. I'm not sure how much error you're introducing with the back-of-the envelope climate assumptions, but that could also be an issue.
All this to say, if estimating the marginal impact (either the mortality cost of carbon or the full social cost of carbon) were as simple as a back-of-the envelope calculation, then there wouldn't be a need to give William Nordhaus the Nobel Prize for his work on the original DICE model (the first one for environmental economics), nor for me to do this work. I think Louis Dixon's original post is basically all you need to do for this exercise (at least for leveraging my paper's results). Or as @jh suggested above, a $1/ton estimate just gets you to $4.4K per life saved using my paper's results.
Also, see this one quote from the end of the paper: "Separate from policy, the MCC and SCC can be useful in informing the decision-making of individuals, households, companies, charities, and other organizations in determining the social impact of the emissions generated by their activities. The emissions contributions of these groups are usually marginal relative to the aggregate emissions of the world economy from the industrial revolution through the twenty-first century. Therefore, the social impact of changes in their activities that either reduce or increase emissions should be quantified using estimates of marginal impacts: i.e. the SCC and the MCC."
Thank you for your responses! I added edits to the essay to reflect this.
Overall, as I noted in the edits, this exercise has made me shift from being skeptical about all climate change interventions to considering shifting some donations from global poverty to climate change interventions. Not entirely convinced, but it seems a lot more plausibly effective than I first suspected.
Some things I don't understand though:
It makes sense that with a convex harms curve, marginal harms will be worse than this back of the envelope linear calculation suggests. But it's surprising that they're 10 times higher. I guess it's just very nonlinear, as you say, but that's surprising to me.
The $1/ton estimate comes from CATF, which is a lobbying organization. Their cost effectiveness calculations account for money they spend lobbying, but not deadweight loss caused by taxes and regulations. How reasonable is it accept that sort of accounting?
1. To your question on accounting for deadweight losses etc., it is true that this is not included, rather this is an estimate of marginal changes from donations. But the factors not included in the calculation are not only deadweight losses (and other costs), but also lots of benefits, e.g. economic benefits from technological leadership. This is parallel to GiveWell analyses which only focus on mortality/direct income gains and ignore a lot of other follow-on benefits and costs.
2. The air pollution benefits of clean energy advocacy are plausibly in the same ballpark as climate benefits (depends on how severe climate change turns out) and benefits from overcoming energy poverty are also very significant (though hard to causally pin-down given the relationship between energy demand growth and human welfare is bidirectional, I explore this a bit more here).
3. One thing that is very different between GiveWell recommendations on global health and FP recommendations on climate is the attitude towards uncertainty -- GiveWell recs have a high uncertainty avoidance whereas CATF and other estimates are meant to be risk-neutral estimates leveraging a fairly indirect theory of change (policy advocacy > policy change > technological change > changed emissions trajectory). So, in that sense the absence of risk-neutral global health recommendations biases the argument in favor of climate.
Thank you for writing this post. I want to point out that your conclusions are highly dependent on your ethical and empirical assumptions. Here are some thoughts about what could change your conclusion:
If you donate to the top charities that are recommended by Founders Pledge, you can probably do much better than 30$/ton. I have not been able to find the precise numbers quickly, but if I remember correctly, 1$/ton is possible under reasonable assumptions. This would change your average estimate to $25,000 per life saved.
Let us assume that the maximal number of happy human beings that could live on earth is reduced by 1 billion by rising sea levels, loss of agricultural land area etc. Let us further assume that this consequences of global warming persist for 100,000 years and that there is a probability of 1% that no game-changing technology such as advanced geo-engineering will be developed. This would mean that 10^12 QALYs are lost and the effectiveness of a dollar would rise. Of course, this argument relies on your rate of temporal accounting.
Climate change could also increase other existential risks. For example, there could be a war about ressources that is fought by nuclear weapons, synthetic pathogens or malevolent AIs.
The message I want to send is not that your analysis is wrong, but that evaluating longtermist interventions is a huge mess since different reasonable assumptions lead to wildly diverging answers.
I'm glad they're looking for charities in the sub $10/ton range! I suspect there is limited room for funding at that value, but it's still marginally good. Finding cheaper climate interventions is really the only part of this equation we can control.
