Cause area: climate adaptation in low-income countries

Summary

This is an investigation into climate adaptation in low- and middle-income countries as a cause area for effective altruists. I use climate adaptation to describe efforts to reduce the damage suffered from any given level of climate change, and climate mitigation to describe reductions in climate change itself (e.g. through reducing carbon emissions, carbon capture, geoengineering, etc).

EAs have historically evaluated climate change as a potential longtermist issue, focusing on its potential existential risk. The 80,000 hours problem profile focuses solely on how climate change could affect the long-term future. From this perspective, it has usually been rejected as unlikely to cause existential risk. In this review, I adopt a different framing of climate adaptation as squarely a global health and wellbeing (“neartermist”) concern, because it addresses issues that are already happening (increased temperatures, storm/flood risk, droughts, etc) and will only become more important in the next few decades. Given this, my subjective summary of climate adaptation is as follows:

Importance − 9.5/10. Using different model estimates of the social cost of carbon, I estimate that climate adaptation could mitigate 44-73% of all climate damages, which makes it quite important even by the most skeptical analyses of climate as a cause area.
Neglectedness − ⁹⁄₁₀. Very few actions are being taken to support climate adaptation in low-income countries. In theory, market incentives support adaptation to the changing climate, but with high adaptation frictions and low incomes to support these incentives, this is a partial solution at best.
Tractability − ⁶⁄₁₀. Unlike climate mitigation, adaptation does not involve global coordination problems, making it far easier to implement solutions. Furthermore, policy has lagged research in this area, which has developed a large knowledge base to support climate adaptation interventions. I outline several research-backed interventions to improve climate adaptation. However, the sparseness of existing work in this area means that there is little capacity to implement them, and also substantial uncertainty about their efficacy.

Conclusion: climate adaptation is very important and neglected. It is tractable in theory, but the shortage of organizations working on adaptation means that donations in this area would likely not beat GiveWell’s top charities. As a result, the most important contributions in this area are likely to come from incubators who can support the creation of new organizations to work on climate adaptation, and people who can advocate for existing climate organizations to develop adaptation projects. Insofar as donors can help this happen, I strongly recommend it as a capacity-building area: climate adaptation is necessary to improve the lives of hundreds of millions of people over the next century.

The problem

Climate change is already here. 2021 saw the world’s most extreme heat wave in North America, resulting in $11 billion of damages, and South Asia’s ongoing heat wave has caused glacial lake flooding, widespread blackouts and at least 100 deaths. Despite this fact, virtually all climate-related grantmaking and advocacy focuses on mitigating the causes of climate change over the next century. rather than adapting to its effects right now. We need to shift our perspective on climate change: it is not just a phenomenon with long-term effects, it is an ongoing set of environmental changes with ongoing damages.

As temperatures rise and extreme heat events become more common, more people are dying of heatstrokes and more crops are failing (Huang et al, 2011). Deadly pathogens like malaria are growing in newly-warmed environments where people have never dealt with them before (Couper et al, 2021). Coastal cities are suffering a reversal of fortune as they become more prone to floods and cyclones (Balboni, 2021). Rather than putting all our eggs in one basket by dedicating all our efforts to climate mitigation, we can substantially reduce these damages by directly adapting to the climate change that is already happening.

Climate adaptation is an enormous topic, with too many effects and too heterogeneous effects to be usefully reviewed as a single cause area. In this review, I narrow the scope in two ways.

I focus on climate adaptation in low-income countries. Climate damages are likely to be much higher in these countries, because they will face more extreme temperature changes and have less governance capacity to handle the effects of climate change. Furthermore, because of contextual differences between low-income countries and richer countries, the best interventions to improve climate adaptation are not necessarily the same in both contexts. For example, rich countries already have widespread access to cooling technology, but diffusing cooling technology is likely to be quite important in low-income countries. Conversely, rich countries are likely to suffer from the poleward movement of insect-borne diseases like malaria, whereas the more tropical low-income countries will be less affected. Thus, in this review I narrow my focus to the highest-impact subset of countries where interventions can be broadly useful across countries.
I focus on adaptation to increased temperatures, and ignore other elements of climate change (e.g. increased natural disasters). This is not because those other effects are less important—in fact, Hsiang and Jina (2014) find that natural disasters cause lasting economic damage, and their increased prevalence could add a breathtaking $10 trillion to the social costs of climate change. Rather, this cut is solely to keep the review short and focused on a coherent set of issues. I would welcome more attention to other elements of climate change that require adaptation.

