Error
Unrecognized LW server error:
Field "fmCrosspost" of type "CrosspostOutput" must have a selection of subfields. Did you mean "fmCrosspost { ... }"?Unrecognized LW server error:
Field "fmCrosspost" of type "CrosspostOutput" must have a selection of subfields. Did you mean "fmCrosspost { ... }"?
Wait a second.
I’m very confused by this sentence. Suppose, for the sake of argument, that all the impacts of emitting a tonne of CO2 are on people about as rich as present-day Americans, i.e. emitting a tonne of CO2 now causes people of that level of wealth to lose $417 at some point in the future. There is then no income adjustment necessary (I assume everything is being converted to something like present-day USD for present-day Americans, but I’m not actually sure and following the links didn’t shed any light), so the post-income-adjustment number is still $417. Also suppose for the sake of argument that we can prevent this for $100.
This seems clearly worse than cash transfers to me under usual assumptions about log income being a reasonable approximation to wellbeing (as described in your first appendix), since we are effectively getting a 4.17x multiplier rather than a 50-100x multiplier. Yet the equation in the quote claims it is 4.17x more effective than cash transfers*.
What am I missing?
*Mathematically, I think the equation works iff. the cash transfers in question are to people of comparable wealth to whatever baseline is being used to come up with the $417 figure. So if the baseline is modern-day Americans, that equation calculates how much better it is to avert CO2 emissions than to transfer cash to modern-day Americans.
I was about to say this and then saw your comment. My impression from the paper is the $417 is a sum of costs to different countries, and for each of them the cost is a present value to the people in that country, with discounting being applied based on the expected amount of economic growth in that country. So I don’t think it’s calibrated to present-day Americans, but I don’t think it’s calibrated to the world’s poorest either, and I agree the argument doesn’t go through.
There’s another problem with the quoted claim, which becomes clear if you pick a value like X = 1/1000. Paying $417,000 to avert a tonne of carbon is a huge net bad and not just a much smaller net good.
It seems to me another problem is that if the social cost of carbon comes from effects on growth, you have to compare that to the effects on growth of cash transfers. It’s generally easy for small changes in growth rate to outweigh small changes to level in the long run, so if you compare the growth effects of one intervention to the level effects of another intervention, it’s no surprise that the former would seem more effective.
Am I right that what we need to know is whether, when assessing the total global social cost of carbon:
i) they multiplied harms to the world’s poorest people, measured in equivalent dollars of their income, by ~100x;
ii) they divided the harms to the world richest people, measured in equivalent dollars of their income, by ~100x;
iii) or something in between?
HH’s argument only goes through if it’s the second of the three.
I agree with this. I would have assumed they would do (i), and other responses from people who actually read the paper make me think it might effectively be (iii). I don’t think it’s (ii).
I mentioned it in my comment elsewhere, but—from a quick look at the paper and the supplementary material—I don’t think it’s much like any of these. They don’t make any special mention that I could find of trying to translate purely economic measures into welfare. The only mention I could find about income adjustment is “rich/poor specifications” which appears to be about splitting the formula for growth of damages into one of two forms depending on whether the country is rich or poor.
Edit: They do mention “elasticity of marginal utility” in the discounting module section which is also known as “intergenerational inequality aversion”.
My nonconfident best guess at an interpretation is that, according to these estimates, for every tonne of carbon:
Future Indians suffer damages utility-equivalent to the present population of India paying a total of $76
Future Americans suffer damages utility-equivalent to the present population of the USA paying a total of $48
Future Saudis suffer damages utility-equivalent to the present population of Saudi Arabia paying a total of $47
Next are China, Brazil, and the UAE, all with $24, and then a lot of other countries, and the sum of all these numbers is $417. So it’s as if the $417 is paid by this particular mix of the world’s people, making it iii), something in between. These numbers are totals that don’t divide by population, so an individual inhabitant of Saudi Arabia or the UAE pays a greater absolute amount than an individual American, who pays a greater absolute amount (but a smaller percentage of income) than an individual Indian.
Thanks for catching this mistake.
I’ve updated the analysis to reflect this.
I emailed the authors but they didn’t reply.
But I think the social cost of carbon figures should generally be interpreted as current US dollars. They are then discounted for decreasing returns to consumption for future people who live in countries with higher consumption.
So we should divide the $417 figure by the 100x multiplier (or more, see my sensitivity analysis).
I’m still a bit worried about this.
It would have been reasonable for them to use the mean global income as the baseline, rather than dollars to the mean US citizen.
If I understand correctly, that would boost things by about a factor of 3 in favour of climate change (mean global income is about $20k, vs. mean US income of about $60k). Though, I suppose that’s a fairly small uncertainty compared to the others listed here.
The assumption is that a policymaker will use these results to shape how strict climate policy is. Stricter climate policies will reduce present-day consumption in the policymaker’s jurisdiction. The goal is to have a climate policy that is just strict enough to balance the future utility gain from improved climate with current utility loss from reduced consumption.
