Main takeaways

Funding research and development of clean technologies, in particular clean protein and clean energy, is probably the most cost-effective method to avoid dangerous climate change. Clean protein is animal-free protein such as cell-based meat and fermentation-based dairy and eggs that can replace their animal-based counterparts. Clean energy includes carbon neutral energy sources that can replace fossil fuels and do not emit greenhouse gases or toxic chemicals, as well as enabling energy infrastructure (e.g. grids, batteries, carbon capture and storage facilities).  

The three factor framework for focus area selection, often used in effective altruism, is useful for back-of-the-envelope cost-effectiveness estimates. According to this framework, the effectiveness is the product of three factors: scale (importance), solvability (tractability) and neglectedness. The three factor framework can be applied in two different ways, resulting in two cost-effectiveness estimates. The results are pretty robust in the sense that for each technology, both cost-effectiveness estimates have the same order of magnitude. 

According to very conservative assumptions (likely underestimating the cost-effectiveness), clean protein R&D-funding has a lower-bound cost-effectiveness of roughly 1 ton CO2 avoided per dollar; whereas clean energy has a lower-bound cost-effectiveness of roughly 0,1 ton CO2 avoided per dollar. The latter is also the (very conservative) estimate by a 2020 Founders Pledge report. Hence, in the short run, clean protein is likely more effective than clean energy R&D. However, in the long run, clean energy becomes crucial: without it, climate targets cannot be met. Clean protein is not sufficient to avoid dangerous climate change. As the development of clean protein that can replace animal protein is probably easier than the development of clean energy that can replace all fossil fuel energy, we can also expect that clean energy R&D has more room for more funding than clean protein R&D, which means that its cost-effectiveness is likely to remain very high for a longer time than clean protein R&D.

Clean protein and clean energy innovations are more cost-effective than e.g. Gold Standard carbon offsetting projects (around 0,05 ton CO2 per dollar), buying emission permits (less than 0,05 ton CO2 per euro for European ETFs), planting trees (0,03 ton CO2 per euro in Belgium), or paying the social cost of carbon to the poorest people who are most affected by climate change (a meta-analysis estimate of 0,02 ton CO2 per dollar, a refined estimate of 0,002 ton CO2 per dollar). The clean tech cost-effectiveness estimates are in the same order of magnitude as other carbon reduction methods that are considered to be very cost effective, such as forest protection (preventing deforestation by paying local communities for ecosystem services, which has a cost-effectiveness around 0,4 ton CO2 avoided per dollar) and reducing human population growth (preventing unwanted pregnancies by voluntary family planning, with a cost-effectiveness around 0,14 ton CO2 avoided per dollar). However, these other cost-effective carbon reduction methods have limited room for more funding, which means as carbon offsetting methods they cannot be universalized: not everyone can offset their own carbon emissions with such offsetting methods. When more people fund such carbon offsetting methods, their cost-effectiveness will strongly increase. There are not enough forests to protect and unwanted pregnancies to avoid. Avoiding carbon emissions from burning forests is not enough to meet climate targets. Reducing the human population is not enough to move to zero carbon emissions, unless the population goes to zero. Only with new, carbon-neutral technologies we can move all the way to zero carbon emissions. 

Clean tech innovation has a lot of room for more funding, is likely to stay very cost-effective, and has other important co-benefits. Clean protein reduces farm animal suffering, environmental impacts (e.g. water pollution, biodiversity loss) and public health risks (from antibiotic resistant bacteria and new zoonotic diseases), whereas clean energy reduces public health risks (e.g. air pollution from burning coal).  

Directly funding clean protein and clean energy R&D is possible by donating to e.g. New Harvest and the MIT Energy Initiative. But cost-effectiveness could increase further by leveraging funding, for example by advocating for more government funding of clean tech innovation, offering strategic support to clean tech companies or facilitating clean tech in other ways. Supporting clean protein advocacy is possible by donating to the Good Food Institute. Clean energy advocacy can be supported by donating to Let’s Fund (which supports the Clean Energy Innovation Program at the Information Technology and Innovation Foundation) and the Founders Pledge Climate Change Fund (which supports the Clean Air Task Force, Carbon180 and TerraPraxis as of 2020).

Calculation methods

Two back-of-the-envelope cost-effectiveness estimates are possible: the acceleration and the contribution estimate. 

The acceleration estimate assumes that the new technology will be developed anyway and will enter the market sooner or later, but an extra unit of R&D-funding accelerates the research such that we have the new technology sooner, such that atmospheric greenhouse gases can be reduced or avoided sooner. If the old technology emits X ton of CO2 each day, and one dollar funding accelerates the new, clean technology with Y days, the cost-effectiveness of R&D-funding equals X times Y ton CO2 per dollar. Once the new technology enters the market at competitive prices, consumers instead of philanthropic funders will pay for the production costs as well as the marketing and advertisement of the new technology.

