Hide table of contents



Over the past year, several figures in EA, such as Will Macaskill, Toby Ord, and Niel Bowerman have discussed the relevance of climate change for the EA community.

Most analyses of the impacts of climate change by EAs so far have focused on whether climate change is a direct existential risk, e.g. Ozymandias, Halstead, Bowerman, and Linch. Few have yet explored other ways in which climate change is likely to impact the long-run future. My hope is that a summary of the key points of Climate Shock will go towards addressing that gap.

Toby Ord's The Precipice highlights that existential risk factors can be equally as important as existential risk. A longtermist is typically interested in reducing total existential risk, not only in addressing specific direct risks.

Climate Shock : The Economic Consequences of a Warmer Planet gives reasons to see climate change as a likely contributor to total existential risk, and important for other reasons from a rationalist longtermist perspective. Climate Shock is recommended in the reading list for The Precipice.

By the way, the European winter of 2019-20 was the hottest on record by far.


Gernot Wagner and Martin Weitzman provide a highly accessible introduction to repercussions of a hotter planet, with a focus on understanding and responding to tail risks, and the potential implications of geoengineering.

Gernot Wagner is a Clinical Associate Professor at the Department of Environmental Studies, New York University. Martin Weitzman (1942-2019) was a professor of Economics at Harvard University and is one of the most highly cited academics in Economics to date.

Key points

  • The severity of climate change is poorly captured in many conventional cost-benefit analyses. The probability of above 6C of warming under current pathways is above 10%. What this would mean for society is highly uncertain. Most conventional analyses extrapolate differences from the 0C-1C range.
  • The causal chain from emissions to human impacts is long and convoluted, and we should also factor in potential costs such as mal-adaptation, and future conflict. Understanding the human impacts of climate change is profoundly complex, as there is both model uncertainty and uncertainty in the applicability of the models themselves. Many costs go unrecognised.
  • Economists tend to model the impact of climate change by deducting costs in the future from growth from a compounded growth rate. However, this misses the true cost. Social costs will be unequally spread, with most harm going to the poorest. Moreover, the increasing deleterious effect of climate change might cause a deterioration in the growth rate, and also increase the share of output affected by climate change. This altered assumption would place more pressure on reducing emissions than models such as the DICE recommend.
  • Geoengineering is being explored as a solution to climate change. But it carries many risks, which could result in doing far more harm than climate change alone. Sensitivities are poorly understood, and the potential global influence of technologies is on a similar scale to that of nuclear weapons. Termination shock and counter-geoengineering make the strategic picture more complex. Asymmetries of impact and unclear attribution could exacerbate conflicts between major powers.
  • The clearest route forward is to push for a price on carbon. Individual actions such as eating less meat and reducing air travel are useful insofar as they lead to this policy objective. Concerningly, geoengineering will soon become within the reach of single states or even wealthy individuals. As a result, a priority should be examining the standards and mechanisms for the governance of geoengineering research.

Summary and selected quotes

Chapter 1: 911

Impacts and attribution

Attribution science is demonstrating the link between human activity and more and bigger storms, the risk of heatwaves, floods, and other extreme weather events. Over time, we risk locking in bad outcomes.

By [2100], it will be too late to act. We can’t re-create glaciers and polar ice caps, at least not in human timescales. The severity of the problems will have been locked in by past action, or lack thereof. Future generations will be largely powerless against their own fate.

Climate change is fiendishly difficult problem to address because it is global, long-term, largely irreversible, and the impacts have significant uncertainty.

Deep uncertainties

  1. Most climate models are unduly skewed toward the known, sometimes making them much too conservative. Until recently, scientific understanding of melting polar ice caps had been so poor that most models simply left it out.
  2. Second, there are fundamental things that we don’t understand about the way the climate works. Temperatures in the past century have been increasing at an increasing rate, temperatures over the poles have warmed more than elsewhere.
  3. The causal chain between from emissions to societal impacts is highly complex. We are uncertain about the links between emissions, atmospheric concentrations, temperatures and physical climate damages, and the link between physical damages and their consequences, and, at least as important, how society will respond: what coping measures will be undertaken, and how effective they will prove to be.