I disagree with your 10^12 QALYs analysis. First, I need a citation on the assumption that livable space will be reduced by 1 billion. Second, the earth isn't at maximum capacity, and I'm not sure population trends are expected to peak above capacity. Third, you shouldn't project out 100,000 years without temporal discounting because our ability to predict the far future is bad and we should use temporal discounting to avoid overconfidence there. For example, it's hard to predict what technology will arise in the future, and assuming a 1% chance that we'll never develop geo-engineering over such a long timespan is a bad assumption.
I agree about existential risks. If climate change causes geopolitical stress that increases the chance of nuclear war by even a small amount, that's obviously bad. I included an x-risk model where we assume climate change kills all humans, but I understand that x-risk would be bad above and beyond the tragic loss of all currently living individuals, so cashing that risk out into dollars per life is maybe incorrect.
About longtermism in general, I basically think EAs are super overconfident about long term predictions, and don't apply exponential discounting nearly enough. Even this analysis going out 100 years is probably overconfident because so much is going to change over that time.
'Climate change could also increase other existential risks. For example, there could be a war about ressources that is fought by nuclear weapons, synthetic pathogens or malevolent AIs.'
To add to this - solar geoengineering could be a major risk (and risk factor for inter-state conflict) that becomes increasingly likely under severe AGW scenarios (people accept more drastic measures in desparate circumstances).
As you noted with McKinsey's GHG abatement curve, there are many interventions that have a negative abatement cost (though some many nitpick and say we need to account for the opportunity cost for the time it takes to change a lightbulb). Marginal mitigation at present is something we can influence, with the best interventions being <$1/ton.
$30/ton seems like the right order of magnitude if we integrate over all abatement costs to get to net zero emissions. Emission reductions are cheap now but will become much more expensive later; that suggests we may want to use the average abatement cost in our calculations assuming we are committed to solving the problem in its entirety. That leaves a high cost per life saved.
I think there is much more we can do on the abatement side. I remember the discussion when the WHO report came out and also read the recent DICE report. These reports assume deaths with no adaptation. Adaptation is something we can influence. Most of the deaths in the WHO report come from issues of poverty: undernutrition, malaria, dengue, diarrhoeal disease. Heat is also listed, but since the mortality rate from cold temperatures is much greater than from hot temperatures, 1-3C of warming may reduce overall temperature-related deaths (https://robjhyndman.com/hyndsight/seasonal-mortality-rates/). Eradicating malaria in the next several decades removes ~1/4 of the mortality risk in the WHO report. Eradicating extreme poverty gets rid of almost all the mortality risk through better nutrition, medical services, and cooling equipment. These are things we can address and are cheaper than mitigation on a per-life-saved basis. This strongly suggests that if your primary concern is human impacts of climate change, it's best to spend your money now on global health and anti-poverty development.
Cost Effectiveness of Climate Change Interventions
AI Use Note: Main body text entirely human written. Claude (Opus 4.8) helped develop models of animal life histories in the appendix.
Cross-posted from Good Structures.
Executive Summary
* Animal advocates sometimes make claims like “there are X of this animal...
“How long have you been v*g*n?”
This is one of the most common icebreakers at animal protection events. It’s a baseline assumption, and it mostly holds true: if you’re out advocating for animals not to be tortured or abused, realistically these days you are v**n, or close. And it makes for good conversation. It seems fairly safe to assume when you meet strangers.
But this assumption is hurting the movement in a way which we don’t always notice: someone new comes into the sp...
Summary
Back in November 2023 I posted here to launch Spiro and raise our first $198k. Two and a half years later this is an update and a fundraiser for the next step.
The short version: we've now reached over-5,900 people with TB preventive medicine, including over 3,000 children under five years old. Our early results have held up well an...
(Epistemic status: Low confidence in end result. I’m not an expert, and I am likely to be overlooking something important. The methodology is informative and there are good links.)
How cost effective is it to help people by preventing global warming? How cost-effective is it, on the margin, to remove CO2 from the atmosphere or prevent carbon from entering the atmosphere? Is it possible to get an answer denominated in cost per life saved?
Edit: Answer: It is possible, but this essay is too oversimplified to get to that answer. Based on discussion in the comments, I think this model is making an inappropriate linearity assumption that hides a steeper gradient on the margin. You might find this exercise informative, but I believe this question's framing distorts the answer by about an order of magnitude.
Let’s look at this equation:
(Cost Per Life Saved) = (Cost Per Ton of CO2) * (Tons of CO2 Per 1°C Warming) * (Expected Degrees Warming) / (Lives Lost Due to Climate Change)
There are a lot of unknowns and many of these terms will change due to technological or social changes, but let’s try to estimate them.
