Hide table of contents

Summary

  • I think decreasing greenhouse gas (GHG) emissions has benefits to humans of 0.00957 DALY/tCO2eq, of which:
    • 68.8 % are strictly linked to decreasing GHG emissions.
    • 31.2 % are linked to decreasing air pollution from fossil fuels.
  • I calculated the cost-effectiveness of:
  • I estimated the cost-effectiveness of CCF is:
    • 3.28 times that of TCF, with a plausible range of 0.175 to 30.2 times. So it is unclear to me whether donors interested in improving nearterm human welfare had better donate to GiveWell’s funds or CCF.
    • 0.217 % that of corporate campaigns for chicken welfare, with a plausible range of 0.0116 % to 2.00 %. Consequently, I recommend donors who value 1 unit of nearterm welfare the same regardless of whether it is experienced by humans or animals to donate to the best animal welfare interventions, such as the ones supported by the Animal Welfare Fund (AWF).
  • I concluded the harm caused to humans by the annual GHG emissions of a random person is 0.0660 DALY, and that caused to farmed animals by their annual food consumption is 4.04 DALY, i.e. 61.2 times as much. In my mind, this implies one should overwhelmingly focus on minimising animal suffering in the context of food consumption.

Calculations

I describe my calculations below. You are welcome to make a copy of this Sheet to use your own numbers.

Benefits to humans of decreasing greenhouse gas emissions

I think decreasing GHG emissions has benefits to humans of 0.00957 DALY/tCO2eq (= 0.00658 + 0.00299), adding:

  • 0.00658 DALY/tCO2eq strictly linked to decreasing GHG emissions, which comprises 68.8 % (= 0.00658/0.00957) of the total.
  • 0.00299 DALY/tCO2eq linked to decreasing air pollution from fossil fuels, which comprises 31.2 % (= 0.00299/0.00957) of the total.

I calculated a component strictly linked to decreasing GHG emissions of 0.00658 DALY/tCO2eq (= 0.0394*10^-3*167), multiplying:

  • A value of increasing economic growth of 0.0394 DALY/k$ (= 0.5/(12.7*10^3)). I computed this from the ratio between:
    • Open Philanthropy’s (OP’s) valuation of health of 0.5 DALYs per multiple of income (= 1/2).
    • The global gross domestic product (GDP) per capita in 2022 of 12.7 k$.
  • A social cost of carbon (SSC) in 2020 of 167 $/tCO2eq (= 17.1*1.14*2.76*2.06*1.51). I determined this from the product between:
    • A partial SSC in 2020 representing just the effects on mortality of 17.1 2019-$/tCO2eq, as obtained in Carleton 2022 for the representative concentration pathway 4.5 (RCP 4.5), and their preferred discount rate of 2 %[1] (see Table III).

      • Carleton 2022 got 36.6 2019-$/tCO2eq for RCP 8.5. Nevertheless, I considered the value for RCP 4.5 because this results in a global warming in 2100 of 2.5 to 3 ºC relative to the pre-industrial baseline, which is in agreement with Metaculus’ median community prediction on 11 March 2024 of 2.81 ºC relative to the 1951-1980 baseline. In contrast, RCP 8.5 leads to a global warming in 2100 of 5 ºC relative to the pre-industrial baseline.

      • Carleton 2022 says “a “full” SCC would encompass effects across all affected outcomes (and changes in mortality due to other features of climate change, like storms)”. However, I believe Carleton 2022’s estimates could be interpreted as encompassing all the impacts on mortality, as I guess additional deaths caused by GHG emissions through natural disasters are much smaller than those via changes in temperature. Based on data from The International Disaster Database (EM-DAT), the mean death rate from natural disasters decreased 97.6 % (= 1 - 0.64/26.5) from the 1920s to the 2010s[2], while temperature increased 1.01 ºC[3] (= 0.735 - (-0.274)). Additionally, the fraction of the damage of extreme weather events attributable to climate change does not appear to present a clear downwards/upwards trend.

    • 1.14 2022-$/2019-$, to adjust the partial SCC for inflation.
    • A factor of 2.76 (= 17.5*10^3/(6.35*10^3)) to account for climate change affecting more people with lower income. Carleton 2022 “allow[s] the VSL [value of a statistical life] to vary with income”, whereas I wanted 1 year of healthy life to always be worth the same (regardless of income). I obtained the factor from the ratio between:
      • 17.5 k 2017-$, which is the global real GDP per capita in 2022.
      • 6.35 k 2017-$ (= (19.1*1.41*10^3*18.2 + 10.1*333*64.6 + 62.7*1.42*10^3*7.11 + 376.0*236*5.38 + 248.5*171*6.26 - 4.7*447*46.0 + 131.8*1.21*10^3*3.77)*10^3/(19.1*1.41*10^3 + 10.1*333 + 62.7*1.42*10^3 + 376.0*236 + 248.5*171 - 4.7*447 + 131.8*1.21*10^3)), which is the real GDP per capita in 2022 of the countries mentioned in Table II of Carleton 2022 weighted by future deaths caused by GHG emissions in 2020[4]. I got the deaths by multiplying the death rates in Table II by the population of the countries in 2022.