The contribution estimate assumes that the new technology requires a fixed amount of R&D-funding to enter the market and has a fixed total greenhouse gas reduction potential, but a unit of funding has a fixed share of the total funding requirement, such that it offers a certain contribution to the total greenhouse gas reduction. If one dollar R&D-funding has a share S of the total funding requirements, and the total greenhouse gas reduction potential of the clean technology is T ton of CO2, the cost-effectiveness of R&D-funding equals S times T ton CO2 per dollar.

Clean protein 

In terms of animal suffering reduction, clean protein R&D, and in particular cell-based meat, is more cost effective than other animal welfare actions. But clean protein is also top effective in terms of greenhouse gas reduction. 

First estimate: acceleration of clean protein development

Scale of the problem: in a business-as-usual scenario, the global annual greenhouse gas emissions from livestock will increase to 10 billion ton CO2-equivalents, which equals 320 ton CO2e per second. (Current livestock emissions are 8 billion ton CO2-equivalents per year. Global protein consumption will increase, but will be slightly offset by replacement of animal protein with currently available animal-free protein).  

Solvability of the problem: 1%. This is a very conservative (low) estimate, composed of two factors: the tractability of research and the tractability of adopting the new technology. Concerning the tractability of research, I assume a 10% probability that extra clean protein R&D-funding is not superfluous, i.e. that accelerated R&D is not ineffective. Concerning the tractability of the new technology adoption, I assume a 10% probability that clean protein will eliminate livestock farming (i.e. that consumers switch from animal-based meat to clean, cell-based meat). The latter means that with 90% likelihood, consumers still prefer animal-based meat above cell-based meat, or that cell-based meat is expected to reduce animal farming with 10%. This is a much lower estimate than most predictions.

Accelerating R&D may be ineffective, because the extra research can be too soon or too late. Funding clean protein research thousand years from now will be too late, because by that time all research will have already been done, or everyone will already eat vegan. But starting clean protein research thousand years ago would have been too soon, because in those days people did not have research technologies such as computers to make progress in clean protein development. Similarly future new research technologies make current research obsolete. Examples of important new research technologies are powerful artificial intelligence and gene editing. Consider how CRISPR-Cas9 gene editing drastically accelerated genetic engineering. Years of genetic engineering research before the discovery of this CRISPR gene editing technology can be done much faster now with CRISPR. Similarly, years of work on the protein folding problem can now be done quickly with artificial intelligence. Suppose with current research technologies, developing a new clean technology takes twenty years, and current extra funding can speed up the development of the clean technology with one year. Hence, we have the clean technology in nineteen years. However, suppose after ten years, a new research technology is invented, that accelerates the research, such that ten years of research can now be done within one year. That means with this new research technology, we will have the clean technology in eleven years. The extra funding to shorten the development from twenty to nineteen years becomes superfluous, because we have the clean technology in eleven years anyway. 

Neglectedness of the solution: extra funding of clean protein R&D accelerates the development of clean meat such that it enters the market (at more competitive price and quality than animal-based meat) by 0,32 seconds per dollar. This assumes in the business-as-usual scenario, clean protein has a constant annual global R&D-funding in the order of 100 million dollar per year (e.g. cell-based meat had investments around  70 million dollar per year in 2019). The lower this amount of annual funding, the more neglected is the research and the more effective is an extra dollar of funding. For example, if current annual funding was only one dollar, adding an extra dollar of funding doubles the R&D efforts and accelerates the research with one year (because the research funded with one dollar next year could be done this year). 

The cost-effectiveness of clean protein development is the product of scale, solvability and neglectedness: 320 ton CO2e avoided per second x 1% x 0,32 seconds per dollar = 1 ton CO2e per dollar. This is a conservative, lower-bound estimate.

Second estimate: contribution to clean meat development

The above estimate is based on the annual greenhouse gas emissions of animal farming, expressed in tons of CO2-equivalents. However, these livestock-related greenhouse gases contain both long-lived gases (CO2 and N2O) and short-lived gases (methane) and it is very tricky to add them up in one measure scale (it is like adding up a stock measure with a flow measure). Hence, the notion of CO2-equivalents can be criticized. The second estimate considers long-lived and short-lived livestock related greenhouse gases separately. 

When considering animal farming, the long-lived gases, especially CO2, relate to the carbon opportunity cost: the lost potential of CO2 sequestration from forest restoration of agricultural land. Animal farming requires a lot of agricultural land, and this land is no longer available for carbon sequestration by reforestation. Switching to clean protein (e.g. cell-based meat, vegan food) decreases agricultural land use, because clean protein is much more land efficient than animal protein. The available land can be reforested.