(Anthoff and Emmerling, 2016, p. 5, emphasis added).
I think it’s conventional in the literature to use the US as the reference region: “In order to make the numerical results comparable with previous studies we take the US as reference region x throughout this paper” (ibid).
My current understanding of how we should interpret a social cost of carbon of $y is:
“Given certain assumptions about future climate effects, emissions trajectories, how utility scales with consumption, and how much we discount future utility, a utilitarian policymaker in country x should be willing to reduce current consumption in his country by up to $y in order to abate 1 ton of carbon emissions”
This also means that, due to assumptions about diminishing marginal utility, the choice of reference region majorly affects the SCC. For example, Anthoff and Emmerling show how reference region affects their results. (Remember that this, like all SCC estimates, is subject to lots of assumptions about the effects of climate change and discounting.)
(Sorry for the huge image)
If the SCC estimate is low enough it can actually be negative for some regions, meaning that any reduction in present-day consumption in those countries to mitigate climate change would reduce utility.
I wanted to jump in here while the discussion is active, but I’ll also flag that John, Johannes and I are working on this and should have a post on comparing climate vs. global health/development interventions in the not-too-distant future.
Great, thank you for this! Look forward to seeing more work also.
And just a quick thought: if we know what the SCC of carbon is for Africa (looks like ~$10), and it’s defined in the way you say, then we could also do the comparison directly with the Africa-SCC figure, rather than converting into US equivalent first e.g.:
1 tonne of CO2 averted → equivalent to $10 of consumption in Africa If it costs $1 to avert a tonne, then $1 → $10 consumption $1 cash transfer → $1 of consumption in Africa (or maybe ~$5 to a GiveDirectly-recipient) $1 to AMF → ~$50 African-consumption-equivalent (thinking of it as 10x GiveDirectly)
So with these figures, carbon offsets are better than cash transfers, but AMF is 5x better than carbon offsets.
I had emailed all the authors of this analysis and asked them, but they didn’t get back to me, so I think it’s ambiguous and not really replicable. But yes I agree it’s a fairly small uncertainty compared to the others.
Thanks for rewriting and republishing this. All very interesting.
On this new revised version, something that stood out to me was the truly extreme range between the optimistic and pessimistic scenarios you describe.
I think the relative cost-effectiveness range you’ve given spans fully ten orders of magnitude, or a range of 10,000,000,000x. Even by our standards that’s a lot. If we’re really this uncertain it seems we can say almost nothing. But I don’t think we are that uncertain.
By choosing a value out in the tail for 4 different input variables all at once you’ve taken us way out into the extremes of the uncertainty bounds. It looks to me like for these scenarios you’ve chosen the 1st and 99th percentiles for SCC, η, cost of abatement, and gain from doing health, all at once.
If that’s right you’re ending up at more like 0.01 * 0.01 * 0.01 * 0.01 --> 0.000001th percentile on the cost-effectiveness output on either end (not really, because you can’t actually combine uncertainty distributions like this, but you get my general point). That seems too extreme a value to be useful to me.
Maybe you could put your distributions for the inputs into Guesstimate, which will do simulations drawing from and multiplying the inputs, and then choose the 5th and 95th percentile values for the outputs? That would go a long way towards addressing this issue.
Hope this helps, let me know if I’ve misunderstood anything — Rob
Thanks Rob for taking the time to comment and my sincere apologies for the delay in replying.
There really is a lot of uncertainty here. Note that all parameters estimates are based on or grounded in empirical and published estimates. Even my adjustment for the social cost of carbon being over- or underestimate by 10x corresponds to values with similar orders of magnitude you can find in the literature—see cell comments of the spreadsheet model. For instance, one recent paper by a renowned climate economist finds that under different model specifications the SCC ranges from $3.38/tCO2e to $21,889/tCO2e. Ditto with what the eta parameter for the income adjustment.
The “realistic estimate” model scenario is what I perceive to use parameters estimates around which there more consensus, but that’s just my opinion and one can reasonably disagree with these choices.
I used the extreme scenarios to highlight the uncertainty and to make statements such as “Even if you believe the true social cost of carbon is higher than most models suggest (i.e. $20k per tonne, the most extreme value in the literature), then that still often is not enough to beat global development interventions”.
Generally, my agenda was probably a bit simpler than people might have supposed. This was not intended to be the last word on whether climate change or development interventions are always better. Rather it’s a starting point and “choose your own adventure” model to help prioritizing between a concrete climate and a concrete development charity.
Note that there are four parameters that drive the results of this analysis (the SCC, the income adjustment eta, the cost to avert CO2, and the effectiveness of global dev/health vs. cash). For the first two, there really is a lot more uncertainty, but for the latter two, it’s more clear. This makes the model actually valuable and with action guiding potential.
For instance, if you’re a small donor and can’t decide between GiveDirectly and the Coalition for Rainforest Nations, then, if you believe that CfRN really has a cost-effectiveness of $0.02 / tCO2e averted, in many scenarios, especially the realistic one around which there is most consensus, it will often beat unconditional cash-transfers, even if you believe that social cost of carbon is quite low.
