Why we have over-rated Cool Earth

[Sanjay]

0. Exec summary

Cool Earth is a non-profit that tackles climate change; it has been enthusiastically supported by many in the Effective Altruism community. In this post, I will say the following:

1. How Cool Earth works

2. According to earlier analysis by Giving What We Can (GWWC), Cool Earth is a high-impact charity

3. However this analysis omits certain aspects/​risks

4. Once you take these into account, Cool Earth’s work is either fairly low impact, or there is (currently) not enough evidence available to say; this depends on which project (i.e. country) you’re looking at

5. Some other options that climate-interested donors may wish to consider

6. At this stage the information gathered doesn’t allow me to give a strong opinion on Cool Earth, but I would recommend that we now consider the earlier GWWC analysis out of date

As for the omitted aspects/​risks (item 3), these include the displacement effect—i.e. when loggers are unable to log the trees protected by one community, they will simply go elsewhere. A secondary concern is risks post-exit. (read on for more explanation of these)

Many thanks to Cool Earth for their input, and note that Cool Earth do not agree with all that has been said in this piece. I will try to represent their views as well.

I would have preferred to investigate these issues more fully before sharing this piece, given that some of the information gaps are quite material, however that has not been possible, largely because of time constraints on the part of Cool Earth. I would also like to thank Cool Earth for the time they have given me for this analysis.

As for whether to donate to Cool Earth or not, I would suggest not to based on the current evidence, although as more evidence arises, this opinion could change.

Donors interested in climate change may be interested in the recent analysis conducted by Founders Pledge. I will also list below some other options that donors interested in supporting climate change might consider, i.e. providing contraception and female education.

1. How Cool Earth works—intro

Cool Earth is a non-profit that tackles climate change by working with rainforest communities to avoid deforestation. They do this by supporting the livelihoods of the people in the rainforest.

The logic is that local communities in the rainforest will be less likely to permit logging if they are not in need of money, since they fundamentally want to preserve the rainforest as well.

This results in rainforest being preserved, which is positive from a climate change perspective.

2. Why is it considered a high-impact choice of charity to donate to?

According to analysis by Giving What We Can (an Effective Altruist organisation which used to do research into charities) Cool Earth was considered a high-impact choice.

The rationale is that the Cost per tonne of CO2 is low compared to other opportunities, i.e.

• Cost per tonne of CO2 for Cool Earth is ~ $1

• Cost per tonne of CO2 for other opportunities is typically more expensive, e.g. ~ $10 (esp direct work, not campaigning)

See Appendix 2 for more about GWWC’s analysis.

My main claim is that GWWC’s claim of c. $1 per tonne of CO2 claim is optimistic. Indeed Cool Earth’s overall impact from a climate change perspective may be negligible, although this depends on various factors.

The GWWC analysis was last updated in 2016, and GWWC no longer conducts analysis of charities, so GWWC were not available to comment on this piece.

3. Reasons to doubt Cool Earth’s impact

• 3a Reason to doubt 1: Displacement risk

• 3b Reason to doubt 2: Risks post-exit

• 3c Polarisation: how calcs lead to all-or-nothing views

3a Reasons to doubt 1: Displacement risk

- Local communities (e.g. the Ashaninka in Peru) will be less likely to permit logging if they are not desperate for money

- Cool Earth helps them to be not desperate for money

- So the loggers will be unlikely to perform logging in *those communities*

- However they would probably just go somewhere else and log there instead.

If we were to model this as saying that loggers would *definitely* go and log somewhere else, then the model should assume that Cool Earth has achieved nothing from a climate change perspective. While it’s unlikely that 100% of loggers would “be displaced” (i.e. just log elsewhere) I suspect it applies to most loggers.

In the appendices, I have considered factors which I thought might mitigate this concern, which I call:

• exogenous-logger-gives-up scenario

• long-term-broad-coverage scenario and the

• endogenous-logger-would-be-material scenario.

Having looked into each of those scenarios (see appendices for more), it seems that their effect is small in Peru and Papua New Guinea, so the displacement effect is likely to dominate. I.e. the displacement effect is likely to mean that Cool Earth is having a small amount of impact on climate change. (See appendix A3(ii) for more on this)

This might not be the case for all the countries where Cool Earth operates—more on this in the next section (section 4).

I have been in touch with Cool Earth, and they have been kind enough to give me some time to discuss these concerns. They disagree with this assessment, especially around the endogenous-logger-is-material scenario. More info in the appendices, especially appendix A3(iii), including why I’m not convinced by their arguments.

3b Reasons to doubt 2: Risks post-exit

Cool Earth’s model involves supporting a community for some years and then leaving and moving on. Once that’s happened, will the support that Cool Earth provides still persist?

By including a probability of logging of only 0.5% with Cool Earth’s intervention (see the summary of GWWC’s analysis), we are essentially placing a high probability that the intervention will work *not only now but in the future*. If in fact the effects are not permanent, then the cost-effectiveness model should be different from the one used by GWWC. Instead it should take into account the expectation that logging will restart, but model it as happening at some point in the future, and discount that back to the present day. Alternatively we could adjust the model by increasing the costs to reflect the fact that Cool Earth will be required to work with those indefinitely, rather than just for a limited period of time as originally assumed.