Scale of the problem: the carbon opportunity cost of animal farming is 750 billion ton CO2. This is the amount of CO2 that can be absorbed by the trees and soils of abandoned farmland, through spontaneous reforestation. This estimate also corresponds with another study, which calculates a carbon opportunity cost of 5 ton CO2 per year per person who eats animal protein (e.g. an average omnivorous diet) instead of clean protein (e.g. an average vegan diet). Multiplying this personal carbon opportunity cost by 7,5 billion people and a 20 year carbon sequestration of forests, gives the estimated 750 billion ton CO2. The carbon opportunity cost of animal farming is huge: it is more than half of the total amount of CO2 that was added to the atmosphere and oceans from the burning of fossil fuels since the industrial revolution.

Solvability of the problem: 10^-11 contribution per dollar. Assume 10 billion dollar R&D-funding is required to have clean protein at competitive market prices (for example 100 years at current yearly funding rate of 100 million dollar per year, which is likely an overestimation). The bigger this required funding, the more difficult (less tractable or solvable) is clean protein research. Once it enters the market, assume a 10% probability that clean meat will eliminate animal farming (that consumers switch from animal-based meat to clean meat). This 10%, divided by 10 billion dollars, gives the total solvability estimate.

Neglectedness of the solution: assume a 10% probability that extra research funding will not be crowded out. When considering a personal contribution to the development of a new technology, there is a risk of funding crowding-out: other people will likely decrease their funding. Suppose if you invest an extra dollar, someone else disinvests a dollar. That means your extra investment is completely crowded-out by someone else’s disinvestment. In this case, your personal contribution increases, but the total amount of funding and hence the likelihood of the development of the new technology stays the same. From a consequentialist perspective, your extra contribution is futile. In reality, extra funding will not always completely crowd out funding by others. This crowding-out effect relates to neglectedness: the more neglected the research, the less crowding-out will occur. It also relates to the additionality: the less crowding-out, the more additive R&D-funding becomes, i.e. the more R&D really increases with an extra unit of funding. Here, I assume a very high crowding-out level of 90%. This gives a conservative, low estimate of the cost-effectiveness. 

The product of scale, solvability and neglectedness gives a lower-bound cost-effectiveness of 0,75 ton CO2 per dollar.

Next to the long-lived greenhouse gases and the carbon opportunity cost of animal farming, we have to consider the short-lived greenhouse gases, and in particular the methane emissions from farm animals. Clean protein (especially cell-based beef and animal-free dairy) has a huge methane mitigation potential.

Scale of the problem: future livestock methane emissions, when kept constant at the current level, contribute to the same global temperature increase as a one-off emission of 200 billion ton CO2. This estimate equals the global temperature increase due to global methane emissions at constant current level (0,3 Kelvin) times the fraction of global methane emissions coming from livestock (30%) divided by the global temperature increase per ton CO2 emitted (0,5 picokelvin).

Solvability of the problem: 10^-11 contribution per dollar, as above. 

Neglectedness of the solution: 10% probability of extra research funding not being crowded out, as above.

The product of scale, solvability and neglectedness gives a lower-bound cost-effectiveness of 0,2 ton CO2e per dollar.

Adding up the cost-effectiveness of carbon sequestration and methane mitigation due to clean protein, the total cost-effectiveness is around 1 ton CO2e per dollar, similar to the acceleration estimate above.

Clean energy

We can make similar back-of-the-envelope cost-effectiveness estimates for clean energy R&D. 

First estimate: acceleration of clean energy development.

Scale of the problem: 34 billion ton CO2 per year annual energy related CO2 emissions from burning fossil fuels.

Solvability of the problem: 10% probability of effectiveness. This corresponds with a high estimate of 90% that extra clean energy R&D is superfluous in the sense that it is either too soon (e.g. it will be overtaken anyway by new research technologies such as artificial intelligence) or too late (e.g. fossil fuel reserves will all be burned before new clean energy technologies enter the market).

Neglectedness of the solution: 22 billion dollar per year annual global funding for clean energy R&D.

The product of scale, solvability and neglectedness gives a lower-bound cost-effectiveness of 0,15 ton CO2 per dollar.

Second estimate: contribution to clean energy development

The second estimate is based on the amount of unused fossil fuel reserves when those fossil fuels are replaced by clean energy.

Scale of the problem: the total CO2 emissions from burning all reported fossil fuel reserves would be 1,6 trillion ton CO2.

Solvability of the problem: 5.10^-13 contribution per dollar. This corresponds with a 50% fraction of reported fossil fuel reserves not used due to clean energy replacement, divided by 100 billion global funding required for replacement of fossil fuels by clean energy. The latter equals 22 billion dollar per year annual global funding of clean energy R&D times 50 years of required research. (As clean energy has more R&D-funding than clean protein, its expected development time is shorter. But as development of clean energy to replace all fossil fuels is likely more difficult than developing clean protein, its development time is not too much shorter.)

Neglectedness of the solution: 10% probability of extra research funding not being crowded out.

The product of scale, solvability and neglectedness gives a lower-bound cost-effectiveness of 0,08 ton CO2 per dollar.

Both acceleration and contribution estimates are around 0,1 ton CO2e per dollar funding of clean energy R&D, an estimate similar to the Founders Pledge.