However, CfRN does lobbying, not a scalable intervention that one could invest a lot of money in. So, in contrast, if you’re a billionaire and are looking to decide between global development and climate change as a cause area for your foundation, then perhaps global development might be a better bet.
Re: Monte Carlos—I think some of the parameter inputs rely on Monte Carlos already. My hunch is that there’s no free lunch here that would reduce the uncertainty much over and above the point estimate / realistic scenario, but this is definitely that I’d like to see other people explore in future research.
I think working this through on guesttimate rather than mulitplying point estimates is really important.
I tried doing it myself with similar figures, and I found the climate change came out ~80x better than global health (even though my point estimate that that global health is better) - which suggests the title of the article could maybe use editing!
When you’re dealing with huge uncertainties like these, the tails of the distribution can drive the EV, so point estimates can be pretty misleading.
Here’s a screenshot of the model: https://www.dropbox.com/s/adtwlz3k2myv8gc/Screenshot 2020-05-25 20.57.11.png?dl=0
I also tried doing the calculations in a different way that I found more intuitive—where I estimate the ‘utils’ of each intervention: https://www.dropbox.com/s/8uczqc1qhi71lte/Screenshot 2020-05-25 20.58.02.png?dl=0
Some other reasons in favour of this approach:
Rob’s point that by multiplying together extreme values, your confidence intervals are unreasonably wide.
Some of the confidence intervals you give for the individual parameters also seem too wide (and seem to not be mathematically possible to fit to a lognormal distribution).
Thanks for the comment!
I think I’ll just leave the title for now, because it is confusing as it is and I’m not sure if it’s worth it to redo/rewrite the analysis. I should probably have just called it “How to compare the relative effectiveness of development vs. climate interventions”. I’ll make a note in the beginning of the post linking to your guesstimate, saying that you found different results.
I can’t quite follow your analysis from the screenshots (perhaps you could link the models and the assumptions for others). For instance, I’m not sure why the input value of money going to Americans vs. GiveDirectly recipients is 23 to 350.
But generally, I agree that Monte Carlo simulations and minding the distributions can be valuable for better error propagation. Also, I was probably being unclear but my analysis was not supposed to be a confidence intervals but rather my the best guess and extreme scenarios.
For instance, in the cell for cost per tonne of CO2 averted in the pessimistic scenario I intentionally picked the extreme value from the Founder Pledge analysis $0.02 and not their mean value (from the cell note: “A donation to CfRN will avert a tonne of CO2e for $0.12, with a plausible range of $0.02 - $0.72.” https://docs.google.com/spreadsheets/d/12lwvxlWLjwuSuXiciFvnBF2bkfcCkrusdqqT37_QWac/edit#gid=1267972809&range=E35 ).
Echoing what Greg Lewis said about hobbyists modelling the C19 pandemic being perhaps not super productive, I’m also not sure how productive further
empirical work such as this is on the EA forum (I don’t even know how many hits the forum gets generally, and this post in particular, how many climate modellers read it, etc.). I think maybe an org with more research capacity would be better suited to do further analysis on this. Or perhaps one could commission researchers with a background in climate modelling to do this (e.g. the author of this paper might be really qualified to do this: https://www.sciencedirect.com/science/article/pii/S014098831930218X ).
Hey Hauke,
That makes sense.
I do think more EA work on this topic would be useful for someone to do, since I don’t think it’s clear from a near-termist perspective that global health is more effective than climate change.
On guesstimate, there was an error and I was unable to save my model. If someone is looking to reproduce this though, I’d suggest they just make their own.
On the value of money to Americans vs. GiveDirectly recipients, my personal estimate was a lower ratio, because I think we should take into account some flow through effects and I think this causes convergence. I don’t think values like 10,000x are plausible for the all-considered tradeoff (even though the ratio could be 10,000x if we’re just considering the welfare of two individuals). More here: http://reflectivedisequilibrium.blogspot.com/2014/01/what-portion-of-boost-to-global-gdp.html
I’m still a bit unclear how useful these are due to Rob’s point.
Below is one important point that I think is extremely difficult to know without being an active researcher in the field. Hauke hints at it in his footnote 6, but I want to expand on it since I think it is important to understand where the social cost of carbon estimates are coming from:
Ricke et al. 2018 (https://www.nature.com/articles/s41558-018-0282-y) are using a climate damage function that predicts much higher damages than the damage function that is used in the main integrated assessment models (IAMs) that predict the social cost of carbon (DICE, FUND, and PAGE). (I’ve included a note with links on this below for those who are interested). This results in an important point:
Most of the difference in the social cost of carbon between Ricke et al. 2018 and the main IAMs is because they are using different damage functions, not because they are accounting for greater marginal utility of consumption to individuals with lower consumption levels.