Both of these restrictions are primarily relevant to the Tractability section, as the guiding framework for my proposed interventions. I believe that the Importance and Neglectedness sections apply to other aspects of climate adaptation, and climate adaptation in high-income countries.

Importance

Defining importance of adaptation

I conceive of the importance of climate adaptation as the proportion of climate damage that could be avoided with substantial adaptation. This is a useful definition because it captures the maximum upside—if the amount of climate damage were independent of our adaptation efforts, then we would say that adaptation has no value. Note that this definition does not distinguish between whether this adaptation needs to be funded, or will occur organically as a result of warming—see Neglectedness for a discussion of organic adaptation. The most important caveat about this definition is that since it focuses on the proportion of avoidable climate damage, it is defined relative to the social cost of climate change. I am not constructing a DALY measure of the costs of non-adaptation, or any measure that can be compared to many different cause areas.

This is a serious tradeoff, but in exchange, this definition buys us an unusually credible estimate of the social costs of non-adaptation. The social cost of climate change is summarized by the social cost of carbon, which is estimated from climate models. In the next section I describe how I use climate model estimates of the SCC to derive the social cost of non-adaptation to climate change.

Estimating importance from climate models

To measure the proportion of climate damage that could be avoided with adaptation, I use the Shared Socioeconomic Pathways (SSPs) developed by the climate research community to represent different socioeconomic scenarios and their resulting climate damages. Each SSP represents an assumed trajectory for critical social variables (fertility, economic growth, etc) and climate variables (emissions, atmospheric pollutants, etc) based on different assumptions about what policies the world will adopt. These social and climate variables interact to determine the social cost of climate change.

There are five SSPs, which can be roughly mapped onto the dimensions of mitigation and adaptation as: (Hsiang and Kopp, 2018)

SSP1 (Sustainability) - high mitigation, high adaptation.
SSP2 (Middle of the Road) - medium mitigation, medium adaptation.
SSP3 (Regional Rivalry) - low mitigation, low adaptation.
SSP4 (Inequality) - high mitigation, low adaptation.
SSP5 (Fossil-fueled development) - low mitigation, high adaptation.

These SSPs provide a neat way to capture the value of adaptation: SSP1 and SSP4 differ only in the extent of adaptation, as do SSP3 and SSP5. This means I can measure the importance of climate adaptation as the reduced damages under SSP1 (SSP5) compared to SSP4 (SSP3). Furthermore, if I interpret the social cost of carbon as the importance of climate mitigation, then I can derive an importance ratio: how important climate adaptation is compared to mitigation without adaptation. This is a useful benchmark for understanding how resources should be allocated between climate mitigation and climate adaptation. The importance ratio for high/low mitigation levels is $I R_{h i g h} = \frac{S C C_{4} - S C C_{1}}{S C C_{4}}$ $I R_{l o w} = \frac{S C C_{3} - S C C_{5}}{S C C_{3}}$ where $S C C_{n}$ is the estimated social cost of carbon under SSP $n$ . I use the low-adaptation scenarios, SSP4 and SSP3, as a baseline to represent how much we can improve on them by facilitating high levels of adaptation.

Yang et al (2018) average the social cost of carbon from nine different climate models under a 1.5% discount rate assumption to calculate the SCC under these different trajectories. They estimate:

SSP	Mitigation?	Adaptation?	Estimated SCC
SSP1 (Sustainability)	High	High	$10/tCO2
SSP4 (Inequality)	High	Low	$18/tCO2
SSP5 (Fossil-fueled Development)	Low	High	$12/tCO2
SSP3 (Regional Rivalry)	Low	Low	$45/tCO2

This yields the importance ratios $I R_{l o w} = \frac{18 - 10}{18} = 0.44$ $I R_{h i g h} = \frac{45 - 12}{45} = 0.73$

These importance ratios imply that high levels of adaptation reduce climate damages by 44-73% relative to a low-adaptation baseline. To be clear, this is all climate damages that are captured by climate models, discounted over the future. This is an enormous fraction and suggests that climate adaptation should be given comparable attention to climate mitigation.