There are actually not very many partnerships that Cool Earth has exited from (only 4) and even those are not good examples because they were imposed by external circumstances. Although the little data we have about those seem positive, we don’t have enough evidence to know whether the preventative effects will persist after Cool Earth’s work has finished.

Quote from Cool Earth on this topic:

“At the very beginning of Cool Earth’s existence ten year’s (sic) ago we worked in Ecuador and Brazil briefly but had to end these early due to political issues beyond our control. For one of the Brazilian communities we worked with for two years, back in 2007, we have satellite images that show the forest to still be intact. This is a good sign that the communities remained resilient but of course we need many more data points from current and future partnerships to be able to provide really strong evidence for the long-term effectiveness. This is something we are putting together now by building a portfolio of partnerships globally which will test how much our support creates resilient communities and how this needs tweaking under different circumstances and slightly different methods (eg direct funding to households instead of to community associations) to improve its effectiveness. ”

In order to get to a clearer picture on this, I would need to better understand the assumptions behind the costings. I have requested this Cool Earth; they have not been able to provide this information.

3c Polarisation: how the calcs lead to all-or-nothing views

One of the quirks of the way the calculations work is that we are likely to believe either that Cool Earth’s work is comfortably more cost-effective than tree-planting or that it has minimal impact. There is only a very small range of input assumptions which lead to conclusions in between. This point is slightly technical, so I have expanded on it in an appendix (see appendix 5).

4 Impactfulness varies by country

The details differ between the different countries where Cool Earth operates. Cool Earth have been kind enough to discuss this with me, although we have not had time to cover everything.

Displacement risk is a key driver of whether Cool Earth has some impact.

(More notes on the “marginal impact of the next pound” in Appendix 6.)

Papua New Guinea: I think it’s unlikely that a palm oil company would be put off from doing their work just because an area can no longer be logged, so it seems the displacement risk is high. Note that Cool Earth argue that there is a potential systemic impact, however I don’t find this convincing. More on this in Appendix 4

Mozambique/​Cameroon: These are new countries that Cool Earth is expanding into. Cool Earth has informed me that the logging that they are tackling is mostly endogenous—if true, this would suggest that the cost-effectiveness is good. I would have liked to have investigated this further, however our conversations ran out of time and I was unable to dig into this deeper, which is unfortunate.

Peru: Cool Earth has suggested that the logging would be endogenous to a material degree. This seems unlikely to be the case, in my view. See Appendix 3(iii) for more detail.

Democratic Republic of Congo: Cool Earth has informed me that the logging that they are tackling is mostly endogenous—if true, this would suggest that the cost-effectiveness is good. I would have liked to have investigated this further, however our conversations ran out of time and I was unable to dig into this deeper, which is unfortunate.

5 Other options for donors to consider

The purpose of this piece is not to provide the full answer to the question of how to donate if you are keen on tackling climate change. However here are a few (not necessarily thoroughly researched) ideas:

• Founders Pledge has given this question plenty of thought and has some recommendations (link)

• Providing contraception for free to people who otherwise don’t have access to it could be competitive with Cool Earth’s cost-effectiveness, even under GWWC’s assessment of Cool Earth. More on this in Appendix 7.

• Funding education for girls has been proposed by drawdown.org as a high-impact choice. Some of the arguments made in Appendix 7 would apply here too. I don’t know how this would stack up in terms of cost-effectiveness, although I’d guess it would likely be on the same order of magnitude as providing contraception, with similar non-climate-change benefits.

The main purpose of mentioning these is to indicate that there *are* other potentially high-impact opportunities for climate-interested donors (i.e. it’s not that we are looking at Cool Earth because there are no other promising donation candidates)

Conclusions

Cool Earth’s work is either fairly low impact, or there is (currently) not enough evidence available to say; this depends on which project (i.e. country) you’re looking at.

For the countries for which we don’t have enough evidence, it may be that more evidence would show Cool Earth to be more cost-effective. While Cool Earth staff have been very kind in giving me their time, there has been a limit to how time they have been able to provide—these conclusions may evolve with more information.

Based on the information as it stands, I would be more inclined to recommend that donors not donate until more analysis is done. I also note that there are several other promising interventions that could be at least as cost-effective as Cool Earth’s work.

----------------------------------------------------------------------------------------------------------------------------------

----------------------------------------------------------------------------------------------------------------------------------

Appendices

There now follow some appendices.

Appendix 1: More background info about loggers and logging

Cool Earth operates in 3 locations:

2 partnerships in Peru

2 partnerships in Papua New Guinea

1 partnership in Democratic Republic of Congo

They plan to expand to Mozambique and Cameroon

In Peru, the loggers are mostly small-scale loggers, possibly one-man-bands, not necessarily operating legally.

(Side note: if loggers are willing to operate outside of the law, it’s possible that they would be willing to log even though the local community doesn’t permit it. I have not investigated this in any detail, however I understand that Cool Earth does careful monitoring of the rainforest in the areas where they work, which gives some cause for reassurance about this risk.)