Policy responses

A carbon tax would be ideal. However, the current structure of the markets pushes in the opposite direction.

The world subsidizes fossil fuels at a rate of over $500 billion per year. That is equivalent to an average worldwide subsidy of some $15 per ton of carbon dioxide emissions.

The authors are unequivocal in pointing the finger at the fossil fuel industry.

It’s incredibly hard to overcome the huge vested interests fighting against... most every economist’s vision of the ideal world.

Chapter 2 : 411 - Bathtub

The importance of net zero

This chapter describes the bathtub analogy, and makes the point that reducing atmospheric concentrations requires switching off the taps, not reducing the flow rate.

It’s not enough to stabilize emissions and avoid annual increases in how much carbon dioxide the world pumps into the atmosphere. We need to decrease emissions to near zero to begin to bring down concentrations.


Bill Nordhaus’s Dynamic Integrated Climate-Economy (DICE) model is the most prominent of those that try to make sense of it all. It takes trade-offs between the climate and the economy as a starting point to calculate an optimal path and price for carbon dioxide emissions...

DICE and other models use an average social cost of carbon of $40, but this misses out on several other major components of cost, such as ocean acidification.

Ocean acidification

Most of the carbon dioxide emitted eventually ends up in the oceans, turning them more acidic... Oceans are likely over 25 percent more acidic than at the dawn of the industrial revolution...
Ocean acidity now is increasing ten times faster than during the last great die-off of some marine organisms around 56 million years ago… too little is known about the full implications of more acidic oceans.

The main weaknesses in the DICE model come from the failure to capture tail risk and unmodellable factors, which are covered in the following chapter.

Chapter 3: Fat tails

This section points out that the severity of climate change is likely to be fat-tailed with a weighting towards higher impacts. Moreover, the current [2015] view is that even moderate assumptions give a significant chance of catastrophe.

Take the latest consensus verdict at face value and assume a “likely” range for climate sensitivity of between 1.5 and 4.5°C (2.7 and 8°F). Equally important, stick to the IPCC definition of “likely” and assume it means a chance of greater than 66 percent, but less than 90 percent. And take the IEA’s interpretation of current government policy commitments at face value. Here’s what you get: about a 10 percent chance of eventual temperatures exceeding 6°C (11°F), unless the world acts much more decisively than it has.

Moreover, this distribution only shows expected average mean surface temperature rises, not the convoluted and complex chain of impacts on human society.

The business of pinning down specific impacts is messy and fraught with its own uncertainties. There are known unknowns aplenty. Unknown unknowns may yet dominate.

Modelling damages

Many studies in climate science choose quadratic damage functions arbitrarily. It seems plausible that exponential damage functions could be more appropriate. These would significantly change the weighting of the severity of bad outcomes.

Instead, DICE mostly relies on something close to quadratic extrapolations. As far out as 6°C (11°F), it all becomes guesswork. Using a quadratic function is a convenient shortcut, but it’s not much more than that. Lots of other extrapolations would fit the observed damages on the lower end of the scale but would yield wildly different results on the upper end. For instance, figure 3.3 shows how estimating exponential rather than quadratic warming yields starkly different results.
When it comes to high-temperature damages, the state-of-the-art economic models simply aren’t much better than fitting a curve around what we know at low temperatures, and extending it into what we don’t— well beyond the range of historically observed temperature increases into ones that mark uncharted territory for human civilization.

Climate change and economic growth

This section explores several issues relating to the relationship between climate change and economic growth.

It’s also not at all clear that we should be thinking about damages as a percentage of output in any given year. Standard practice for DICE and other models is to assume that the economy hums along just fine until damages from climate change get subtracted at some point in the future.