One might worry that this result might be driven by the 1.5% discount rate used in this paper, which is much lower than in many others. By normalizing the effect as a ratio between two social costs of carbon which are computed with the same discount rate, I can avoid the effects of different discount rates (more nuance in the footnote)^[1].

Comparison to other cause areas

As emphasized before, this estimate of the importance of adaptation is strictly based on comparison to climate mitigation. Ideally, I would use the raw SCC estimates from each SSP to compute a monetized value of adaptation, and use that for a cost-effectiveness analysis that could be compared across global health and wellbeing causes. Unfortunately, there is no “unit of adaptation” which can be given a dollar cost, because adaptation takes many different forms. This is unlike climate mitigation, which can be framed in terms of tons of CO2 averted. Moreover, such an analysis would be highly uncertain due to the influence of the discount rate on the SCC estimate, whereas the comparisons I make are within the same discount rate and thus avoid that uncertainty.

In a broader sense, my goal is not to precisely estimate the importance of climate mitigation or climate adaptation. Even the most skeptical assessments of climate as a cause area acknowledge that climate mitigation is tremendously important, and diminish it only by comparison to longtermist/existential risk areas. Rather, my goal is simply to sketch out the optimal allocation of resources between climate mitigation and climate adaptation, and show how far we are from that optimal allocation today due to underinvestment in climate adaptation.

Sources of Uncertainty

Since I draw estimates of importance from climate models, all sources of uncertainty in climate models apply here. Pindyck (2013) critiques climate models for using arbitrary damage functions and being sensitive to arbitrary factors like the discount rate. Using a ratio of SCC estimates partially addresses these concerns, since these factors affect the numerator and denominator comparably, making the ratio less sensitive to these arbitrary decisions. Nonetheless, I agree with Pindyck’s caution against interpreting climate model outputs with any kind of certainty, and the inherent uncertainty of climate models is not something I can fully solve in this analysis.

Another concern is that this importance ratio does not represent a very precise concept; after all, we are just comparing scenarios with heuristic ideas of “high” and “low” adaptation. This is certainly true. I am leaning heavily on the climate research community’s categorization of socioeconomic trajectories, which are ultimately the best comparison we have to mark the effect of adaptation.

Tractability

Unlike mitigation, climate adaptation is not a coordination problem. If one place reduces emissions while others do not, that place will hardly benefit from mitigation. In contrast, if one place adapts to climate change while others do not, that place will fully benefit from its adaptive measures. This makes climate adaptation much more tractable for an individual organization, compared to climate mitigation, which requires global coordination to be effective.

In the following sections, I outline some promising interventions to help with climate adaptation. It’s important to acknowledge that most of these are not taken from the practices of real organizations, but rather from research that strongly suggests their efficacy. In practice, research need not translate exactly to effective interventions (e.g. due to scaling concerns), so this gap increases my uncertainty about the efficacy of these interventions. Nonetheless, I think they are a proof-of-concept for what effective climate adaptation interventions would look like.

Crop-switching support

Agriculture is the sector likely to be most damaged by climate change, and switching crops planted has a strong potential to mitigate these damages. Rising and Devineni (2020) find that crop-switching would reduce US agriculture’s damage from climate change by 50%, and Costinot, Donaldson and Smith (2016) estimate that the global losses in agriculture from climate change could be reduced by 66% just by farmers reallocating production optimally to the new climate. This makes crop-switching an important margin for climate adaptation: if it can occur organically, then climate change becomes substantially less worrying for agriculture—and for the countries whose economy and poverty alleviation depends on the agricultural sector.