The more appropriate control would be to compare Ricke et al. 2018 to the gro-DICE model developed by Diaz & Moore 2017 ( https://www.nature.com/articles/nclimate2481 ), which uses a damage function that is more similar to Ricke et al. 2018. gro-DICE projects a social cost of carbon of $220, compared to Ricke et al.’s $415. However, the gro-DICE damage function is still less damaging than Ricke et al’s damage function. Ideally, we would want to do a comparison of the social cost of carbon using the same damage function (so we could isolate just the effect of differences in marginal utility), but unfortunately we can’t readily do this because these papers are all using different damage functions. Given that Ricke et al. are using the most damaging damage function, we do know that the effect would be less.
Note on climate damage functions: Ricke et al. 2018 use a damage function based off of Burke, Hsiang, and Miguel 2015 ( https://www.nature.com/articles/nature15725 ). This is the “most damaging” of the damage functions in the literature. Diaz & More 2017 uses a damage function based on Dell, Jones, and Olken 2012 ( https://www.aeaweb.org/articles?id=10.1257/mac.4.3.66), which is more damaging than the damage functions used in traditional IAMs, but not as damaging as the Burke, Hsiang, and Miguel 2015 damage function used in Ricke. et al. 2018.
These analyses predict much higher climate damages than the analysis that make up the DICE-2016 damage function (the DICE-2016 damage function is derived here: https://cowles.yale.edu/sites/default/files/files/pub/d20/d2096.pdf ). The reason for this is differences in econometric strategy, which is often called the “levels vs. growth debate.” See Auffhamer 2018 ( https://www.aeaweb.org/articles?id=10.1257/jep.32.4.33 ) for a nice discussion of the reason why these vary. In general, this is an active debate in the field and there is not consensus on the appropriate way to predict economic damages from climate. 2018 Nobel winner William Nordhaus, for instance, would fall in the “levels” camp, and young superstar researchers Solomon Hsiang from Berkeley and Marshall Burke from Stanford would fall in the “growth” camp.
Thank you for this comment. Relevant quote from my updated analysis above:
“The new paper’s social cost of carbon figure is controversial and has been criticized for being too high for various methodological reasons.[6] For instance, one very critical new paper also now estimates the social cost of carbon on a country-level, suggesting that the global social cost of carbon is only $24 (and, using various sensitivity analyses, values ranging from $3.38/tCO2e to $21,889/tCO2e).[7]
To account for the new paper overestimating or underestimating the social cost of carbon, below, we use sensitivity analysis to show how our model responds to over- or underestimating the true social cost of carbon by 10x.”
It might be helpful to flag where exactly you made the changes, and what the previous values were, and what key assumptions/estimates you changed so that it’s clear to readers who read the previous version of this post.
The changes have been very substantial, because the first version was a much more simple model.
The main difference are that the first version had an error as pointed out by AGB in the comments.
Here’s a comparison doc between the two version:
https://docs.google.com/document/d/1a-6xC7WjSq2-AtSuUPKaQf1ielxYFrElxfATVDBfMA4/edit?usp=sharing
Thanks Hauke! The comparison doc was very helpful.
Just an extra thought for those following up on this analysis:
I was wondering if this analysis stacks the deck against global health.
The basic idea is that SCC estimates aim to include all the costs of CO2 – these are discounted, but many of the damages come ~100 years in the future.
On the other hand, the analyses of global health mainly try to quantify the immediate effects on health and income. They don’t include the idea that greater health and income now can lead to compounding economic benefits in the future.
Another way of seeing this is that the SCC estimates include ‘medium-term effects’, whereas the the global health ones might not.
Or another way of seeing it is that once we’re willing to include long-term benefits in the equation, we’re actually in the longtermist regime, and should focus mainly on existential risks.
In future attempts to compare climate change to global health, I think it would be useful to distinguish different worldviews used to make the assessment, which might be something like:
Near termist
Long termist
Conventional economic CBA
Very interesting.
In my experience the debate about climate damage is quite ideologically loaded so I am a priori very skeptical of using a single study for these kind of estimates, they always come with a host of assumptions that are ultimately fairly arbitrary.
In addition when you have a causal chain of at least three steps (climate sensitivity > impact of warming > human response) each with significant uncertainty that affects the next step it seems easy to get the estimate wrong by much more than one order of magnitude so the strategy of saying ‘even if we take a tenth of the estimate’ to save us from overestimation does not seem sufficient to me.
I would agree with this. My understanding is that the IAMs are so unmoored from reality as to be basically useless. They don’t try to account for the risk of catastrophic impacts, and the damage functions are chosen in part for mathematical tractability rather than fidelity to what climate change will actually be like. This is why I would object to claims such as “new research shows that the social cost of carbon is $477”.
This also seems like an area in which expert elicitation won’t be very accurate. We’re talking about impacts 100 years into the future for a problem heavily dependent on political developments which are extremely difficult to predict.
I agree that climate modelling is very uncertain, but we should not throw out the baby with the bathwater.
