top view photo of cropland

Exaggerating the risks (Part 3: Climate change, lessons from Halstead)

Given progress in emissions, the risk of human extinction from the direct effects of climate change now seems extremely small … I construct several models of the direct extinction risk from climate change but struggle to get the risk above 1 in 100,000 over all time.

John Halstead, Climate change and longtermism

1. Recap

Many authors think that humanity currently faces high levels of existential risk, often between 10-50% in this century. In this series, I argue that leading estimates of existential risk are often exaggerated.

Part 1 introduced the series along with key concepts such as existential and catastrophic risk. Part 2 looked at Toby Ord’s claim that there is a 1/1,000 chance of irreversible existential catastrophe from climate change by 2100.

Climate change is a serious problem, and there is no doubt that it poses a substantial catastrophic risk to humanity. However, I argued, the case for existential risk from climate change is not so clear.

In particular, I argued, Ord gives almost no reasons to think that climate change poses a 1/1,000 chance of irreversible existential catastrophe by 2100, and Ord gives us many reasons to think that it does not.

However, readers might complain that this isn’t enough. Even if Ord gives no strong reason to accept a 1/1,000 risk of existential catastrophe from climate chance, might there be other reasons to accept a high estimate of climate risk? Enter Halstead.

2. The Halstead report

One reason that I am often skeptical of existential risk estimates is that many estimates are largely disconnected from established scientific methodology, modeling techniques, and matters of empirical fact. When we take a long, careful look at the facts, we begin to see that what first struck us as very risky is in fact much less risky than we might suppose.

John Halstead completed a DPhil in Political Philosophy at the University of Oxford. He has held a number of research positions within and around the EA community, most recently as head of applied research at Founders Pledge.

As part of the supplementary materials for MacAskill’s book, What we owe the future, Halstead prepared a 400-page report, Climate change and longtermism. The report systematically and carefully reviews the scientific evidence to conclude that existential risk from climate change is many times lower than Ord proposes. Drawing on interviews and advice from dozens of scientists, as well as hundreds of scientific papers and reports, Halstead finds himself unable to identify plausible avenues of support for high risk estimates.

Here are Halstead’s conclusions in his own words:

Given progress in emissions, the risk of human extinction from the direct effects of climate change now seems extremely small … I construct several models of the direct extinction risk from climate change but struggle to get the risk above 1 in 100,000 over all time.

My best guess estimate is that the indirect risk of existential catastrophe due to climate change is on the order of 1 in 100,000.

Note that these are estimates of all-time risk from climate change, not of risk in this century. Many of Halstead’s arguments suggest that risks in this century are significantly lower than these.

In the next few iterations of this series, I want to take a careful look through every risk pathway reviewed by Halstead. I want to show, as slowly and carefully as I can, why I think that the risk of existential catastrophe along each pathway is limited. This will, for the most part, be an opinionated summary of Halstead’s report. I would highly encourage readers to make their own way through the Halstead report and make up their own minds.

These posts will be long, boring, full of charts and largely devoid of philosophical argumentation. I won’t apologize for that. Facts matter. Evidence matters. Existential risk estimates must be grounded in evidence, and examining the evidence is often a long, boring process. It must be done.

3. Agriculture

Chapter 5 of Halstead’s report looks at the effects of climate change on agriculture. Climate change will directly threaten crops in many regions through heat stress and drought risk. Climate change will also harm agricultural workers by exposing them to heat stress and risks such as flooding.

There is no doubt that climate change will bring changes to agriculture. These changes will be challenging, especially to the economies and food supplies of many of the earth’s poorest nations. But for agricultural changes to amount to an existential risk, we would have to be faced with threats such as mass starvation of a sizable fraction of the human population.

Even then, it is far from clear that mass starvation would lead to human extinction or to the permanent curtailment of our potential for future development. Humans have already survived many large-scale famines, so unless climate change were to be severe enough to make agriculture unsustainable in the vast majority of habitable regions, the path from agricultural impacts to existential risk is far from clear.

But let us just focus on the risk of starvation. How likely is it that climate change will lead to extreme, worldwide famines by 2100? Here a few points are relevant.