The challenge is that it is murky whether crop-switching organically occurs when it is necessary for climate adaptation. The evidence on this question is mixed. On the one hand, some papers have found cross-sectional evidence of adaptation: Seo and Mendelsohn (2008) found that temperature and precipitation are highly predictive of South American farmers’ crop choice. On the other hand, Burke and Emerick (2016) found that changes in the US’s climate over the twentieth century did not reduce the share of farmland planted in corn, despite significantly reducing its yields. In adjudicating between these studies on crop-switching, I find the latter evidence more credible as a basis for predicting future crop-switching. Seo and Mendelsohn compare different locations at the same time, so the effect they find is plausibly the result of long-run local knowledge in each community about what the right crop for their climate is. Indeed, one expert I spoke to said that farmers in India pick what crops to plant based on collective knowledge and learning from their peers, supporting this hypothesis. However, what we want to know is how this local knowledge will adapt to changes in climate. So Burke and Emerick’s comparison of the same locations across time is better suited to answering that question. Moreover, frictions to crop-switching in developing countries are likely to be higher than in the US, so their setting offers a credible lower bound. Thus, I don’t believe that long-run adaptation in crop choice is likely to happen on its own—I think there is substantial room for philanthropic work to reduce frictions to adaptation.

There is very little evidence on what frictions matter most for crop-switching, except for one survey conducted by Abid et al (2017) where farmers reported that lack of information were the largest constraint on their adaptation to changing climate. Additionally, an expert I spoke to highlighted a key dynamic they had observed: small farms are more risk-averse and tend to follow their neighbors’ choices, while large farms are more likely to experiment with new crops on small subplots of their land. This expert also emphasized the role of price expectations in farmers’ crop choices, stating that farmers were likely to plant crops even if they hadn’t planted them before, if they anticipated good prices (usually inferring from past prices).

These facts point towards two types of interventions. First, crop-switching could be achieved through a network-based intervention, seeding large farmers with information and more resilient crops to cultivate, in the hopes that their success will diffuse adoption to smaller farmers. Second, subsidies for more heat- and drought-resistant crops could lead farmers to adopt them despite existing frictions.

Cooling technology diffusion

Extreme heat is becoming an enormous global health issue, causing deaths, illness and reduced productivity. Burkart et al (2021) show that in 2019, 65,000 people died from extreme heat-related morbidities in nine countries they studied, compared to 41,000 in 1990. Ebi et al (2021) show that over a billion workers worldwide are exposed to extreme heat, and a third of them suffer from negative health effects. Somanathan et al (2021) find that an increase in average temperature of 1 degree reduces industrial output by 2%, driven by reduced output per worker.

Given these worrying outcomes, access to cooling technology is quickly becoming an essential coping tool for climate change-induced periods of extreme heat. In fact, it has always been essential in this way: Barreca et al (2016) show that temperature-induced mortality in the US reduced by 75% over the 20th century, with the spread of air conditioning accounting for essentially all of the decline. Given the rising frequency of extreme-heat periods, diffusing cooling technology could save lives, and is a viable candidate for an intervention.

The major caveat is that we can only recommend interventions to diffuse cooling technology if they a) don’t substantially worsen the problems of emissions and electricity demand, and b) are not already happening. Indeed, AC diffusion is happening: the IEA projects that by 2050, India and China will add 1.1 and 0.9 billion AC units respectively, and energy demand for ACs will triple. This, along with the emissions effects of AC, suggests that AC diffusion is not a good candidate intervention.

Given this rapid expansion of ACs, it’s natural to ask whether cooling technology diffusion might be a solved problem. However, I believe that aggregate AC adoption projections are misleading with regards to whether the most heat-vulnerable households will actually adopt AC and avoid heatstrokes. The most vulnerable households usually lack grid connections to power an AC unit, and even when they are connected, an AC unit is a large piece of infrastructure with a high fixed cost of adoption. Indeed, Miguel, Lee and Wolfram (2021) subsidize rural electrification and none of the newly-electrified households in their sample purchase an AC unit. The IEA’s projected “cold crunch” is not exclusive with a large number of vulnerable households continuing to lack access to cooling technology. Thus, I think that there is room for interventions that promote access to cooling technology to save lives from heatstrokes.