Quote from my analysis above:
“one study estimated a lower bound of the global social cost of carbon at US$125 and argues that:
“Quantifying the true SCC value is complicated because of various difficult-to-quantify damage cost categories and the interaction of discounting, uncertainty, large damages and risk aversion [...] The best that can be offered is a lower bound based comes from a conservative meta-estimate that aggregates studies using high and low discount rates, it does not account for various climate change damages owing to a lack of reliable information, and it does not consider a minimax regret argument addressing damages associated with extreme climate change.”
Also, as an aside, outside of prioritization, for optimal policy (e.g. carbon pricing) the social cost of carbon should be:
Set to the marginal abatement cost, which can be optimal and easier to estimate.[17] or
Set to err on the side of overestimating externalities[18] (while reducing other non-Pigovian taxes).”
I now include more optimistic estimates (in the sense that the SSC won’t be that high) in my sensitivity analysis.
Wow, I have never seen someone do this before! This is really impressive, excellent job being willing to reverse your conclusions (and article). Max upvote from me.
Tl;dr: This assumes pure rate of time discounting. I curious how well your analysis works for anyone who do not think that we should discount harms in the future simply by virtue of being in the future.
--
1.
THIS IS SO GOOD
This is super good research and super detailed and I am hugely impressed and hope many many people donate to Let’s Fund and support you with this kind of research!!!
--
2.
LET’S BE MORE EXPLICIT ABOUT THE ETHICAL ASSUMPTIONS MADE
I enjoyed reading Appendix 3
• I agree with Pindyck that models of the social cost of carbon (SCC) require a host of underlying ethical decisions and so can be highly misleading.
• I don’t however agree with Pindyck that there is no alternative so we might as well ignore this problem
At least for the purposes of making decisions within the EA community, I think we can apply models but be explicit about what ethical assumptions have been made and how they affect the models conclusions. Many people on this forum have a decent understanding of their ethical views and how that affects decisions and so being more explicit would support good cause prioritisation decisions of donors and others.
Of course this is holding people on this forum to a higher standard of rigor than professional academic economists reach so should be seen as a nice to have rather than a default, but lets see what we can do...
--
3.
DISCOUNTING THE FUTURE, AND OTHER ASSUMPTIONS
3.1
My (very rough) understanding of climate analysis is that the SCC is very highly dependent on the discount rate.
(Appendix 3 makes this point. Also the paper you link to on SCC per country says “Discounting assumptions have consistently been one of the biggest determinants of differences between estimations of the social cost of carbon”).
The paper you draw your evidence from seems to uses a pure rate of time discounting of 1-2%. This basically assumes that future people matter less.
I think many readers of this forum do not believe that future people matter less than people today.
I do not know how much this matters for the analysis. A high social cost of carbon seems, from the numbers in your article, to make climate interventions of the same order of magnitude but slightly less effective than cash transfers.
3.2
I also understand that estimates of the SCC is also dependent on the calculation of the worse case tail-end effects and there is some concern among people in the x-risk research space that small chances of very catastrophic affects are ignored in climate economics. I do not know how much this matters either.
3.3
I could also imagine that many people (especially negative leaning utilitarians) are more concerned by stopping the damage caused from climate change than impressed by the benefits of cash transfers.
SO:
I do not have answers to what effects these things have on the analysis. I would love to get your views on this.
Thank you for you work on this!!!
Thank you for your kind words Sam.
On discounting:
Of course, philosophically, pure time discounting is wrong, but:
“Another reason to discount is that far future benefits are more speculative, and changes to the world in the meantime can disrupt your project or make it irrelevant. For example, a vaccine development project that hopes to deliver a vaccine in a few decades faces a higher risk of being defunded or the disease in question disappearing, than does a similar project that expects to deliver a vaccine in a matter of years. This is a good reason to discount future benefits and costs, but the appropriate rate will vary dramatically depending on what you are looking at, and will not necessarily be the same every year into the future.”
https://www.givingwhatwecan.org/post/2013/04/was-tutankhamun-a-billion-times-more-important-than-you/
The social cost of carbon is generally highly sensitive to the pure rate of time preference.
But:
“The national social costs of carbon of faster growing economies are less sensitive to the pure rate of time preference and more sensitive to the rate of risk aversion” from Tol
and from the Ricke paper:
“CSCCs were calculated using both exogenous and endogenous9 discounting. For conventional exogenous discounting, two discount rates were used, 3 and 5%. the results under endogenous discounting were calculated using two rates of pure time preference (ρ=1, 2%) and two values of elasticity of marginal utility of consumption (μ=0.7, 1.5) for four endogenous discounting parameterizations.”
So maybe it’s not highly sensitive to just discounting anymore.
But both the Ricke and the Tol paper use sensitivity analyses on their SCC and different parameterizations make SCC 10s to 1000s of $ per tonne and I guess they’ll use “sensible” ranges for this.
I would love others to look into this more as well and could well imagine new research uncovering facts that would dominate this analysis.
I have included this estimate along with some commentary about why I disagree with your conclusion in an overall review of climate cost-effectiveness analyses.