Edit: TL;DR: If we guess at these numbers, it costs about 50,000 $750,000 per life saved if you want to save lives by eliminating carbon from the atmosphere. However, I believe this is an overestimate by a factor of 10. See here for a different analysis which yields a significantly lower cost.
Part 1: Cost Per Ton of CO2
How much money do you have to spend to make there be one less ton of CO2 in the atmosphere?
Edit: Most of these numbers are misleadingly large, and cheaper interventions appear to be available. See note at end.
This turns out to be a relatively easy question to answer because there are people actively working on producing technology to improve this number. Here’s an article about direct carbon capture, where they say the current cost is about $600/ton, but they’re hoping to get it down to $100/ton. The Department of Energy has the goal of reducing this to $30/ton.
That’s the marginal cost, and if they scaled it to address all the carbon put into the atmosphere annually, there’s reason to think that the cost might be significantly lower (because of economies of scale) or higher (because they’d run out of manufacturing capacity and land).
But that’s the cost for sucking carbon out of the atmosphere once it’s already there. What about things we can do to prevent carbon from entering the atmosphere in the first place?
Climate change economists have produced a graph called an abatement cost curve, which shows a comparison of the cost of different interventions denoted in dollars per ton of carbon.
Many of these interventions have negative cost, meaning that they reduce carbon and also cost less in the long run. Switching incandescent bulbs to LEDs is a good example of that; they last a lot longer and therefore they’re cheaper. Looking at the chart, I see there are a lot of interventions in the €15-50 range.
Table 2 in this paper contains more such cost estimates, which are generally a bit more pessimistic; it notes many interventions are in the $30-300 range. For example, “Livestock Management Policies” are estimated to cost $71 per ton of carbon, while the Cash for Clunkers program in the US had a cost of about $270-420 per ton of carbon, which is worse.
Here’s an article about Biden’s climate policy, which includes a particularly cost effective climate change intervention in the form of a tax credit, which is estimated to cost $33 to $50 per ton of carbon.
Cost estimates vary!
In general, the cheapest interventions will not scale to fix the entire climate problem. Some industries like steel transport are very expensive to decarbonize.
So at the end of this section, we’ve got to put a price on a ton of carbon. I’m going to go with $30, which is representative of some of the most effective interventions we have available today. Most interventions will be more expensive than this, but some will be cheaper.
EDIT: Commenters point out that some lower cost interventions are available in the sub ten dollar range, and potentially in the sub one dollar range. I've updated this essay to include that. I'm going to use $2, although some cheaper interventions may be or may become available. In particular, see the Founders Pledge on the Clean Air Task Force, page 95.
(Side note: Tons vs Tonnes. Some of you might remember that the metric tonne at 1,000kg is different from the English ton at 2,000lb, or 907kg, and you might have noticed that I was glossing over that. However, they’re pretty close and the difference is much less than the margin of error on the conclusion, so I’m treating them as the same. Sorry!)
TL;DR: $2/ton (Edit: was $30)
Part 2: Tons of CO2 Per 1°C Warming
How much CO2 do we need to put into the atmosphere to warm the planet by one degree celsius?
It turns out this answer is a little bit complicated because greenhouse gasses don’t warm the planet directly, instead they help the sun warm the planet. But increased carbon does increase the steady state temperature the earth will reach. Someone on Quora ran a simulation where you can see that. Their answer was 5 trillion tons.
Some environmentalists have computed a carbon “budget” to hit various warming targets. This source suggests about ~500 gigatonnes CO2 to get 1.5°C, or ~1,500 gigatonnes CO2 to get 2°C of global warming. That works out to about 2 trillion tons per degree, or 2Tt/1°C.
Other posters on that Quora question roughly agree with that estimate, from various lines of argument, so that’s the number we’ll use.
TL;DR: 2Tt/1°C
Part 3: Expected Degrees Warming
How much do we expect the earth to warm up by? This is a complicated question that I’m not qualified to address, so I’m going to present this infographic from NASA and then interpret it badly.
Looks like about 2 or 3°C degrees of warming.
TL;DR: Let's say 2.5°C.
Part 4: Lives Lost Due to Climate Change
If we get 2 or 3°C of global warming, that will be bad, for all the reasons that climate change is bad. But how bad, specifically? Can we put a number on how bad it will be? In particular, how many people do we expect to die because of global warming?