Alternative cooling technologies exist that could be significantly lower-cost, lower-infrastructure and non-emissions-generating, such as electric fans. Jay et al (2021) state that electric fans reduce physiological heat strain and improve comfort (although they caution that fans cannot be used above 45 degrees Celsius). Beyond electric fans, all cooling technologies rely on electricity, which means that electricity access interventions are critical to improve climate adaptation.

These two facts point to two angles of approach to cooling technology diffusion. First, electricity access interventions are essential for adapting to extreme heat under climate change, and advocacy for electricity access could be a key approach for effective interventions. Second, promoting access to electric fans is important, especially fans that can work with unreliable electricity supplies that are characteristic in developing countries—e.g. subsidies for battery-operated or solar-powered fans could be impactful.

Development of climate-resistant seed varieties

Soon, agriculture in low-latitude areas is likely to suffer from higher temperatures, more extreme heat days, and lower rainfall. One important adaptive measure could be the development of heat-resistant seed varieties so that crop failure and the resulting food insecurity are much less likely to occur under climate change. Indeed, Moscona and Sastry (2022b) show that innovation in agriculture has averted 20% of the potential damage to US agriculture from extreme heat since 1960, showing that it can be a powerful force for climate adaptation.

Even now, the development of climate-resistant seeds is an area with some funding by governments and philanthropists. The Foundation for Food and Agriculture Research has launched a program to develop climate resilience in crops, in partnership with the Gates Foundation. So far this program has disbursed $5 million in grants. In addition, the US government funds agricultural research that includes climate-resistance (example). This success points to a role for philanthropists in funding research on heat-/drought-resistant seed varieties.

Neglectedness

Climate adaptation is very neglected, with almost no organizations and limited market support for adaptation interventions.

Nonprofits

Using recommendations and grants as a metric indicates that climate adaptation is extremely neglected compared to climate mitigation. All the seven climate charities recommended by Vox’s Future Perfect or the three recommended by The Life You Can Save focus on climate mitigation. Founder’s Pledge has never given a grant to an adaptation-focused nonprofit, and Giving Green focuses on US policy change. This neglect is not limited to EA funders: Charity Navigator does not recommend any adaptation-focused organizations, and large environmental nonprofits like Greenpeace and the Environmental Defense Fund do not have any verticals dedicated to adaptation.

There are two exceptions to this neglect. As described above, the development of climate-resistant seeds has active funders like the Foundation for Food and Agriculture Research, the Gates Foundation and the US government. Nonetheless, I think climate-resistant seeds are still a good funding opportunity. The expert I spoke to opined that the field is still small relative to its scientific potential, and it can still absorb much more funding. Moreover, there is strong reason to believe that development of climate-resistant seeds happening in the US (which is where most funding goes) does not actually benefit low-income countries. Moscona and Sastry (2022a) show that agricultural innovations in the US (the global hub of agricultural research) are mismatched to countries with different climates from the US: mismatch reduces the adoption of improved seed varieties among African farmers by 43%. As a result, innovation that protects US agriculture may not help farmers in more tropical countries, and we should prioritize funding agricultural R&D taking place in tropical countries. Moscona and Sastry (2022a) show that agricultural patents have been growing rapidly in Brazil, Russia, India and China, although their level is still far below the US. This points towards opportunities to fund local research that will aid in the development of drought- and heat-resistant seeds. The Gates Foundation’s materials state that they support public-sector crop breeding taking place in Africa and Asia, an approach that is strongly supported by Moscona and Sastry’s mismatch hypothesis, and which I would recommend for grantmaking in this area.

The second exception to this neglect is that, although they do not use climate adaptation framing in their materials, Clean Air Task Force’s energy access vertical could help people adapt through electrification (and the associated access to cooling technology). However, currently CATF’s energy access vertical seems to be only piloting, and they are present only in sub-Saharan Africa. This review justifies scaling up their energy access work substantially to help them move beyond piloting and potentially expand their regional scope in the future.