    • A ratio between the full SSC and the partial SSC of 2.06 (= 185/90), as in Rennert 2022 (see Fig. 3).
      • From Fig. 3, impacts on agriculture account for 45.4 % (= 84/185) of the cost, on energy consumption 4.86 % (= 9/185), on mortality 48.6 % (= 90/185), and on sea-level rise 1.08 % (= 2/185).
      • “The mortality damage functions are based on the results of Cromar et al. 202219, in which a panel of health experts was convened to conduct a meta-analysis of peer-reviewed research studying the impacts of temperature on all-cause mortality risk [including conflicts and natural disasters?], which includes human health risks related to a broad set of health outcomes including cardiovascular, respiratory and infectious disease categories. The meta-analysis combined studies to produce regionally disaggregated estimates of the effects on all-cause mortality of each degree of warming across a broad range of baseline temperatures, including both increased mortality risk at high temperatures and reduced risk at cooler temperatures”.
      • I did not directly use Rennert 2022’s full SSC of 185 $/tCO2eq because I agree with David Friedman’s critique of this study. Crucially, it assumes the same relationship between income and mortality until 2300 despite significant increases in income, thus underestimating mitigation (e.g. via greater ability to afford air conditioning). For context, according to data from the Global Burden of Disease study (GBD), the disease burden per capita from non-optimal temperature decreased 35.7 % (= 1 - 486/756) from 1990 to 2019. I guess part of this decrease is explained by an increase in mitigation measures enabled by higher incomes.
    • A factor of 1.51 (= 2.54/1.68) to integrate the effect of morbidity[5]. I obtained this from the ratio between:

      • 2.54 billion DALYs in 2019.

      • 1.68 billion (= (2.54 - 0.861)*10^9) years of life lost (YLL) in 2019, which I got from the difference between the above and 0.861 billion years lived with disability (YLD) in 2019.

  • A cost-effectiveness of decreasing GHG emissions of 3.41 tCO2eq/$, with a plausible range of 0.182 to 31.4 tCO2eq/$.
    • These were obtained in the context of Founders Pledge’s 2018 climate report (see section 3.2), and respected the future work of Clean Air Task Force (CATF), which is the organisation that has received the most money from CCF.
    • Johannes Ackva, who is the manager of CCF, “thinks this [3.41 tCO2eq/$] is in the right ballpark and not worth getting more precise for an analysis like this (where most parameters are much more uncertain)”.

I estimated a component linked to decreasing air pollution from fossil fuels of 0.00299 DALY/tCO2eq (= 164*10^6/(54.8*10^9)), dividing:

  • The disease burden connected to air pollution from fossil fuels in 2019 of 164 MDALY (= 5.13*10^6*31.9). I obtained this multiplying:
    • 5.13 M deaths caused by air pollution from fossil fuels in 2019, in line with Lelieveld 2023.
    • 31.9 DALY/death (= 213*10^6/(6.67*10^6)), which I got from the ratio between the disease burden and deaths from air pollution in 2019 of 213 MDALY and 6.67 M.
  • The GHG emissions in 2019 of 54.8 GtCO2eq.

Cost-effectiveness of the Climate Change Fund

I calculated the cost-effectiveness of CCF is 0.0326 DALY/$ (= 0.00957*3.41), with a plausible range of 0.00174 (= 0.00957*0.182) to 0.300 DALY/$ (= 0.00957*31.4), multiplying:

  • The benefits of decreasing GHG emissions of 0.00957 DALY/tCO2eq (see previous section).
  • A cost-effectiveness of decreasing GHG emissions of 3.41 tCO2eq/$, with a plausible range of 0.182 to 31.4 tCO2eq/$.
    • These were obtained in the context of Founders Pledge’s 2018 climate report (see section 3.2), and respected the future work of Clean Air Task Force (CATF), which is the organisation that has received the most money from CCF.
    • Johannes Ackva, who is the manager of CCF, “thinks this [3.41 tCO2eq/$] is in the right ballpark and not worth getting more precise for an analysis like this (where most parameters are much more uncertain)”.