Edit: This section obscures the difference between overall mortality due to climate change and marginal mortality due to climate change. Models suggest that harms due to climate change are nonlinear and that marginal harms will increase at higher temperatures.
This turns out to be a hard question to answer, but there are some good estimates. There are also some bad estimates, which I’m presenting also. Here are 4 methodologies, ordered from most optimistic (but still bad) to most pessimistic. Coincidentally, the most optimistic methods are the highest quality.
Detailed Multi-Factor Analysis from the WHO
The World Health Organization put out an estimate that between 2030 and 2050, 250,000 additional people will die each year because of climate change. They created this infographic:
The full report is worth reading if you’re interested in this question. They did their research and accounted for many different potential vectors of climate change. I was originally going to say that nobody has a good answer to this question, but this research is detailed and convincing.
If we project that number over a full century, it adds up to 25,000,000 people losing their lives due to climate change.
Economic Modeling
This excellent article from R. Bressler attempts to answer the question “how many tons of carbon in the atmosphere kills one person”, and their answer is “4,434 metric tons of carbon dioxide in 2020—equivalent to the lifetime emissions of 3.5 average Americans—causes one excess death globally in expectation between 2020-2100”.
If I understand the paper correctly, they use a modification of the DICE model to get an estimate that 83 million people will die between 2020 and 2100. I think they incorporated the effect of economic losses and harms due to non-death losses into this figure, which you can see in equations 7 and 8 in their paper. They're actually summing harms out until the year 2510, but they're using temporal discounting so they don't weight it very heavily. I'm not sure why they chose such a long time horizon.
One thing I like about this paper is that they try to balance the economic benefits of industry and economic growth against the harms of climate change and conclude that 2° of warming is “optimal”.
I’m going to extend their estimate to a full century at 100 million dead.
All the Wars and Famines
This model is pure guesswork, but it’s based on how I hear people talk about their concerns about climate change.
Let’s assume that climate change will cause global unrest which leads to terrible wars, and also that it will disrupt food production and cause mass starvation.
How many people do we expect to die in these terrible wars and famines? In the 20th century, about 100,000,000 people died from famine, and a similar number died in all the wars. Let’s assume that the next hundred years will be similar, and that 100% of it can be blamed directly on climate change. This is probably an overestimate, because there might be other causes of war[citation needed].
That works out to 200 million dead over the century, or 2 million people per year on average.
Worst Case Scenarios
This article thinks a billion people will die because of climate change. It was published in the journal of Environmental Psychology and spends a lot of the article talking about how to use its results in political messaging. Here’s a quote about their methodology: “A more rigorous multivariate analysis [...] will not be attempted here. Instead, I will present a big-picture, top-down estimate.” Their numbers for death estimates are summarized in their section titled “Worst-Case Estimate”, and include a variety of potential harms.
I did not find their methodology convincing, but I am including it because it was widely cited in the press. 1 billion dead over a century.
Vague, Spread Out Harms
You may have the intuition that climate change may not kill that many people, but it will reduce the quality of life of many people across the world in a way that's worse than the deaths it causes. With the exception of Bressler, the analyses above don't go into that.
I find this hard to estimate, and I wasn't able to find a good study on this. Please post in the comments if you're able to find a rigorous analysis of this!
For the sake of estimation, let's assume that climate change affects everyone's lives for the worse. We need a number in the end, so let's say that everyone becomes 5% worse off. Some people won't be affected that much, some people will be affected worse, but 5% on average. For reference, this study (page 44-45) indicates that having a tension headache or shingles causes a 5% decrease in well-being. Following the logic of QALYs, we can say that everyone being 5% worse off is sort of equivalent to 5% of people dying, or about 500 million deaths. This might be on top of the deaths estimated by some other method.
Temporal Discounting
(Side note about time: I think temporal discounting is appropriate here, not because we care less about people in the future, but because we have less ability to predict how we affect them. Using a too-conservative but convenient 1% annual discount rate and the formula for the sum of an infinite geometric series, 1/(1-0.99), we find that the infinite sum is equal to 100 years, helpfully letting us use one century as a stand-in for all the rest of time.)
Summary
Putting these together, we have a range of death estimates from 25-1,000 million deaths over the course of a century. I think numbers toward the lower end of that range have the best support from good research, but that numbers toward the higher of that range match up better with environmentalist intuitions.