Hauke’s calculation simply determines a standard Benefit/Cost ratio. If it costs $10 to avert a tonne of CO2 that provides benefits of $417 (in damages averted), this Benefit/Cost ratio equals 41.7. This ratio should be directly comparable to Copenhagen Consensus ‘Social, economic, and environmental benefit per $1 spent.’ For the Post-2015 Consensus, ‘Climate Change Adaption’ is listed as providing a Benefit/Cost ratio of 2 while climate-related ‘Energy Research’ has a ratio of 11. I would weight these results from meta-level research must more strongly than that from a single study. But even if we believed Hauke’s study, a benefit/cost ratio of 41.7 still lags ‘Reduce Child Malnutrition’ (ratio of 45) or ‘Expanded Immunization’ (ratio of 60). This hardly suggests that “we should consider prioritizing climate change over global development interventions.” The unconditional cash transfer benchmark that Hauke uses is a minimum and not representative of highly cost-effective interventions in global development. Using GiveWell’s estimates, deworming and malaria nets are more than 10x more cost-effective than cash. Before rushing to replace well-established priorities and interventions that are based on decades of research, we need to have substantial confidence in the new priority/intervention. This study is far from it.
Note that the Copenhagen Consensus and GiveWell results do not apply utility adjustments. If this new climate change study does so, its Benefit/Cost ratio would be distorted by improperly inflating Benefits, which make the ratio larger than it actually is.
Thank you- I’ve now included this in my model:
“Some global development interventions have been estimated to be 17.5x more effective than cash-transfers (e.g. deworming).[34] We use this as the optimistic case.”
When a post like this on cause prioritization appears on EA forums, I expect it to go something like this: “the marginal costs of the best interventions are lower in category A than category B, under the following assumptions (though if we tweak the assumptions in a certain way, B looks better)… so it’s likely that we should fund those best interventions in category A rather than B.”
This post, however, doesn’t appear to be based on the current EA thinking about the marginally-best interventions for climate change or global development. On climate, the last thing I heard was that clean energy R&D is most sorely needed… which agreed with my prior that molten salt reactors are awesome, scalable and sorely needed, plus the potential of things like enhanced geothermal energy to provide a transition path for the oil & gas industry, and then there’s alternative fusion technologies like DPF and EMC2′s Polywell (whose value to society is either zero or astronomical). Yet I don’t see these on the chart of interventions. I don’t recall what the best available interventions for global development are thought to be, but I thought I heard someone say that cash transfers likely weren’t it.
And although we should be skeptical of “free lunch” interventions with negative costs, that doesn’t mean they aren’t real. Some of them could be great investments, and if so, should we not identify which ones are realistically negative-cost and invest in them?
(I only read the first half; sorry if I missed something important.)
You’re absolutely right—generally, my agenda was probably a bit simpler than people might have supposed. This was not intended to be the last word on whether climate change or development interventions are always better. Rather it’s a starting point and “choose your own adventure” model to help prioritizing between a concrete climate and a concrete development charity.
Note that there are four parameters that drive the results of this analysis (the SCC, the income adjustment eta, the cost to avert CO2, and the effectiveness of global dev/health vs. cash). For the first two, there really is a lot more uncertainty, but for the latter two, it’s more clear. This makes the model actually valuable and with action guiding potential.
For instance, if you’re a small donor and can’t decide between GiveDirectly and the Coalition for Rainforest Nations (or perhaps the Let’s Fund Clean Energy campaign that you cite), then, if you believe that CfRN really has a cost-effectiveness of $0.02 / tCO2e averted, in many scenarios, especially the realistic one around which there is most consensus, it will often beat unconditional cash-transfers, even if you believe that social cost of carbon is quite low.
However, CfRN does lobbying, not a scalable intervention that one could invest a lot of money in. So, in contrast, if you’re a billionaire and are looking to decide between global development and climate change as a cause area for your foundation, then perhaps global development might be a better bet.
Thanks for this. One thing that perplexes me about the Ricke et al. (2018) analysis is that the SCC for most African countries looks to be lower than for the USA (fig.2), whereas the general consensus seems to be that the impacts of climate change will have far worse effects on individuals’ utilities in Africa. So this makes me wonder have they properly captured the effect of marginal utility changing with income? I’m not an economist, so I don’t know how to judge this myself.
Excellent point
I think it’s a mixture of the following:
1. African countries are relatively small
2. the social cost of carbon measures the cost to GDP—and if your GDP is not very big to start with then there’s not a big cost to you.
It is quite unintuitive/ disconcerting that the official social cost of carbon for the DRC (one of the poorest countries, 80 million people, close to the equator and particularly affected by climate change), only has a social cost of carbon of 30 cents per tonne, whereas the US has one of $40 - see:
https://country-level-scc.github.io/explorer/
As I said above, there are contributors to (true) social cost of carbon not fully captured by empirical, macroeconomic damage functions, and their likely impacts on the social cost of carbon (see Table S5 in the paper’s supplementary material and Table 1 in[21]). For instance:
Adjustment costs (short-term costs of adaptation)
Non-market damages (biodiversity loss, cultural losses, etc.)