In Papua New Guinea, the loggers are more likely to be large multinational companies.

(For DRC I have not checked—it is unfortunate that I’ve not been able to get more time with Cool Earth)

Appendix 2: Summary of GWWC’s $1 per tonne of CO2 analysis

The below table summarises the analysis done by Giving What We Can (GWWC). You can see the full sources for Giving What We Can’s (GWWC’s) analysis here: https://​​www.givingwhatwecan.org/​​report/​​cool-earth/​​

Appendix 3: Displacement risk—more details

Displacement risk is the risk that someone intending to log in a given area is rebuffed from that area because of Cool Earth’s work, but then the logger simply goes elsewhere.

At first glance, I think that the most likely scenario is that most loggers are not put off (i.e. they are simply displaced and perform logging elsewhere), largely because they have already travelled a long distance to get there, so to expect them to be put off by the thought of travelling just a little further seems surprising.

However I consider here some possible mitigating factors:

• A3(i) Exogenous-logger-gives-up scenario: The logger may have attempted logging somewhat opportunistically, and once the nearest/​easiest logging opportunity has been rebuffed, they simply choose to do something else instead of logging.

• A3(ii) Long-term broad coverage scenario: The immediate impact of logging may be that the logger just goes elsewhere. However if Cool Earth were to continue to expand enough, eventually there would be so few areas left, that the logger may be forced to give up on logging. If that occurs, it seems fair to give some credit now to that future possibility.

• A3(iii) Endogenous-logger-would-be-material scenario: By “endogenous loggers” I mean people from within the community that Cool Earth supports. If Cool Earth has materially reduced the number of endogenous loggers, it would suggest that Cool Earth’s intervention is impactful, because those people are less likely to go elsewhere.

A3(i) Logger-gives-up scenario

For our analysis, we need an estimate of the probability of the logger-gives-up scenario. I have tried discussing this with Cool Earth, however I’m conscious that they are experts on the communities that they support, and not experts on the communities who do the logging. So they are not best placed to answer this question.

However I’m conscious that the logger has travelled some distance to perform the logging. For example, in Peru the loggers are typically small-scale loggers who may have travelled from the Andean region—i.e. potentially a long distance across the width of the country—in order to reach the forests. In Papua New Guinea, the loggers are large, serious logging companies. Neither of these groups are likely to be casual/​opportunistic about the intent to log.

I would be inclined to assume that this sort of giving up doesn’t happen to any material extent, although if there was evidence that this was an (at least) somewhat common occurrence, it might change the overall analysis. Similarly if we believed that loggers who have travelled a long distance, are forced to travel a bit further, and are therefore disinclined to travel again and more inclined to seek income from elsewhere, this would also change the analysis.

A3(ii) Long-term broad coverage scenario

If Cool Earth were to continue to expand enough, eventually there would be so few areas left where logging was possible that the logger may be forced to give up on logging. How far away are we from that?

The below bullet points indicate why I think that we are very far away from that. Hence considers the long-term broad coverage scenario to be far away enough that it gives it negligible credit.

• Peru

• The loggers are typically coming from the Andes and travelling several hundred km to get to there

• So the amount of extra travel needed would perhaps need to be several hundred more kilometres before it deters a logger

• One community will typically occupy around 100 acres of land

• This might require around 600m to traverse (or possibly a bit less)

• So they would need to add another several hundred communities

• They are currently covering 2 communities in Peru

• Papua New Guinea

• The loggers are typically large multinational logging companies

• So if it became harder to access forest in one location, the company would simply look elsewhere

• Democratic Republic of Congo

• Cool Earth has one partnership in DRC. I have not considered DRC closely in this analysis

A3(iii) Endogenous logger would be material scenario

Currently there is very little “endogenous” logging (i.e. logging by people from within the community which Cool Earth supports) -- it accounts for around 3% of the logging which occurs, according to Cool Earth.

The key question is whether endogenous logging would be material if Cool Earth weren’t working in those communities—if so, then the displacement risk is less of a concern.

From my conversations with Cool Earth I understand the following:

• They believe that endogenous logging would be material without them

• They were unable to provide data about the extent to which endogenous logging occurs in other nearby communities

• They were able to give an anecdote about another community who logged 25% of the land around them

This has not been enough to convince me that endogenous loggers are material. In particular:

• Cool Earth works with communities who are particularly prone to value the rainforest, this suggests that endogenous logging is likely to be low there. Indeed they have argued that the community’s desire to preserve the rainforest is part of the reason for their success.

• If the communities would have logged 25% of the rainforest around them without Cool Earth’s work, this would be more than enough to push Cool Earth into cost-effective territory; however I don’t believe the community that did this was typical.

• In my conversations with Cool Earth, they have referenced some more data about the deforestation that’s been occurring both within and outside of Cool-Earth-supported communities, and this might help to shed some light on this. At the time of writing, Cool Earth have not had time to share this with me.

Appendix 4: More on Cool Earth’s opinion on Papua New Guinea

Cool Earth have suggested that the work in Papua New Guinea is still valuable, even though displacement risk is high for the loggers. They have noted that a major palm oil company—New Britain—has decided that they will stop logging and only create palm oil on existing sites rather than perpetrate further deforestation. They believe that Cool Earth has a valuable role to play in achieving outcomes like these.