Compounding growth over the next century is often argued to outpower the costs of climate change. However, Wagner and Weitzman argue that climate change could decrease the growth rate itself.

Instead assume that damages affect output growth rates rather than output levels. Climate change clearly affects labor productivity, especially in already hot (and poor) countries. Then the cumulative effects of damages could be much worse over time.

More broadly, lowering the growth rate could reduce both the compounding growth rates over time, decreasing the long-run direct effect of growth, and also increasing the length of the time of perils (see Aschennbrenner).

Astronomical stakes

Finally, given that there are potentially extremely bad outcomes [even if they would not directly caused human extinction], then standard economic analyses fail to capture very large, if not quite astronomical, stakes.

We ought to focus on avoiding the possibility of these kinds of catastrophic damages in the first place. Some call it a “precautionary principle”—better safe than sorry. Others call it a variant of “Pascal’s Wager”— why risk it, if the punishment is eternal damnation?

Chapter 4: Willful Blindness

This section opens with a short section on ignorance and culpability, arguing that the clear evidence of the harms of climate change means that inaction (omission) is equivalent to a harmful act itself. It then moves to looking at climate change compared to other existential risks.

Climate change and existential risks

The chapter briefly discusses five other sources of existential risk, listed as 'biotechnology, nanotechnology, nukes, pandemics, and robots'. They conclude as follows.

In the final analysis, climate change is far from the only potential catastrophe humanity ought to be worrying about. Others, too, deserve more attention, and funding.

However, Wagner and Weitzman argue that work on climate change is worthwhile from an existential risk perspective for the following reasons.

  • First, a high chance of catastrophe. Current models [in 2015] indicated a greater than 10% chance of over six degrees of warming.
  • Second, the gap between our current efforts and what’s needed on climate change is enormous.
  • Third, climate change has firm historical precedence. Humans have never experienced it, but the planet has.

Chapter 5: Bailing out the planet


The eruption of Mount Pinatubo on June 15, 1991 put around 20 million tons of sulfur dioxide into the stratosphere. As a direct result of the volcanic explosion, global temperatures temporarily decreased by about 0.5°C (0.9°F). However, it did not address other direct effects of carbon pollution, such as ocean acidification, and potentially many others.

Geoengineering seems to be a tempting solution to climate change, but there are several reasons to be sceptical of it as a solution. If anything, the peril of the path ahead should make us even more focused on reducing emissions now, for the following reasons.

1. Extreme leverage ratios

The sulfur dioxide released by Mount Pinatubo reduced temperatures by about 30,000 times as much as the same amount of carbon dioxide would have increased them.

This extreme leverage ratio means that some geoengineering interventions should be 'likened to nuclear weapons', or other potential weapons of mass destruction. And their capability will only increase over time.

Even using today’s technology, a more targeted geoengineering intervention could possibly achieve leverage ratios near a million to one.

2. Free drivers / the unilateralist's curse

While carbon emissions are caused globally, it will be in the hands of states and non-state actors to interfere with the climate (see the unilateralist's curse).

Most put the direct engineering costs on the order of $1 to 10 billion a year... It’s not nothing, but it’s well within the reach of many countries and maybe even the odd billionaire.

3. Termination shock

Geoengineering is a symptomatic treatment, and the underlying mechanism causing warming would not be abated. If geoengineering was suddenly stopped, global temperatures could rocket back up again, causing significant climatic instability.

A sudden jump from abruptly ending geoengineering would result in all sorts of additional issues.

4. Governance of geoengineering

If one state conducts geoengineering, and extreme weather events are caused elsewhere, then political dynamics are likely to lead to blame even when attribution is unclear.

The complexity and asymmetries of geoengineering mean global governance is required.

Whatever geoengineering we could ever do on a sustained, global scale would require almost unprecedented global governance systems. It’s easy to imagine that falling apart for all sorts of reasons. War is one.

Chapter 6: Climate wars

Interstate conflict

Climate change will have an unequal effect on different countries, and so incentives will push in different directions. Geoengineering could become a divisive issue and a powerful weapon, driving tensions between states, and increasing the risk of great power war.