Cost-effectiveness of the Top Charities Fund

I concluded the cost-effectiveness of TCF is 0.00994 DALY/$ (= 51*0.195*10^-3), multiplying:

  • GiveWell’s implicit valuation of saving lives of 51 DALY/life. According to OP, “GiveWell uses moral weights for child deaths that would be consistent with assuming 51 years of foregone life in the DALY framework (though that is not how they reach the conclusion)”.
  • The mean reciprocal of the cost to save a life of GiveWell’s 4 top charities of 0.195 life/k$ (= (1/5 + 1/5.5 + 1/5 + 1/5)*10^-3/4).

Cost-effectiveness of corporate campaigns for chicken welfare

I arrived at a cost-effectiveness of corporate campaigns for chicken welfare of 15.0 DALY/$ (= 8.20*2.10*0.870), assuming:

  • Campaigns affect 8.20 chicken-years per $ (= 41*1/5), multiplying:
    • Saulius Šimčikas’ estimate of 41 chicken-years per $.
    • An adjustment factor of 1/5, since OP thinks “the marginal FAW [farmed animal welfare] funding opportunity is ~1/5th as cost-effective as the average from Saulius’ analysis [which is linked just above]”.
  • An improvement in chicken welfare per time of 2.10 times the intensity of the mean human experience, as I estimated for moving broilers from a conventional to a reformed scenario based on Rethink Priorities’ median welfare range for chickens of 0.332[6].

  • A ratio between humans’ healthy and total life expectancy at birth in 2016 of 87.0 % (= 63.1/72.5).

In light of the above, corporate campaigns for chicken welfare are 1.51 k (= 15.0/0.00994) times as cost-effective as TCF.

Discussion

Cost-effectiveness

I infer the cost-effectiveness of CCF is 3.28 (= 0.0326/0.00994) times that of TCF, with a plausible range of 0.175 (= 0.00174/0.00994) to 30.2 (= 0.300/0.00994) times. The actual plausible range is wider because I only modelled the uncertainty in the cost-effectiveness of CCF in tCO2eq/$. So it is unclear to me whether donors interested in improving nearterm human welfare had better donate to GiveWell’s funds or CCF[7]. At the same time, I do not recommend Founders Pledge’s Global Health and Development Fund nor EA Funds’ Global Health and Development Fund.

Note CCF is mostly trying to minimise the damage strictly linked to GHG emissions, but interventions explicitly targeting decreasing deaths from air pollution, like those of OP’s area of South Asian air quality, may be more cost-effective. I established 31.2 % of the benefits of decreasing GHG emissions are linked to decreasing air pollution from fossil fuels, but the best interventions to mitigate the damage strictly linked to GHG emissions being the best to mitigate air pollution would be a little bit of a surprising and suspicious convergence. For reference, I was also expecting a greater fraction of the benefits to come from decreasing air pollution.

In addition, I conclude the cost-effectiveness of CCF is only 0.217 % (= 0.0326/15.0) that of corporate campaigns for chicken welfare, with a plausible range of 0.0116 % (= 0.00174/15.0) to 2.00 % (= 0.300/15.0). The actual plausible range is wider for the reasons I mentioned above, and the difference may be attenuated owing to indirect effects[8], but I believe corporate campaigns will still come out on top given reasonable assumptions[9]. Consequently, I recommend donors who value 1 unit of nearterm welfare the same regardless of whether it is experienced by humans or animals to donate to the best animal welfare interventions, such as the ones supported by AWF. As a side note, I currently also prefer these over ones explicitly targeting existential risk mitigation, like those supported by the Long-Term Future Fund[10] (LTFF).

It can still make sense for Founders Pledge to recommend CCF to donors partial to climate change:

  • Even if Founders Pledge’s researchers are mostly cause neutral, many people donating to CCF may be partial to climate change.
  • The counterfactual donations of CCF’s major donors might go to less effective organisations in its absence, and supporting CCF might be a pathway towards moving to more pressing areas.

Nonetheless, I would say cause neutral organisations offering philanthropic advice would do well to make a serious effort to influence the areas their donors support, which may well be the greatest driver of impact (relatedly). Furthermore, I encourage such organisations to invest more resources into cause prioritisation, in order to better understand how to compare climate change, global health and development, and animal welfare interventions.