Tipping points in the climate system (catastrophic climate events, hysteresis etc.)
High inertia effects of CO2 (ocean acidification, sea level rise)
General equilibrium effects (spillover, trade, etc.)
Macro-scale adaptation (long-term restructuring of economy)
Political instability and violent conflicts
Large migration flows
More extreme weather and natural disasters
Bresler finds that explicitly accounting for climate mortality costs triples the welfare costs of climate change.[22]
The highest social cost of carbon estimate in the literature is on the same order of magnitude ($1687[23]), and the highest figure amongst many in a recently published paper find that for 6 degrees of warming the cost will be (which has a substantial probability) is $21889 / per tonne) [24]
That’s why in the pessimistic version of my model increased the SCC by 10x (higher than most estimates).
Even after the update, I’m confused by this model. I’ll restrict my attention to the “Realistic case” for simplicity.
The global social cost of carbon when constructed from the country-level estimates is effectively:
CCG=CCa+CCb+CCc+…
where CCG is the global social cost of carbon and CCa, CCb, CCc, etc. are the country level social costs of carbon for countries ‘a’, ‘b’, ‘c’, etc.
The spreadsheet model then uses an income adjustment factor of 1,260 expressing that a dollar means more to a poor person than a rich person. We can also think of this as a unit conversion factor: 1,260 American median income dollars = 1 GiveDirectly recipient dollar. We can write this conversion factor as 1260$A1$GD=1.
The next step in the model is to divide the global social cost of carbon by the income adjustment factor. If we expand global social cost of carbon, this looks like:
CCG1260=CCa1260+CCb1260+CCc1260+…
If we just look at the units, this is:
$G$GD$A=$a$GD$A+$b$GD$A+$c$GD$A+…
where $G is a country-level-cost-of-carbon-weighted dollar, $a is a dollar in country ‘a’, etc. Converting country ‘a’ dollars to GiveDirectly recipient dollars via $GD1260$A is only appropriate if country ‘a’ is in fact America. Otherwise, the units don’t line up.
It seems like what we’d actually want as far as income adjustment is something like:
CCG−GD=$a$GD$a+$b$GD$b+$c$GD$c+…
where CCG−GD is the global social cost of carbon expressed in GiveDirectly recipient dollars.
In other words, we can’t just apply the same income-adjustment factor to every country’s social cost of carbon because not every country has the same income. We need per-country income adjustment factors. Applying the 1,260 income adjustment factor to American social cost of carbon works but in applying this adjustment to the global cost of carbon, we are also implicitly saying that the social cost of carbon in Burkina Faso ought to be discounted just as heavily.
I think the only way the current approach works is if we assume that country-level costs have somehow already been adjusted into American median income dollars.
Thanks for this comment and apologies for the delay in replying.
As mentioned I had emailed the authors and they didn’t get back to me, and also the author of this paper this paper, “A social cost of carbon for (almost) every country” had apparently trouble replicating their analysis.
But both these papers do cite individual costs of carbon for each country—for instance here:
https://country-level-scc.github.io/explorer/
There you can for instance see that India’s SCC is $85.4 . I think this is based average income till the end of the century taking into account growth (“socioeconomic projections” (perhaps $20k/capita ).) and how much they’re hurt by climate damages given how rich they are. This figure is then income adjusted so that it is comparable with present day Americans (it might actually downward adjust the damages of rich countries which will be richer than present day US by the end of the century). Because then it can be interpreted as “this is what the Indian government should spend right now in USD to avert per tonne”, which seems to make most sense.
These costs are then all added together to arrive at the global social cost of carbon and what we should spend collectively to avert a tonne of CO2. This is why I think we can compare this with Givedirectly and the analysis is roughly correct.
I think my current model is suboptimal in that it mixes up different income adjustment factor eta = 1.5 and combining it with a different eta they use in the paper. It might be good to standardize this in one analysis. This might change the results.
I’m sorry that I didn’t understand quite the point you’re making, but you might legitimately be onto something. I think to get to the bottom of this it might be best to create a new model with open and transparent code to figure this out. I saw that you did very good quantitative analysis in a different post, so perhaps you could take this on. But maybe this would be more something for a slightly bigger org with more research capacity to take on. Not sure how productive this is or whether there’s really demand for such an analysis though.
Not sure which explanation will work better so I’ve included both a prose description of what I think is going on and a model. Look at whichever seems most useful.
Words
I think this is the crux of the discrepancy. I interpret the country-level social cost of carbon as expressing the cost in terms of that country’s citizens’ consumption. More concretely, if the social cost of carbon is $85/tonne of CO2 in India and everyone in India emits 5 tonnes of CO2, everyone’s welfare in India is as though their income is actually $425 (5 * 85) lower. I understand you to be saying that if everyone in India emits 5 tonnes at $85/tonne, it’s actually as though their income is reduced by $12.5 (5 * 85 / 34 where 34 is the poverty multiplier using an eta of 1.5 and an Indian income of $2000). The lower number comes because you believe the $85/tonne to be quoted in terms of dollars for the American median income and it needs to be adjusted to account for the fact that each dollar “means more” in lower-income India. Is that description intelligible? Is it an accurate account of your views?