What if optimizing geoengineering for India hurts China, and vice versa? Would we want a geoengineering match between two nuclear powers, each with over a billion people?...
Let the climate war games begin.


Moreover, states could try to counter other the geoengineering actions of others. Even under optimistic scenarios, the side effects could be drastic.

Yet both would likely come with their own sets of unpleasant side effects, which are unlikely to cancel each other out. It’s also possible to imagine nonlinear responses.

Chapter 7: What you can do

What should we do, faced with this challenge? Reduce emissions now.

On an individual level, Wagner and Weitzman recommend that readers recycle, cycle, eat less meat, and offset their emissions, but conclude that ultimately policy reform is the most meaningful lever.

The goal, in the end, is to enact the best overall policies that will guide market forces in the right direction. So if asking one more person to recycle more is the foot in the door for their going to the polls and voting for the right policies that are in the common interest, great.


Climate Shock provides several reasons why climate change is relevant from a longtermist perspective. First, they argue that there is a significant chance of a global catastrophic risk, with a 10% chance of over six degrees of warming. While this might not be a direct existential risk, it is likely to be pretty bad. The impact of climate change for humans is based on a long and poorly understood causal chain, from emissions, to atmospheric concentrations, weather patterns, local weather, and the adequacy (or inadequacy) of social responses. There are many known unknowns and so there could be many more unknowns unknowns.

Second, costs are poorly captured in standard CBAs for many reasons. Climate change could decrease the economic growth rate, putting long-run compounding on a lower trajectores; the full impacts include many unmodellable factors; and unabated emissions may lead to global catastrophes from which recovery is uncertain, and it is hard to quantify the impact of a global catastrophe.

Third, geoengineering could exacerbate tensions between great powers. Single countries or wealthy individuals may choose to act unilaterally. The asymmetries of the impacts of climate change, and the potential impacts of geoengineering means that global decisions will be fraught with conflict. The difficulty of establishing attribution of extreme weather events could lead to public outcry and international pressure. This could raise the probability of a great power conflict, and harm efforts to govern other risks, such as nuclear security, biosecurity, and AI.

In terms of solutions, policy reform is needed because climate change is a collective action problem. Global problems need global solutions. A carbon tax is potentially the best way forward. Geoengineering should only be used as a last-ditch essential measure, as it is fraught with uncertainties and could potentially lead to much worse scenarios than conventional climate change trajectories. Given that it might be used at some point in the future, more work is needed to establish the best ways to govern geoengineering at a global level.

What does this mean for effective altruism?

The relevance of climate change for EA is an open question, discussed by Will, Toby, and Niel in the links at the top of this article. A longer discussion took place one year ago here, with interesting arguments for and against.

As far as I'm aware, at present the main organisations working on climate change in the EA community are CSER, and Founders Pledge, who recommend CATF.





More posts like this

Sorted by Click to highlight new comments since:

Something that the authors of this book perhaps should have highlighted is that DICE's main virtue is its simplicity: it is far from being either the only or the best IAM for most analyses. However, to appreciate how badly calibrated the damage function is, here's a note from the documentation:

"However, current studies generally omit several important factors (the economic value of losses from biodiversity, ocean acidification, and political reactions), extreme events (sea-level rise, changes in ocean circulation, and accelerated climate change), impacts that are inherently difficult to model (catastrophic events and very long term warming), and uncertainty (of virtually all components from economic growth to damages). I [Nordhaus] have added an adjustment of 25 percent of the monetized damages to reflect these non-monetized impacts. "

(Quote is from the manual of DICE 2013R, http://www.econ.yale.edu/~nordhaus/homepage/homepage/documents/DICE_Manual_100413r1.pdf , still valid for 2016R version as per https://www.pnas.org/content/114/7/1518/tab-figures-data .)

More from Ben
Curated and popular this week
Relevant opportunities