I am not so sympathetic to this claim. Cool Earth is not a campaigning organisation, and does not allocate a material amount of its resource to influencing governments or corporates. I suspect that New Britain would have made the same decision had Cool Earth never existed, and that Cool Earth’s continued existence is unlikely to influence other such palm oil companies.

Appendix 5: Polarisation—why Cool Earth is likely highly cost-effective or not at all

Appendix 2 sets out the model used by Giving What We Can (GWWC) when they did their analysis. This post has focused on how much reduction in overall probability of logging is actually generated by Cool Earth’s work, in other words:

• if the probability of logging without Cool Earth is 28% (line (B) from Appendix 2)

• how much does that probability reduce to with Cool Earth? (line (C) from Appendix 2)

The below chart shows how the cost per tonne of CO2 averted varies with assumption (C) from Appendix 2, while keeping all other assumptions fixed.

GWWC’s assumption was that the probability of logging *with* Cool Earth doing their work was 0.5%. This means that almost any reduction in the scale of Cool Earth’s work (e.g. instead of bringing the probability of logging down to 0.5%, let’s say you thought it was actually 10%) would still leave Cool Earth more cost-effective than a straightforward tree-planting scheme. In other words, for most of the chart, the green line is lower than (i.e. more cost-effective than) the blue line.

It’s only if Cool Earth’s work is having very little impact that we get a cost-effectiveness which is much worse than tree-planting.

There is still a zone where Cool Earth’s cost-effectiveness happens to be close to that of tree-planting, however this is a narrow window.

Note that I’ve included here a comparison with tree-planting mostly because this is an easy comparison to make—it’s not because I expect tree-planting to be the best climate change intervention outside of Cool Earth.

Appendix 6: What is the marginal impact of the next pound donated?

At the margin, the next pound donated to a charity is typically likely to support the following types of work:

-- those projects which the charity is seeking to grow or get off the ground (i.e. which are somehow “expanding” beyond what they are currently doing)

-- those projects which are less “core” to the charity (and therefore more likely to be at risk if money were short)

These are subtle things, and often not easy to determine. In the case of Cool Earth, because they are expanding in certain areas (Papua New Guinea, Mozambique and Cameroon) these are likely to be good candidates for the marginal impact of the next pound.

If we happened to know that the impact of the next pound was to fund work in Papua New Guinea, say, then the impact analysis on (say) Peru would be almost irrelevant. It would not be entirely irrelevant because it says something about the approach of the organisation being funded. And there is always some chance that the funds could end up supporting (say) Peru, or indeed anything else Cool Earth does.

Appendix 7: Providing contraception

This is a very shallow review, and needs more work before it could be considered definitive. However here are some rough and ready calcs to indicate that the cost per tonne of CO2eq averted is probably only around $1:

[EDIT: Some commenters have argued that the post would be better without this section. They are probably right. The spirit of this section is meant to be: “if we keep on thinking creatively about climate change we will probably keep finding creative solutions—I don’t know if contraception is really a good idea, but I’m just throwing an idea out to illustrate this.” I’m keeping this section in anyway.]

Cost per pregnancy averted?

1. Cost to fully meet the need for modern contraceptives: $9.4 billion [A]

2. Number of unwanted pregnancies per year: 74 million [B]

3. Cost per pregnancy averted ~ $100 (= (1) /​ (2) with some rounding)

Average CO2 footprint per person?

1. Total CO2 emissions = c 53 billion tonnes of CO2eq per annum [C]

2. Total world population = c 7 billion

3. Average CO2 footprint = c 7.5 tonnes per person p.a. (= (4)/​(5) with some rounding)

4. Average CO2 footprint for someone in the developed world = 10 tonnes per person p.a. [D]

5. Average CO2 footprint for someone in the developing world ~ 5 tonnes per person p.a. (rough estimate, see [E])

Cost per tonne of CO2eq averted?

1. Lifetime CO2eq footprint ~ 100 tonnes per person (roughly calculated, see [F])

2. Cost per tonne of CO2eq averted ~ $1 per tonne of CO2eq (= (3)/​(9))

[A] source: Guttmacher Institute, p5 of https://​​www.guttmacher.org/​​sites/​​default/​​files/​​report_pdf/​​addingitup2014.pdf . The source did not say that this was a per annum cost, but I’m guessing it probably is.

[B] source: Pathfinder International: “In developing countries, more than half of all women of reproductive age want to avoid pregnancy. 214 million women are not using an effective method of contraception. And millions of women lack the services they need, leading to 74 million unintended pregnancies in developing countries each year.” https://​​www.pathfinder.org/​​focus-areas/​​contraception-family-planning/​​

[C] World bank gives a figure of c 53,526,300,000 tonnes as at 2012 https://​​data.worldbank.org/​​indicator/​​EN.ATM.GHGT.KT.CE

[D] Online source says: “The average person’s carbon footprint is about 10 tonnes” https://​​www.carboncalculator.co.uk/​​averages.php

[E] I didn’t find a source for the average carbon footprint for someone who is living in a context where contraception is not readily available and needs to be provided by a charity. However as a rough guess I just chose 5 tonnes because the average of 10 tonnes and 5 tonnes comes to the desired 7.5 tonnes—this does not weight correctly for population sizes, but I think this will probably mean I’m understating the carbon footprint.