Markets/Organic Adaptation

Some of the interventions above are market-based interventions, which makes it natural to ask whether market incentives will help solve the problem without grant funding. For example, shouldn’t there be a market incentive to spread air conditioning to people in warming countries who are going to value it highly? The key obstacle to the market implementing these interventions is income. If vulnerable people could easily afford air conditioning or pay large sums for heat-resistant seeds, then purely market solutions might address the problem. But this is obviously not the case, and over the time horizon we need adaptation (the next few decades) income growth is unlikely to be a sufficient solution. Indeed, according to the IPCC Sixth Assessment Report released in 2022:

Despite progress, adaptation gaps exist between current levels of adaptation and levels needed to respond to impacts and reduce climate risks (high confidence). Most observed adaptation is fragmented, small in scale, incremental, sector-specific, designed to respond to current impacts or near-term risks, and focused more on planning rather than implementation (high confidence). Observed adaptation is unequally distributed across regions (high confidence), and gaps are partially driven by widening disparities between the estimated costs of adaptation and documented finance allocated to adaptation (high confidence).

Thus, we cannot rely on organic adaptation to address climate impacts.

Beating GiveWell Top Charities?

The hundred-million-dollar question for this review is—does climate adaptation beat GiveWell top charities as a funding area? In my opinion, this area does not admit a good back-of-the-envelope calculation of this question, because each intervention targets different people (farms, households, R&D institutions) and has a totally different outcome (crop yields, income, reduced mortality), so each intervention will admit its own BOTEC independently of the others.

That caveat aside, this analysis leads me to two conclusions. First, climate adaptation has the potential to beat GiveWell top charities: climate change seems intuitively likely to cause (at least) hundreds of millions of DALYs of harm, and averting 44-73% of that would be hugely impactful, especially since there is very little work towards doing so right now. Second, the lack of capacity means that funding climate adaptation on the margin will almost certainly not beat GiveWell top charities. The primary candidates for funding right now would be Clean Air Task Force’s energy access vertical, or agricultural R&D taking place in low-income countries. I would support grants in these areas, but they seem too limited in scope to realize the potential of climate adaptation on their own.

These two conclusions paint a nuanced picture in which climate adaptation is a vital area for philanthropic work, but it currently admits capacity-building more than simply funding large-scale interventions. The most important work in this area is to create organizations and projects that could use funding to implement adaptation solutions. This means it could be very valuable if funders used grant money to incentivize/support the setup of new organizations working on climate adaptation. Such organizations could plausibly beat GiveWell top charities if they achieve scale.

Conclusion

Adaptation to climate change’s ongoing effects is an important, neglected and moderately tractable cause area. Policy has lagged research in this area, leading to a robust evidence base for interventions to improve climate adaptation, even as very few actions are actually taken by climate action organizations. The shortage of existing organizations and projects means there is not a lot of capacity of absorb funding in this area, and marginal donations would not likely beat GiveWell’s top charities. Nonetheless, the importance and neglectedness of climate adaptation means that we should make it a priority to start new projects and organizations focused on climate adaptation: to the end that philanthropic funding can accelerate this, I would recommend it as a funding area.

Acknowledgements: Thanks to Liza Brover, Tejas Subramaniam, Oliver Kim, Cassandra Melax, and Trevor Woolley for helpful feedback on this essay.

↩︎
One wrinkle to this claim is that these SCC estimates are derived by averaging nine model estimates: when there is more variability between them, some of them will be extremely high and push up the average SCC. This matters because each SSP has a different range of variability between models, so a low discount rate will push up the average SCC for SSPs with higher tail risk. I was worried about this because each SSP has a different tail risk, and a lower discount rate amplifies uncertainty and pushes the SCC upwards for SSPs with more tail risk. From visually inspecting the paper’s model estimate graphs, SSP1 and SSP4 have relatively similar estimates across models, so the comparison would not likely be altered by a different discount rate. SSP3 has large variability but SSP5 has even larger variability. Thus, adjusting the SCCs downward with a higher discount rate would likely decrease SSP5 more than SSP3, which would only increase the importance ratio. Thus, the low discount rate is unlikely to be inflating the importance of climate adaptation.