Carbon and farmed animal welfare footprint

For the global GHG emissions per capita in 2021 of 6.9 tCO2eq, my estimate for the benefits of decreasing GHG emissions of 0.00957 DALY/tCO2eq implies the annual personal emissions of a random person cause 0.0660 DALY (= 6.9*0.00957). Equivalently, given the healthy life expectancy in 2019 of 63.7 years, the annual GHG emissions of 965 (= 63.7/0.0660) random people cause an harm comparable to 1 human death.

estimated the scale of the annual suffering of all farmed animals is 4.64 times the scale of the annual happiness of all humans. As a consequence, for a ratio of 87.0 % between human’ healthy and total life expectancy at birth (see previous section), I think the harm caused to farmed animals by the annual food consumption of a random person is 4.04 DALY (= 4.64*0.870). Equivalently, the annual food consumption of 15.8 (= 63.7/4.04) random people causes an harm comparable to 1 human death.

Based on the above, the harm caused to farmed animals by the annual food consumption of a random person is 61.2 (= 4.04/0.0660) times that caused to humans by their annual GHG emissions. In my mind, this implies one should overwhelmingly focus on minimising animal suffering in the context of food consumption.

One could neutralise:

  • The harm caused to humans by the annual GHG emissions of a random person donating just 2.02 $ (= 6.9/3.41) to CCF, or 0.00440 $ (= 0.0660/15.0) to THL, which has been playing a major role in corporate campaigns for chicken welfare.
  • The harm caused to farmed animals by the annual food consumption of a random person donating just 0.269 $ (= 4.04/15.0) to THL.

I see these as nice opportunities to donate more instead of just offsetting the impact of one’s emissions.

Wild animals

I estimated the scale of the welfare of wild animals is 10.9 M times that of farmed animals. Nonetheless, I have neglected the impact of GHG emissions on wild animals due to their very low resilience. According to Brian Tomasik:

On balance, I’m extremely uncertain about the net impact of climate change on wild-animal suffering; my probabilities are basically 50% net good vs. 50% net bad when just considering animal suffering on Earth in the next few centuries (ignoring side effects on humanity's very long-term future).

In particular, it is unclear whether wild animals have positive/negative welfare.

Acknowledgements

Thanks to Anonymous Person, Johannes Ackva, Matthew Rendall and Rafael Latham-Proença for feedback on the draft[11].

  1. ^

     “Our preferred estimates use a discount rate of ”. This is 1.08 (= 0.02/0.0185) times the 1.85 % (= (17.5/9.72)^(1/(2022 - 1990)) - 1) annual growth rate of the global real GDP per capita from 1990 to 2022. The adequate growth rate may be higher due to transformative AI, or lower owing to stagnation. I did not want to go into these considerations, so I just used Carleton 2022’s mainstream value.

  2. ^

     TV adoption was pretty low during the early 20th century, which could explain the greater salience of natural disasters today.

  3. ^

     0.735 ºC (= (0.68 + 0.54 + 0.58 + 0.62 + 0.67 + 0.83 + 0.93 + 0.85 + 0.76 + 0.89)/10) is the mean temperature anomaly in the 2010s relative to the mean one during period from 1961 to 1990, and -0.274 ºC (= (-0.30 - 0.24 - 0.34 - 0.32 - 0.31 - 0.28 - 0.12 - 0.23 - 0.21 - 0.39)/10) is that in the 1920s relative to the same period.

  4. ^

     I set the real GDP per capita of Europe to that of the European Union (EU), which is the one provided by the World Bank. This assumption has little importance because Europe’s weight is only -0.515 % (= -4.7*447/(19.1*1.41*10^3 + 10.1*333 + 62.7*1.42*10^3 + 376.0*236 + 248.5*171 - 4.7*447 + 131.8*1.21*10^3)).

  5. ^

     Carlson 2023 called for quantifying the current disease burden of climate change. Cromar 2022 says the following about morbidity and mortality:

    Morbidity costs are generally lower [in agreement with the aforementioned factor being lower than 2] and more difficult to estimate than mortality costs due to the lack of data available at the country level (with the notable exception of birth outcomes). A single endpoint, such as health care expenditures, may be a useful metric in this regard in the short term for valuing morbidity outcomes that are difficult to estimate. Mortality impacts will continue to be of greater importance in economic modeling.

  6. ^

     Ideally, I would use a number encompassing both broiler and chicken-free campaigns.

  7. ^

     GiveWell has 3 funds, TCF, All Grants Fund and Unrestricted Fund, which should ideally have the same marginal cost-effectiveness.

  8. ^

     The CCF has not funded interventions aiming to decrease GHG emissions from animal agriculture, which could have significant effects on animal welfare.

  9. ^

     For example, using best guess welfare ranges 0.316 (= 10^-0.5) to 3.16 (= 10^0.5) times those from Rethink Priorities.