[F] The simplistic way to calculate the lifetime carbon footprint is to multiply the annual carbon footprint by the life expectancy in years, which would be c. 50 years for countries with the shortest life expectancies. However we should discount the future years because there is uncertainty about the future (e.g. maybe we will have solved climate change in 30 years’ time; or maybe we will have caused a climate catastrophe and wiped out humanity in 30 years’ time). So to account for this I’ve used a discount rate of 5%, and to get the right discounted present value would involve dividing by 5% (i.e. multiplying by 20) assuming people live in perpetuity, which would give a similar number to assuming that people live for 50 years.

Further notes:

• This is based on the cost to fully meet the contraceptive need—this may include more expensive “last mile” contraceptive needs and the best use of funds at the margin might be more cost-effective (although I haven’t checked how the $9.4 bn figure was calculated, so I don’t know if that’s correct)

• Calculations of this sort often ignore an important factor: each person adds a demand on the world’s resources, but also brings a small (maybe only 0.00001% or whatever) chance that the person will create a phenomenal breakthrough which will solve the problem. As an example, many have worried over the years about having enough food to feed a growing population. However if we had limited population growth as a result of this, we might have never had Norman Borlaug, and never had the Green Revolution. That said, for people born in the developing world, it’s likely that the best way to unlock their potential to help the world is probably to have fewer of them and make sure they are well educated rather than having more of them in countries with still-developing education infrastructures.

• A separate study has found that if you met the unmet need for contraceptives (no coercion whatsoever) then you could achieve large scale abatement at less than a dollar a ton. Source: https://​​www.vox.com/​​energy-and-environment/​​2018/​​11/​​16/​​18096352/​​climate-change-clean-energy-policies-guide. Note that I had seen this quote before I put together my numbers, so I probably sub-consciously tweaked my choice of discount rate to agree to this. (I.e. if you think that two independent assessments have come to the same figure, it’s probably better to say that they are not entirely independent)

• Note that this gives no credit to the many other benefits of the intervention, notably reducing maternal mortality, enabling female empowerment, and other sustainability benefits.

Many thanks to John Halstead and Johannes Ackva who reviewed earlier drafts of this post.

[John G. Halstead]

Thanks for writing this. I think this may be a cautionary tale of EA hubris, which we should beware of in the future. GWWC had two interns write reports on climate change presumably over the course of about two months, who had no prior knowledge of the area. The movement then took on the Cool Earth recommendation as though it was definitely the best thing within climate change. (This is not a criticism of the authors of the reports, who I think did a good job given the constraints).

There are a lot of foundations distributing billions of dollars in this area, with grant-makers working full time to find opportunities. Climate change is a very complicated area and it is very hard to know who is diong what. I think it would be surprising if GWWC found the best possible thing given its constraints.

I also think it is an argument forchecking lots of conclusions and premises with relevant experts.

We might want to consider whether we are doing something similar in other areas. e.g. is our approach in development, far future or animal welfare naive according to some experts that we haven’t fulled engaged with yet?

[Taymon]

I don’t think I would call this hubris. We all knew that the Cool Earth recommendation was low-confidence. But what else were we going to do? To paraphrase Scott Alexander from another recent community controversy, our probability distribution was wide but centered around Cool Earth.

I do think that that nuance occasionally got lost when doing outreach to people not already very informed about EA, but that’s a different problem. We haven’t solved it, but I feel like that’s because it’s hard, not because nobody’s thought about it.

(One could also argue that outreach to mainstream audiences about EA shouldn’t discuss climate change at all, given its place in the movement, but the temptation to make those mainstream audiences more receptive by talking about something they already care about is strong.)

[John_Maxwell]

We all knew that the Cool Earth recommendation was low-confidence.

I just glanced at the part of Doing Good Better that discusses Cool Earth. It doesn’t seem that low-confidence to me, and does seem a bit self-congratulatory/​hubristic.

We haven’t solved it, but I feel like that’s because it’s hard, not because nobody’s thought about it.

I think I’ve observed information cascades in Effective Altruism relating to both global poverty and AI risk. The thinking seems to go something like: EA is a great community full of smart people. If a lot of smart people agree on something, it’s probably right. It’s hard to beat a big group of smart people, so it’s probably not worth my time personally to consider this issue in detail, so I will use the opinion of a big group of smart people instead. Then when I offer arguments against some position, the person is like: “I know you’re wrong because a big group of smart people disagrees with you”, without considering my arguments.

Big groups of smart people are great, but only because they have intelligent disagreements in order to come to correct opinions. You should trust an idea because it has survived a lot of diverse challenges, not because a lot of people profess it (and especially not because a lot of people in your social circle profess it, since social circles are self-selected). IMO, if you aren’t personally knowledgable regarding a diverse set of challenges an idea has survived, you shouldn’t confidently say that idea is correct, otherwise you are part of the problem. Instead you can say: “some people I trust believe X”.