TL;DR: I’m going to err on the low side because it's supported by better research and use an estimate of 200 million deaths.
Math: Putting it All Together
Math time! Remember our formula:
(Cost Per Life Saved) = (Cost Per Ton of CO2) * (Tons of CO2 Per 1°C Warming) * (Expected Degrees Warming) / (Lives Lost Due to Climate Change)
(Note: AGW stands for anthropogenic global warming, and it’s a dummy unit indicating that 1 global warming will have occurred. It reflects my failure to denominate harms in deaths/°C.)
The answer we get is $50,000 per life saved through spending to reduce global warming by removing carbon from the atmosphere.
Edit: See conclusion.
Obviously, there’s a lot of uncertainty here. The environmental science around how much warming will occur and how much CO2 will cause that warming is pretty good, and it has less than a degree of magnitude in error bars. The estimate of how many lives will be lost due to global warming is worse, and I could easily see it be off by an order of magnitude in either direction, especially when you account for non-death harms.
I’ve tried to include slightly conservative estimates throughout, so I think this number represents an optimistic, low end estimate for the cost.
Note 1: Ranges and Uncertainty
Edit: Outdated.
We saw uncertainty in each step of the model. Here's the ranges of plausible values we found:
Cost Per Ton of CO2: $10-$50. This is only considering a range for the very best interventions. Most climate interventions will be more expensive than this.
Tons of CO2 Per 1°C Warming: 1.5-2.5 Tt.
Degrees of Warming: 1.5-5°C. Seems like a lot of uncertainty here.
Deaths Caused by Global Warming: 25-1,000 million.
From these we can get a low-end estimate and a high-end estimate using the formula above.
This is the best case scenario (for cost effectiveness, not for the future), where we assume that climate change will be really bad really quickly, and it's really cheap to fix:
This is the worst case scenario (again, for cost effectiveness), where we assume that climate change won't be that bad and also it's expensive to deal with:
Edit: Climate change is still expensive, but this analysis overstates it. Cheaper interventions are available than I initially thought. However, the cheapest interventions wouldn't scale to fix the whole problem.
One takeaway I’ve gotten from this exercise is that climate change is a really expensive problem. Reversing climate change would cost about 120 trillion dollars, or about 6 times the GDP of the USA, and slightly more than global GDP. If the US federal government spent its entire budget every year for 10 years on addressing climate change, it might be expected to address 1°C of warming, out of the 2 or 3°C that we’re expecting to experience. I can think of other ways in which we might want to spend that.
Note 3: Worst Case Scenario
Edit: Outdated
What if climate change is an existential risk that will lead to the death of all humans?
If we assume that climate change will immediately cause the death of all 8 billion humans on earth unless we pony up the money to reverse 2 degrees of warming with our best present technology, we get:
This provides an upper bound for how cost-effective climate change interventions can be.
Note 4: Comparisons to Other Cause Areas
GiveWell estimates that its top charities in global health need about $5,000 to save one life or life equivalent, which includes quality of life impacts from illness besides death, such as injury or loss of ability to work, primarily through preventative medicine in the developing world. The cost I’ve computed above is significantly worse than that.
Edit: See commentary in conclusion, climate change may be cheaper to influence than this line of analysis suggests. Some climate change charities may be considerably cheaper than GiveWell's top charities.
Climate change interventions might become competitive with global health charities if two things become true:
We update our models toward increased cost-effectiveness for climate change. This might mean that climate change starts to look more harmful, or we develop much better technology to deal with it.
GiveWell's top charities run out of room for more funding, and the marginal available direct intervention gets much worse.
If carbon removal technology improves significantly or the impacts of climate change start to look worse, or if global health loses room for additional funding, then climate change may enter the ballpark of being comparable to GiveWell’s top charities.
If you’re optimistic about technology and pessimistic about the harms from climate change, it starts to look comparable to GiveDirectly, which is about ten times less cost effective (per life saved) than top malaria charities.
Note 5: Non Human Harms
Koalas are at risk from forest fires, and their well-being is hard to calculate
Climate change affects animal welfare and nature in a way that malaria or global poverty don’t. A full accounting of the cost of climate change needs to account for animal suffering, but that’s outside the scope of this post. In addition, climate change is likely to reduce biodiversity, which might be seen as intrinsically valuable.