  10. ^

     For donors interested in interventions explicitly targeting existential risk mitigation, I recommend donating to LTFF, which mainly supports AI safety. I guess existential risk from climate change is smaller than that from nuclear war (relatedly), and estimated the nearterm annual risk of human extinction from nuclear war is 5.93*10^-12, whereas I guess that from AI is 10^-6.

  11. ^

     The names are ordered alphabetically.

Comments13
Sorted by Click to highlight new comments since:

"A cost-effectiveness of decreasing GHG emissions of 3.41 tCO2eq/$, with a plausible range of 0.182 to 31.4 tCO2eq/$."

This is not a credible number, and Founders Pledge as of several years ago said they no longer stand behind the cost-effectiveness calculation you link to in your post.

It is based on an assumption that CATF nuclear advocacy will result in cheap enough reactors to replace coal in thermal electric power production. That is not credible now, and it wasn't at the time when the BOTEC was made. Note the 0.5%/1%/2% assumptions that nuclear will displace coal that are doing quite a lot of heavy lifting in getting the numbers to work out. The percentages are far lower than that. Be careful of arbitrary bounding your analysis in whole numbers between 1-100%. I made a copy of the sheet when it came out to capture any changes or alterations - my copy has some cells labeled that are missing in yours.

The supposed climate benefits of nuclear advocacy are contested, and far more credible and sophisticated modeling shows the possibility that there are some zero-sum trade-offs in scaling that mean more nuclear power could result in higher cumulative emissions. I see it as even odds whether CATF nuclear advocacy increases or reduces emissions in expectation. But extremely likely (>95%) that CATF's nuclear program was a total waste of philanthropic dollars.

The lesson is to be careful with BOTECs - use probability distributions instead of single-point numbers, and have several people red-team the analysis both in the numbers and the structure of the calculation.

Another issue is you are taking values derived from speculation and comparing them to measured cost-effectiveness from RCTs with a strong evidentiary basis.

While I think it is a mistake to motivate this estimate with a 2017 BOTEC (here we agree!), it is also mistaken to claim that such a range – spanning more than two OOMs and high and fairly low cost-effectiveness – is implausible as a quite uncertain best guess.

As discussed many times, CCF grantmaking does not rely on 2017 BOTECs and neither does my best guess on cost-effectiveness (Vasco operationalized it one specific way I am not going to defend here, I am just defending a view that expected cost-effectiveness is roughly in the 0.1 USD/t to to 10 USD/t range).

Why an estimate in this range seems plausible

This seems plausible for many reasons, none of which depending on the specific BOTEC:

I. Outside-view multiplier reasoning

  • (1) It is clearly possible to reduce tons of CO2eq for USD 100/t through direct and high-certainty action.
  • (2) If you only assumed a conventional advocacy multiplier – of the form that many EA orgs assume when modeling policy work (e.g. OP) and that is well-substantiated by empirical political science research and many studies on successes in philanthropy – you would assume a 10x multiplier.
  • (3) You now “only” need another 10x multiplier to get to USD 1/t and there seem many plausible mechanisms to get there – e.g. focusing on actions with transformative potential such as innovation, avoiding carbon lock-in etc. or – more meta – driving in additional funding from other donors / foundations when supporting early-stage organizations.
  • (4) Obviously, one also needs to discount for things like funding additionality, execution risk. Etc.
  • (5) This will result in a very uncertain range, but it is well-approximated by what Vasco has chosen to model this.

Note that these are overall quite weak assumptions and, crucially, if you do not buy them you should probably also not buy the cost-effectiveness analyses on corporate campaigns for chicken welfare.


II. Observations of grants and inside-view modeling

  • (1) While I generally put less stock in them than in comparative analysis, we also do more inside-view cost-effectiveness analyses that often have a range close to USD 0.1-USD 10t/CO2e.
  • (2) While the CCF does not exist long enough to be confident in long-run emissions outcomes – we generally invest in theories of change that need time – there is a lot of reason to expect that some of those bets will pay off at the very high cost-effectiveness:
    • (a) Many of the charities supported – such as TerraPraxis, Future Cleantech Architects etc. – have crowded in multiples of the funding we allocated to them often as a direct result of our recommendation and/or organizational development we enabled with early grants.
    • (b) While hard to disentangle, they also play key roles in many policy changes – e.g. Carbon180 was a leading advocacy org on carbon removal in the IRA/IIJA window, two of our grantees are pushing a conversation on repowering with advanced heat sources (nuclear or geothermal) and one of our grantees (FCA) had several policy wins in the EU (not all they can talk about). 
    • (c) More nascent work is focused electricity market liberalization to advance renewables in emerging economies (Energy for Growth), a stronger climate civil society on the right (DEPLOY/US), as well as geothermal innovation in Canada (Cascade).
    • (d) This is a diversified sets of bets that leverages different mechanisms, with the uniting theme of leveraging advocacy, the focus on actions / spaces that are neglected, that have the potential to change trajectories, and that have a risk-reducing quality (hedging).  