Supposing that’s right, I can now explain why I believe my version. I think the relevant parts of the paper and model are the damages module and the discounting module.
I don’t think the sort of adjustment-to-America (i.e. “income adjusted so that it is comparable with present day Americans”) would happen in the damages module. Indeed, a quick skim of the damage function reference shows it to be talking purely about macroeconomic indicators. So the output of the damages module very likely is in local terms (e.g. a 1% decline in India’s GDP) rather than American terms.
Based just on the name, if this sort of adjustment-to-America were present, I’d expect it to take place in the discounting module. But the Ricke paper doesn’t describe any sort of adjustment for disparate incomes at time T=0. The income-relevant adjustments they perform are an: (1) an optional rich-poor damage specification which affects how damages grow over time for countries in each bin; (2) an optional elasticity of marginal utility adjustment to account for how the cost diminishes as economies grow [1]. So it doesn’t look like the discounting module does the required adjustment-to-America either.
Model
I’ve also made a quick spreadsheet model trying to explain what I think is going on.
The first two tabs are just data—the country level social costs of carbon and the GDP per capita (median income probably would have been better but alas).
The next tab to look at is probably “Hillebrandt Analogue”. It replicates your calculation using country-level data. It applies your uniform “Realistic” poverty multiplier to each country and then sums rather than summing before applying the poverty multiplier. Reassuringly, we get basically the same result: a global social cost of carbon of $0.32.
The other two tabs show roughly how I think the modeling should be done.
“CSSC in American Dollars” uses American GDP per capita as a reference income to compute separate poverty multipliers for each country based on its GDP per capita. The multiplier is 1 for America, less than 1 for countries richer than America and much greater than 1 for countries much poorer than America. Applying each country’s poverty multiplier to its country-level social cost of carbon and summing gives us a global social cost of carbon $36,834. If 100% of the social cost of a tonne of CO2 fell on an American making $62,641 (US GDP per capita), this would be the welfare-equivalent of a $36,834 reduction in income.
“CSSC in GiveDirectly Dollars” follows the same procedure but uses GiveDirectly annual consumption ($180) as the reference income. Thus the poverty multiplier is near 0 for most developed countries and is only over 0.5 for extremely poor countries like Burundi. Applying each country’s poverty multiplier to its country-level social cost of carbon and summing gives us a global social cost of carbon of $5.67. If 100% of the cost of a tonne of CO2 fell on a typical GiveDirectly recipient, this would be the welfare-equivalent of a $5.67 reduction in income.
Interestingly, this $5.67 figure makes climate change interventions within a factor of 2 of GiveDirectly in terms of cost-effectiveness (assuming $10 tonne/CO2 averted).
I think it’s pretty clear from context that this income adjustment is only applied within countries over time rather than across countries at time T=0: “We thus used growth-adjusted discounting determined by the Ramsey endogenous rule, with a range of values for the elasticity of marginal utility (μ) and the pure rate of time preference (ρ), but we also report fixed discounting results to demonstrate the sensitivity of SCC calculations to discounting methods.”
Thank you so much for this thoughtful comment.
Unfortunately, I won’t have the time to give it the time it deserves to engage with it properly (e.g. rework the analysis etc.).
I think you raise interesting points and people looking to extend this work further should take your comment into account.
Okay, thanks. I will likely write this up as a post then so that the concern doesn’t get lost here in the comments.
Thanks this is a great idea!
I’m really sorry that I don’t have more time for this.
Thanks, this is interesting! I quickly read through the core paper and am a bit confused.
It seems like you’re understanding income adjustment to be one of the main additions in the paper. Where are you seeing that? The title/abstract/etc. seem to be pitching greater spatial resolution as the main contribution. Greater spatial resolution helps with income adjustment but isn’t sufficient. As far as I can tell the paper primarily uses regular old GDP per capita (with the de rigeur acknowledgement that GDP isn’t a great welfare measure). The only income adjustment I see is a couple of mentions of rich/poor specifications and the supplementary information suggests that this is just splitting the formula for the growth of damages based on whether a country falls into the rich bin or poor bin.
They explain the increased cost not as due to income adjustment (as I understand things) but because:
All that said, I wish it did use logarithmic utility because that seems like an important improvement!
(FYI: All the inline footnotes still link to a Google doc.)
This post was awarded an EA Forum Prize; see the prize announcement for more details.
My notes on what I liked about the post, from the announcement:
This is really interesting. I’m curious about crowding out and marginal dollar effects. i.e. the smart money spends all its resources on this, allowing the dumb money to free ride and keep on with the status quo (or even get worse with less perceived consequences). Meanwhile, there are now far less smart dollars available to fund weird moonshots that only the smart money can think about.
One solution: more funding for geoengineering moonshots (and please, with fewer assumptions that geoengineering automatically means that safety and reversibility aren’t major design criteria).