Conclusion
I did not find that climate change interventions are cost effective compared to GiveWell's top charities, at about $50,000 per life saved. My main source of uncertainty was in estimating the harms expected to be caused by climate change, and more research there is needed. Because climate change is so expensive to mitigate, a lower bound of about $15,000 per life can be established in the case where climate change is treated as an X-risk.
Edit: See DannyBressler's comment below (same Bressler cited in "Economic Modeling"). His analysis finds a considerably lower number, and his analysis is more sophisticated, so I recommend believing it over this essay's.
Surprisingly to me, this cost is smaller than I expected. I went into this essay expecting to find that climate change would be extremely ineffective to influence, and while I still believe it’s much worse than GiveWell’s top charities, the gap is smaller than I expected.
Edit: This exercise has shifted me from believing "it is not cost effective to do anything about climate" to believing "maybe I should donate money to climate change interventions."
I’d also like to point out that there are likely to be climate interventions that are much more effective than the $2/ton estimate that I use. In particular, some interventions like building solar panels are likely to cost negative dollars, and it’s sort of a coordination problem to make that happen. Also, it’s likely that technical research is likely to have the leverage to drive down those costs and be much more cost effective. Estimating the cost effectiveness of research is outside the scope of this essay.
I’d like to see more analysis of this question. I think that non-expert EAs could improve on my estimate by doing a more detailed reading of the literature.
Thank you for this post on a very important topic! And thank you for the kind words on my Mortality Cost of Carbon paper.
I think that, at least from the perspective of using my paper, the analysis is actually much simpler than what you do above. Instead of using the 83 million cumulative 2020-2100 excess deaths, use the mortality cost of carbon itself: i.e. the number of lives saved per ton of carbon dioxide reduced, which is provided by the paper. So instead of the equation you show above, the equation now becomes:
Marginal Cost Per Life Saved = (Marginal Cost Per Ton CO_2 reduced)/(The Mortality Cost of Carbon)
@Louis_Dixon performed this analysis before in a really nice post Does using the mortality cost of carbon make reducing emissions comparable with health interventions? - EA Forum (effectivealtruism.org) and found that using carbon dioxide reduction estimates from a Founder's Pledge report:
The issue with your original equation above is that you are implicitly assuming linearity, i.e. assuming that the marginal cost of saving a life from marginally reducing emissions is equivalent to the average cost of saving a life if we were to reduce emissions all the way to zero. However, one of the findings of the Mortality Cost of Carbon paper is that the system is actually nonlinear and highly convex, so the number of lives saved from marginally reducing emissions is actually much greater than the average number of lives saved that you would get per ton if you were to reduce all the way to zero (see figure 4). This is all a fancy way to say that there are diminishing marginal returns in terms of saving lives from reducing carbon dioxide on a planetary scale. So to determine the marginal impact of reducing emissions, use the marginal estimates provided by the paper (the mortality cost of carbon).
And of course, as you mention above, the mortality cost of carbon is just the projected number of excess deaths from 2020-2100 caused by marginal emissions due to temperature-related mortality - i.e. the net effect of more hot days (bad for mortality) and fewer cold days (good for mortality). It leaves out potentially important climate-mortality pathways such as the effect of climate change on infectious disease, civil and interstate war, food supply, flooding, as well as the co-benefit from less air pollution. Despite these limitations, Louis was still finding that these projections were cost-competitive with Givewell's top recommendations.
*Note that Louis was using the 2020 Working Paper version of the Mortality Cost of Carbon, which included all mortality sources from the 2014 WHO paper (one of three papers used to construct the mortality damage function, which Andrew also mentions in the post), whereas the 2021 published version of the paper in Nature Communications used just the temperature-related mortality estimates from the 2014 WHO report. This ends up leading to a slight difference in the mortality cost of carbon estimate, from 1/4255 in the working paper version to 1/4434 in the published paper version. Recalculating Louis's analysis with the published paper version numbers yields:
Thank you for commenting! I felt like I was relying on your paper without fully understanding it. I'm afraid that much of my post is just an attempt to reinterpret your work.
It's encouraging that Founder's Pledge thinks they can get such a low price on carbon! Interventions at that price really might be effective.
One major question I had about your paper is; what's the breakdown of harms between direct deaths, economic harms, and other losses (like non-fatal hunger)? When the WHO estimates 250,000 deaths each year from 2030-2050, should I interpret a multiplier on that to account for other harms like productivity lost from sickness?