III. Learning from other areas of philanthropy

Most areas of philanthropy seem structured such that, when being alright with risk neutrality and leveraged theories of change, one can get significant multiplier.

For example, I am quite confident that the implied multiplier for the case of chicken welfare campaigns compared to direct action is likely similarly large for what we are assuming for the case of Climate Fund. I also do not think any nuclear risk grant-maker would find it implausible that they could reduce nuclear risk 100x more cost-effectively (in expectation) than whatever the direct action equivalent would be. Or a global health grant-maker that would expect that their grants are 100x more cost-effective by influencing advocacy to have government invest in vaccine RD&D rather than buying equipment for their local hospital.


Bottom line: This cost-effectiveness range as a risk-neutral best guess does not depend on a 2017 BOTEC, but rather can be motivated via different streams of reasoning and evidence.

(I also think the critique of the 2017 BOTEC is way over-confident but this would be a separate comment)
 

Thanks for clarifying, Johannes. Strongly upvoted.

Note that these are overall quite weak assumptions and, crucially, if you do not buy them you should probably also not buy the cost-effectiveness analyses on corporate campaigns for chicken welfare.

I think estimates of the chicken-years affected per $ spent in corporate campaigns for chicken welfare may be more resilient than ones of the cost-effectiveness of CCF in t/$. According to The Humane League:

  • Nearly 2,500 corporate cage-free welfare policies, which protect hens from the most intensive forms of confinement, have been secured around the world.
  • 1,157 companies have successfully transitioned away from using battery cages, resulting in meaningful change for millions of animals.
  • 89% of all corporate cage-free commitments with a 2022 deadline (meaning companies promised to completely phase out battery cages by 2022 or earlier) have been fulfilled.
  • Just 4% of companies that pledged to be cage-free by 2022 are not yet fully cage-free (though they are publicly reporting on their progress), while only 7% are not yet publicly reporting on the status of their transition.
  • We’ve seen a complete transition to cage-free systems across industry sectors: 509 restaurants, 269 manufacturers, 174 retailers, 118 foodservice and caterers, 95 hospitality companies, 53 producers, and 19 distributors

As a side note:

  • Saulius estimated campaigns for broiler welfare are 27.8 % (= 15/54) as cost-effective as the cage-free campaigns concerning the above.
  • OP thinks “the marginal FAW [farmed animal welfare] funding opportunity is ~1/5th as cost-effective as the average from Saulius’ analysis”.
  • However, I accounted for both of these effects in my analysis.

Thanks, this updates me, I had cached something more skeptical on chicken welfare campaigns.

Do you have a sense of what "advocacy multiplier" this implies? Is this >1000x of helping animals directly?

I have the suspicion that the relative results between causes are -- to a significant degree -- not driven by cause-differences but by comfort with risk and the kind of multipliers that are expected to be feasible.

FWIW, I also do believe that marginal donations to help farmed animals will do more good than marginal climate donations.

Thanks for the follow-up! It prompted me to think about relevant topics.

Do you have a sense of what "advocacy multiplier" this implies? Is this >1000x of helping animals directly?

By helping animals directly, are you talking about rescuing animals from factory-farms, and then supporting them in animal sanctuaries? I am not aware of cost-effectiveness analyses of these, but here is a quick estimate. I speculate it would take 2 h to save one broiler. In this case, for 20 $/h, the cost to save a broiler would be 40 $ (= 2*20). Broilers in a conventional scenario live for 42 days, so saving one at a random time would in expectation avoid 21 days (= 42/2) of life in a conventional scenario, and contribute to perhaps 7.5 years of life in a sanctuary[1], such that the remaining life expectancy would become 130 (= 7.5*365.25/21) times as long. Based on my assumptions here, and supposing the welfare of a broiler in a sanctuary as a fraction of chickens' welfare range is similar to the welfare of a typical human as a fraction of humans' welfare range, I estimate going from a conventional scenario to a sanctuary is 22.8 (= (3.33*10^-6*130 + 2.59*10^-5)/(-5.77*10^-6 + 2.59*10^-5)) times as good as going from a conventional scenario to a reformed scenario. So the rescue would have a benefit equivalent to changing 479 days (= 21*22.8) of a broiler in a conventional scenario to one in a reformed scenario. Assuming the cost of maintaining the broiler in the sanctuary is much smaller than the cost of the rescue, which may be optimistic, the cost-effectiveness would be 0.0328 chicken-years/$ (= 479/365.25/40). If so, corporate campaigns for chicken welfare would be 250 (= 8.20/0.0328) times as cost-effective.