If I’m right, fixing this problem just requires talking about it more. I’m not saying it is a huge problem, but I think I’d prefer that people discussed it a bit more on the margin.

[Taymon]

I don’t think nobody delved into the Cool Earth numbers because they assumed a bunch of smart people had already done it. I think nobody delved into the Cool Earth numbers because it wasn’t worth their time, because climate change charities generally aren’t competitive with the standard EA donation opportunities, so the question is only relevant if you’ve decided for non-EA reasons that you’re going to focus on climate change. (Indeed, if I understand correctly the Founders Pledge report was written primarily for non-EA donors who’d decided this.)

Whatever’s been going on with global poverty and AI risk, I think it’s probably a different problem.

(And yes, Doing Good Better was part of what I was referring to with respect to nuance getting lost in popularizations. It’s that problem specifically that I claim is difficult, not the more general problem of groupthink within EA.)

[Arepo]

I think nobody delved into the Cool Earth numbers because it wasn’t worth their time, because climate change charities generally aren’t competitive with the standard EA donation opportunities

This claim seems exactly what people felt was too hubristic—how could anyone be so confident on the basis of a quick survey of such a complex area that climate didn’t match up to other donation opportunities?

[CarlShulman]

I actually happen to think that the report was too dismissive of more leveraged climate change interventions that I expected could be a lot better than the estimates for Cool Earth (especially efficient angles on scientific research and political activity in the climate space), but the OP is suggesting that the original Cool Earth numbers (which indicate much lower cost-effectiveness than charities recommended by EAs in other areas with more robust data) were overstated, not understated (as the original report would suggest due to regression to the mean and measurement error).

[John G. Halstead]

I actually think the best climate charities are better than a lot of other things EAs donate to. From a long-termist pont of view, it looks better than global health, zoning in San Fran, macroeconomic policy, criminal justice reform, and animal welfare. I also think funding ordinary climate policy orgs is a better bet than funding solar geoengineering research, which Open Phil has done. Climate change of >6 degrees is very much on the cards, and this would do tremendous damage in the long-term.

Donations in other long-termist areas, like EA field building, AI, bio and nuclear security are probably better, but I think more could be going into the really outstanding climate organisations out there. I’m aware of at least four really good climate policy orgs whose funding gaps I would like to see filled.

[juliemack169]

Really very great answer.

[Kirsten]

“I think nobody delved into the Cool Earth numbers because it wasn’t worth their time, because climate change charities generally aren’t competitive with the standard EA donation opportunities”

I’ve heard this a lot on EA circles, but I’m not sure why. I tried to compare the Coalition for Rainforest Nations and the Against Malaria Foundation in terms of lives saved and could barely start. It’s just too uncertain and complex for me—I don’t know how to translate CO2 reductions into lives saved, although I’m certain that climate change will kill people. Has anyone even tried to do this kind of comparison? A link would be appreciated.

[John G. Halstead]

John Broome has also tried to create a conversion factor from DALYs to CO2. I don’t think any particular estimate is credible. Estimates of the social cost of carbon are for the most part completely made up, unmoored from information on impacts. It’s also very hard because I think probably most of the costs of climate change are very indirect and highly uncertain, stemming from the political risks of unprecedented mass migration

[Denkenberger🔸]

There are two ways for climate change reduction to be considered effective by EA frameworks: long-term future and saving lives/​improving utility in the presentish generation. There is some discussion here about long-term future. For saving lives, I agree it is tricky. When I attempted this in 2005, I tried to do it based on increased utility. Even though it is true that climate change will likely fall disproportionately on less-developed countries, when you look at the actual economic impacts, they accrue mostly to richer people because they make up the majority of the economy. This is especially true in the longer term, when it is likely that even current less-developed countries will be significantly richer than today. For typical cost climate interventions, I was getting they are about 2.5 orders of magnitude lower cost effectiveness than direct global poverty interventions. Another attempt is here (though you may not agree with his discounting). If Cool Earth really is significantly lower cost, of course that would improve the comparison. But I still think it is very unlikely to be better than direct global poverty interventions.

[CarlShulman]

One thing to emphasize more than that writeup did, is that in EA terms donating to such a lightly researched intervention (a few months work) is very likely dominated by donations to better research the area, finding higher expected value options and influencing others.

On the other hand, the point estimates in that report favored other charities like AMF over Cool Earth anyway, a conclusion strengthened by the OP critique (not that it excludes something else orders of magnitude better being found like unusual energy research, very effective political lobbying, geoengineering, etc; Open Philanthropy has made a few climate grants that look relatively leveraged).

And I agree with John Maxwell about it being oversold in some cases.

[Larks]

Thanks for writing this. This sort of evaluation, which has the potential to radically change the consensus view on a charity, seems significantly under-supplied in our community, even though individual instances are tractable for a lone individual to produce. It’s also obviously good timing at the start of the giving season.