I knew I was making a bad linearity assumption, but I think I might have underestimated how much error it was introducing. If I use my naive model to match your work, I get 50,000 tons/death, which is an order of magnitude off from your estimate. Is that because of an improper linearity assumption?
(50,000 tons/life) = (2 Tt/°C) * (2.5°C/AGW) / (100 million lives / AGW)
When I have time later, I'll edit the post to include some of your feedback.
No worries! I'm glad you found the paper useful and interesting!
The mortality cost of carbon is just the number of excess deaths from temperature-related mortality in units of excess deaths from emitting one metric ton of CO_2. So it's just excess deaths and nothing else. The social cost of carbon is the full monetized value of all climate impacts from emitting one ton of CO_2, which includes the monetized value of those excess deaths in addition to other sources of climate damages. You can see that before the model accounted for temperature-related mortality, the social cost of carbon was $37, but after accounting for temperature-related mortality, it is $258. However, note my caveat from the conclusion: "It is important to note that recent literature has identified other shortcomings in the DICE model including other issues with the climate-economy damage function and the climate module. Besides adding the effect of climate change on mortality and subsequent feedbacks, DICE-EMR takes the rest of the DICE model as given without updating other factors."
It's hard for me to determine how much the different simplifying assumptions from your back-of-the-envelope formula are affecting your estimate. The linearity assumption is certainly causing a big difference because the system is highly convex. Also, the DICE-EMR model has the DICE climate model built into it that can show the climatic effect of changes in emissions. I'm not sure how much error you're introducing with the back-of-the envelope climate assumptions, but that could also be an issue.
All this to say, if estimating the marginal impact (either the mortality cost of carbon or the full social cost of carbon) were as simple as a back-of-the envelope calculation, then there wouldn't be a need to give William Nordhaus the Nobel Prize for his work on the original DICE model (the first one for environmental economics), nor for me to do this work. I think Louis Dixon's original post is basically all you need to do for this exercise (at least for leveraging my paper's results). Or as @jh suggested above, a $1/ton estimate just gets you to $4.4K per life saved using my paper's results.
Also, see this one quote from the end of the paper: "Separate from policy, the MCC and SCC can be useful in informing the decision-making of individuals, households, companies, charities, and other organizations in determining the social impact of the emissions generated by their activities. The emissions contributions of these groups are usually marginal relative to the aggregate emissions of the world economy from the industrial revolution through the twenty-first century. Therefore, the social impact of changes in their activities that either reduce or increase emissions should be quantified using estimates of marginal impacts: i.e. the SCC and the MCC."
Thank you for your responses! I added edits to the essay to reflect this.
Overall, as I noted in the edits, this exercise has made me shift from being skeptical about all climate change interventions to considering shifting some donations from global poverty to climate change interventions. Not entirely convinced, but it seems a lot more plausibly effective than I first suspected.
Some things I don't understand though:
It makes sense that with a convex harms curve, marginal harms will be worse than this back of the envelope linear calculation suggests. But it's surprising that they're 10 times higher. I guess it's just very nonlinear, as you say, but that's surprising to me.
The $1/ton estimate comes from CATF, which is a lobbying organization. Their cost effectiveness calculations account for money they spend lobbying, but not deadweight loss caused by taxes and regulations. How reasonable is it accept that sort of accounting?
(Working at Founders Pledge)
1. To your question on accounting for deadweight losses etc., it is true that this is not included, rather this is an estimate of marginal changes from donations. But the factors not included in the calculation are not only deadweight losses (and other costs), but also lots of benefits, e.g. economic benefits from technological leadership. This is parallel to GiveWell analyses which only focus on mortality/direct income gains and ignore a lot of other follow-on benefits and costs.
2. The air pollution benefits of clean energy advocacy are plausibly in the same ballpark as climate benefits (depends on how severe climate change turns out) and benefits from overcoming energy poverty are also very significant (though hard to causally pin-down given the relationship between energy demand growth and human welfare is bidirectional, I explore this a bit more here).
3. One thing that is very different between GiveWell recommendations on global health and FP recommendations on climate is the attitude towards uncertainty -- GiveWell recs have a high uncertainty avoidance whereas CATF and other estimates are meant to be risk-neutral estimates leveraging a fairly indirect theory of change (policy advocacy > policy change > technological change > changed emissions trajectory). So, in that sense the absence of risk-neutral global health recommendations biases the argument in favor of climate.