Relatedly, I Fermi estimated corporate campaigns for chicken welfare are 22.5 times as cost-effective as School Plates, which is a program aiming to increase the number of plant-based meals at schools and universities in the United Kingdom that is seemingly regarded as successful in advancing their intervention. It makes sense to me its cost-effectiveness is higher than that I estimated for rescuing broilers, but lower than that of corporate campaigns for chicken welfare. A direct rescue operations targets a single animal, School Plates presumably targets a university or schools in a given small region, and corporate campaigns target companies, which intuitively affect even more animals than the latter.

The 2 shallow analyses above seem qualitatively in agreement with what Founders Pledge's approach of focussing on impact multipliers would predict. On the other hand, I would be nice to have more monitoring and evaluation of animal welfare interventions to calibrate heuristics.

I have the suspicion that the relative results between causes are -- to a significant degree -- not driven by cause-differences but by comfort with risk and the kind of multipliers that are expected to be feasible.

I suspect Ben Todd's analysis underestimates variations in cost-effectiveness within causes, at least if one excludes indirect effects[2]. At the same time, it still seems like animal welfare interventions are generally more cost-effective than climate and global health and development ones. If corporate campaigns really are in the ballpark of 1.44 k times as cost-effective as GiveWell's top charities as I estimated, and Open Philanthropy's human welfare grants in their Global Health and Wellbeing portfolio, which supposedly takes advantage of multipliers, are 2 times as cost-effective as GiveWell's top charities, then corporate campaigns for chicken welfare are around 720 (= 1.44*10^3/2) times as cost-effective as such grants.

FWIW, I also do believe that marginal donations to help farmed animals will do more good than marginal climate donations.

Thanks for sharing!

  1. ^

    According to Goodheart Animal Sanctuaries:

    Most chickens have a natural lifespan of around 10 years but same may live up to 15 years! Hens rescued from industry sadly may not live this long, even when kept as pets, due to their stressful start in life.

  2. ^

    As an example of an indirect effect, rescues of farmed animals can be filmed, and the videos used to pressure companies to sign and fulfill their animal welfare pledges.

My estimation of the cost-effectiveness of rescuing a broiler had an error which I have now corrected. I was assuming the rescue would not change the lifespan of the broiler, but it would live longer in a sanctuary. Assuming 7.5 years (see 1st footnote above), corporate campaigns are just hundreds of times as cost-effective as the rescue (instead of thousands).

Thanks for the comment, Matthew.

This [3.41 tCO2eq/$] is not a credible number, and Founders Pledge as of several years ago said they no longer stand behind the cost-effectiveness calculation you link to in your post.

What is your best guess for the expected marginal cost-effectiveness of CCF? For the purpose of this analysis, it does not matter much whether it is 10 % or 10 times that I assumed, because I think in this case the qualitative conclusions would be the same.

You may be right that Founders Pledge no longer stands behind the cost-effectiveness analysis (it would be helpful if you could link to where they say that). However:

Johannes Ackva, who is the manager of CCF, “thinks this [3.41 tCO2eq/$] is in the right ballpark and not worth getting more precise for an analysis like this (where most parameters are much more uncertain)”.


Note the 0.5%/1%/2% assumptions that nuclear will displace coal that are doing quite a lot of heavy lifting in getting the numbers to work out. The percentages are far lower than that.

Interesting. In that case, GiveWell's interventions would be better than CCF, and corporate campaigns for chicken welfare would be many orders of magnitude more cost-effective than CCF.

Be careful of arbitrary bounding your analysis in whole numbers between 1-100%.

Great point! I do think this is a major source of error in cost-effectiveness analysis, and quantitive analyses more broadly:

In general, I suspect there is a tendency to give probabilities between 1 % and 99 % for events whose mechanics we do not understand well [unlike e.g. lotteries], given this range encompasses the vast majority (98 %) of the available linear space (from 0 to 1), and events in everyday life one cares about are not that extreme. However, the available logarithmic space is infinitely vast, so there is margin for such guesses to be major overestimates.


The supposed climate benefits of nuclear advocacy are contested, and far more credible and sophisticated modeling shows the possibility that there are some zero-sum trade-offs in scaling that mean more nuclear power could result in higher cumulative emissions.