I think the post would be improved without the section on contraception, however. There are many simple environmental interventions we could benchmark against instead, that don’t involve population ethics. Preventing a future human from being born has many impacts—they will probably have a job, some probability of inventing a new discovery, and most importantly they will probably be grateful to be alive—of which emitting some CO2 is likely to be one of the smaller impacts. Any evaluation of contraception that only looks at direct environmental impact is going to be so lacking that I suspect you’d be better off choosing a different intervention to compare to.

[Taymon]

Also, the cases for contraception and female education as climate-change interventions seem much, much more speculative than the case for rainforest conservation, so much so that their respective cost-effectiveness numbers probably ought not to be directly compared.

[Aaron Gertler 🔸]

I agree on a meta level that the section on contraception shouldn’t be here.

On the object level, this isn’t the first time I’ve seen a charity analysis make the assumption that a problem can be completely solved if you multiply “cost to solve X% of the problem” by 100/​X. This seems unlikely to be true in most cases, and is an important mistake to avoid.

For example, even if Guttmacher is correct that $9.4 billion will allow us to meet all unmet demand for contraceptives, contraceptives aren’t always used correctly, and even when used correctly, they don’t always work, so we won’t stop all 74 million unintended pregnancies. Also, not every unintended pregnancy leads to a child being born. And I suspect (from a brief skim of the report, I could be wrong here) that the $9.4 billion number involves some basic multiplication on Guttmacher’s part, without fully accounting for how difficult it is to reach different populations (in communities where contraceptives are forbidden, how much will it cost to change that policy?).

It’s still valuable to estimate and round, but the “multiply the number of people by the cost per person” fallacy (is there a catchier name?) seems like it routinely underestimates the cost of fully “fixing” a problem for reasons like these: Some % of your target population is hard to reach, and the thing that you think will always fix the problem only fixes it some % of the time.

[Sanjay]

Aaron, I think we agree that the contraception section is not the strongest part of the post. On your specific point though, I don’t think it’s actually so bad to use the $9.4bn divided by 74 million.

You suspect that $9.4bn is calculated by looking at data from a much smaller sample and performing a simple extrapolation. So do I. But that’s actually exactly what we want—this advice is intended for people donating thousands, not billions. (This point is what I was trying to get at with my note about more-expensive “last-mile” contraceptive needs).

Also remember I’m aiming for a rough estimate.

[Aaron Gertler 🔸]

Good points all. Yours was a reasonable estimate, but the topic made it a good way to discuss certain general problems with this type of estimate, which are often more prominent in estimates made by other people in other situations.

(Also, I didn’t spot the “last mile” comment; sorry for missing that, and thanks for calling it to my attention.)

[Sanjay]

Thanks Larks and Taymon. Your comments about the section on contraception is probably fair.

[Kirsten]

Cool Earth was the EA community’s default response to anyone who wanted to donate to climate change for years, without particularly good reason. Sanjay’s work overturned that recommendation and shortly after more rigorous recommendations were published.

[hegesippe]

I’m in fact shocked on many levels by the contraception section.

a. as you said you guestimated the average lifetime carbon impact. In doing so you underestimated or silenced the immense difference between countries and “classes”: basically your “target” population for contraception (super-sic) has a carbon footprint that is 100 times lower than, say people in the UK. Indeed, if you check there, https://​​en.wikipedia.org/​​wiki/​​List_of_countries_by_carbon_dioxide_emissions_per_capita and take the per capita emissions of UK and Papua New Guinea between 1980 and 2014, an average british person born in 1980 will have emitted approximately 300 tons of CO2, while an average Papuan will have emitted approximately 25 tons of CO2. This is “per country”, and thus doesn’t discriminate between richer and poorer classes ⇒ to put it simply, a person who takes the plane 10 times /​​ year will have a much higher impact that one who doesn’t. So even in Papua New guinea, the poorest and most marginalized—whom you want to target for contraception—will have a much lower impact than the national average. So basically, an “effective altruism” would more effective in highly emitting countries such as… the UK, the US, Canada… or Qatar, Kuwait...

b. obviously you see where I’m getting at: this section on contraception is ethically very dubious, as contraception is a matter of health and human rights, not a matter of strategically deciding which poor people would be better of not being born so that rich people can continue polluting but call it “effective altruism”.

[Taymon]

I suspect that it was widely recognized for quite some time that GWWC’s analysis of Cool Earth was outdated enough not to be trustworthy. People donated to Cool Earth anyway because it was the only climate-change charity that we had any particular reason to believe was better than others. This, of course, has changed with the Founders Pledge report, and as such I predict that EA interest in Cool Earth will fade with time.

I looked a little to try to figure out why the criticisms of Cool Earth don’t also apply to the Coalition for Rainforest Nations. It sounds like the primary reason is because CfRN influences nationwide policy, so the loggers can be displaced only to a different country, which is inconvenient enough that most would give up.

[John G. Halstead]

The Rainforest Coalition have advocated for measurement of deforestation prevention at the national level which means that you don’t get intra-national displacement. Since all rainforest countries have signed up for REDD (and it is enshrined in the Paris Agreement), you also don’t get international displacement.

Project-based approaches to deforestation as carried out by Cool Earth were rejected for years by the UN as verified CO2 reductions because of the problems outlined in the post, amongst others.

[Aaron Gertler 🔸]

Strong upvote to this post for being exactly the kind of research we need more people to do.

Also, I like that you contacted Cool Earth and gave them a chance to respond, though I wish you’d also asked them for permission to publish the correspondence or to at least quote them directly. In cases like these, I prefer seeing both sides discuss an issue in their own words.

[Sanjay]

Publishing the correspondence would be tricky because lots of it happened over phone calls. However it may be possible to get them to respond to this post, and then you can see their point of view from the horse’s mouth.

[RayTaylor]

I think a step back might help.

First we might decide that effective campaigning on climate might have more per dollar effectiveness in medium and long term than local action. Jim Hanssen himself, or WWF, or EIA, or Operation Noah, or COIN, or Plane Stupid might therefore be better causes even if they cant demonstrate short term carbon reduction.

Second, carbon is not the only or even the most powerful warming gas, so in the long term, we might think organisations like WWF and EIA (UK) working on super-powerful greehouse gases like HCFCs have done more than most of the 200 year time scale.

There are other climate forcings such as Black Carbon, reduction of water in landscapes (ref NASA’s GLACE study and UK’s AMMA 1 & 2), urban albedo and so on, any of which could be more tractable currently.

Mass behaviour change and local action has been singularly ineffective against climate chage in the face of massive industrial devlopment and government pro-growth policies, so clearly some action is needed at government level. And again, governments have been ineffective because of powerful lobbying and effective “the science is uncertain” campaigns by the coal lobby, using lobbyists like Fred Singer who previously worked for Big Tobacco.

COINnet.net is a small Oxford NGO that has done a lot of work on this, and George Marshall now works directly with big corporate type people to try to establsh a communication bridge.

It may be that our political and econmic systems are so structured that nothing truly effective can be done on climate until people in wealthy countries are much more impacted. In that case, research on recovery, restoration and geoengineering (ie things that can be implemented in 20 /​ 40 /​ 60 years time) might be more on-point, so that we are ready to act effectively when the will is really there?

That leaves us with a gap of time, in which there may be severe consequences. ALLFED.info has an approach to try to mitigate such consequeces, by recovering food systems as fast as possible. (Declaring an interest: I co-founded and work for ALLFED.)

[imoatama]

Fairly recent EA here (interested for ~2 years, took the Try Giving pledge a year ago, went to one EAGx earlier this year)

FWIW, I took the GWWC assessment to be both low confidence and out of date, but chose to support Cool Earth anyway despite considering it lower impact than eg health or poverty reduction because I wanted to contribute specifically in aid of climate change mitigation.

My decision process was basically to decide that I wanted climate change to be a focus area for some giving, then go and try and find the least worst charity from an EA perspective. But I still only ever thought of Cool Earth as a least worst sort of option.

I was already planning to cancel my regular donation before I read this (I bought a house and have paused all monthly donations, I will instead donate in a lump sum at EOFY), and will stick with that plan having read this.

[John G. Halstead]

Really good that someone took the time to write this

[Aaron Gertler 🔸]

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:

“Why we have over-rated Cool Earth” is a classic example of independent EA research. The author consults public data, runs his own statistical analyses, and reaches out to a charity with direct questions, bringing light to a subject on which the EA community doesn’t have much knowledge or experience. He also offers alternative suggestions to fight climate change, all while providing enough numbers that any reader could double-check his work with their own assumptions.

[Sanjay]

Commenters appear to be not unanimous about whether “We all knew that the Cool Earth recommendation was low-confidence.” I don’t have a dispositive answer on this, but my impression (based on anecdotes/​conversations with EAs) is that many people consider the Cool Earth conclusion definitive.

[Robert_Wiblin]

I’ll just add that I always read it as a highly tentative recommendation that you’d expect to be overturned if serious resources were put into investigating climate change charity—though I’m unusually in-the-know.

[Jamie_Harris]

There are also important considerations about the risk that rainforest preservation efforts might indirectly increase suffering.

Many in the effective altruism community believe that a large proportion of wild animals, especially invertebrates and other r-selected species, have net negative lives. Recently, this was the conclusion of a recent report by charity entrepreneurship. If you believe that there is a non-trivial chance that these animals can suffer or have morally relevant experiences, then the short- and medium-term effect of rainforest protection might be a counterfactual increase in wild animal suffering (see here for Brian Tomasik’s discussion of a related question).

More widely, encouraging concern for habitat protection might encourage people to value non-sentient entities even where the interests of these non-sentient entities conflict with the direct interests of individual animals. In general, this seems to be a step in the wrong direction if you agree that moral circle expansion is desirable. This might encourage the likelihood of future dystopian scenarios which involve astronomical levels of suffering.

In a sense, by promoting environmentalism via conservation, you might be reducing the chance of a global catastrophe via climat change but increasing the chance of S risk.

[Brian_Tomasik]

Good points. I mentioned Cool Earth specifically here, with a tentative calculation suggesting that even if greenhouse-gas emissions increase wild-animal populations (and it’s not clear that they do), preserving rainforest to sequester CO2 probably increases wild-animal populations even more.