Note the goal is decreasing cumulative deaths rather than GHG emissions. I think there is no difference if one conditions on a given emissions trajectory, but the goals may come apart accounting for uncertainty in the emissions trajectory. It is more valuable to decrease emissions in scenarios where emissions and temperature are higher, and nuclear power may be specially valuable here if those are scenarios where renewables did not scale as much as is currently anticipated.

The lesson is to be careful with BOTECs - use probability distributions instead of single-point numbers, and have several people red-team the analysis both in the numbers and the structure of the calculation.

I agree all of these are useful. Just one note. I often use probability distributions in my analyses, but at the end of the day one has to compare interventions by boiling down their cost-effectiveness distributions to a single number corresponding to the expected cost-effectiveness. If one decides to fund A over B, one is implictly saying the expected cost-effectiveness of A is higher than or equal to that of B.

Another issue is you are taking values derived from speculation and comparing them to measured cost-effectiveness from RCTs with a strong evidentiary basis.

This is one reason I do not conclude CCF is better than TCF even though I estimated the cost-effectiveness of CCF is 3.28 times as high. However, I think one can robustly conclude that corporate campaigns for chicken welfare are more cost-effective than TCF because I calculated their cost-effectiveness is 1.44 k times as high, which is a lot (similar to the ratio between the cost-effectiveness of TCF and unconditional cash transfer in high income countries, as TCF is 10 times as cost-effective as unconditional cash transfers to people in extreme poverty, and people in high income countries earn around 100 times as much as people in extreme poverty, which leads to a multiplier of around 1 k).

Love a good cost-effectiveness calculation.

Has anyone done a calculation of the (wild) animal welfare effects of climate change? Or is this so ungodly intractable that no one has dared attempt it.

Love a good cost-effectiveness calculation.

Thanks! Me too.

Has anyone done a calculation of the (wild) animal welfare effects of climate change? Or is this so ungodly intractable that no one has dared attempt it.

I am not aware of estimates of the impact of climate change on wild animal welfare. However, Brian Tomasik discussed many relevant factors qualitatively, and summarised his positition as follows. "On balance, I'm extremely uncertain about the net impact of climate change on wild-animal suffering; my probabilities are basically 50% net good vs. 50% net bad when just considering animal suffering on Earth in the next few centuries (ignoring side effects on humanity's very long-term future)".

Brian also has many estimates of the years of wild animal lives affected by changing land use (e.g. a change from rainforest to crops). These can be converted to human years using welfare ranges, but then there is the super hard question about what is the level of welfare of wild animals relative to their welfare range. I have calculated the impact of saving lives on wild animal welfare assuming the lives of wild insects are as intense as those of broilers relative to their respective welfare ranges. "All in all, I can see the impact on wild animals being anything from negligible to all that matters in the nearterm". Despite this, I believe there is lots of uncertainty about whether wild animal welfare is positive or negative, so I did not include impacts on wild animals in my post.

In any case, if one thinks the impacts of climate change on humans may well be dominated by those on wild animals, interventions to help these will look better than ones to decrease GHG emissions. Likewise, if one thinks the impacts of saving human lives may be dominated by impacts on farmed animals, interventions to help these will look better than ones to decrease GHG emissions. So I believe interventions to help animals are better than ones to decrease GHG emissions under any worldview.

I admire and encourage your mathematical diligence.

What I'm worried about is the error bars - multiplying errors can cause wild differences between the estimated and actual numbers. If the error bars of the two funds (CCF and TCF) overlap significantly, it might be too soon to judge which one is best.

Thanks, Joris, and welcome to the EA Forum!

What I'm worried about is the error bars - multiplying errors can cause wild differences between the estimated and actual numbers. If the error bars of the two funds (CCF and TCF) overlap significantly, it might be too soon to judge which one is best.

Agreed:

  • I estimated the cost-effectiveness of CCF is:
    • 3.28 times that of TCF, with a plausible range of 0.175 to 30.2 times. So it is unclear to me whether donors interested in improving nearterm human welfare had better donate to GiveWell’s funds or CCF.

Thanks, Pat. Readers interested in air pollution may want to check Founders Pledge's report on it. They do not have cost-effectiveness estimates, but I guess the best marginal interventions to decrease air pollution are less than 10 times as cost-effective as GiveWell's top charities, whereas I estimated corporate campaigns for chicken welfare are 1.44 k times as cost-effective as those. So I think it is quite safe to say that the best marginal animal welfare interventions are more cost-effective than the best marginal air pollution interventions.

Curated and popular this week
Relevant opportunities