First, almost all experts expect global food production to increase steadily over this century, even under many of the most pessimistic climate scenarios. That is because the harms from climate change will be offset by the spread of yield-increasing technologies such as improved farming equipment, synthetic fertilizers and pesticides, irrigation, and novel crop varieties.

For example, a recent report by the Food and Agriculture Organization of the United Nations reviews changes in crop yields across a variety of crop types, regions, and climate change scenarios. Here are the results for non-irrigated systems:

In this model, there may be a modest downward influence of climate change on crop yields, although some regions are expected to benefit. However, technological change is expected to substantially outpace climate change, dragging up crop yields in even the worst scenarios. None of this should be taken to deny that agriculture is in for a rough ride, but it is hardly evidence that a world-ending famine is on the horizon.

Second, impacts on agriculture can be partially mitigated. We can irrigate our crops; migrate crops to more temperate regions; or improve water management to deal with drought. We can also develop new crop varieties that are better-suited to a changing climate. When these changes are built into agricultural models, the future looks better still. For example, here is what the same report by the Food and Agriculture Organization of the United Nations projects for crop yields in irrigated systems:

On these projections, the impact of climate change on yields for irrigated crops may actually be net positive.

Third, levels of warming that would be lethal for agriculture are well out of reach on even the most pessimistic scenarios for warming by 2100. Here are Halstead’s estimates of the levels of warming needed to wipe out 8 months of the growing season in different localities:

Almost all of these lethal limits are out of reach on even the most pessimistic warming trends. The latest IPCC report forecasts warming across a range of scenarios:

In even the most pessimistic scenario, warming is projected to stay within five degrees, and many scientists expect warming to be far lower than this. (Halstead reviews the evidence and suggests that the most likely scenarios lie on the yellow line).

All of this is well below the levels needed to make agriculture impossible (typically 10-30 degrees), and as we saw in the Food and Agriculture Organization Report, some regions will actually see yield benefits from climate change because they are currently too cold to support agriculture during many months of the year.

Summing up, most of the world is unlikely to starve due to climate change in this century. The poorest among us may well experience increased food insecurity, especially if effective altruists continue to divert resources from anti-poverty efforts towards existential risk mitigation. Some workers will literally die in the fields from heat stress. All of this will be a great, and preventable catastrophe. But it will not be an existential catastrophe.

4. Heat stress

As the world grows warmer, humans will be exposed to worrying amounts of heat stress.

Here are the safe wet-bulb globe temperature (WBGT) limits on activities by acclimatized people, as summarized by Halstead.

ActivityWBGT limit for acclimatized people
Intense physical activity26

At present, WBGT around the globe is usually comfortably within these limits, even during the hottest months of the year. Here is a map of average WBGT from 1981-2010 during the hottest month in each local climate (Kjellstrom et al. 2017).

On the most realistic scenarios, capped at 2-3 degrees of warming by 2100, heat stress will be bad but manageable. Tens, but perhaps not hundreds of millions will be exposed to temperatures making it impossible to work outdoors, or even to survive without air conditioning during the hottest months (Quiggin et al. 2021):

On less plausible, but more extreme scenarios, heat stress becomes fairly dire. Here is a map of average WBGT on a range of extreme climate scenarios (King et al. 2015):

How should we react to these scenarios? The first thing we should do is cry. Should any of these scenarios come about, there will be sharp increase in heat-related deaths, sustained outdoor labor will become infeasible during the hottest months, crop yields will diminish, and some populations may be forced to migrate to cooler areas.

However, all of this stops well short of an existential catastrophe. Even in the most implausible scenarios, there are large land masses that will remain temperate, even cool during the hottest months of the year. Technologies such as air conditioning can, and even today have begun to make warmer countries inhabitable during hot summers. Migration, while a great human tragedy, need not lead to extinction.

Furthermore, it is important to remember that the world currently suffers from greater numbers of cold-related deaths than heat-related deaths. Here, again, is Halstead:

For the Global Burden of Disease study, Zhao et al (2021) find that between 2000 and 2019, on average there were around 500,000 heat-related deaths and around 4.5 million cold-related deaths each year. The literature suggests that the benefits of climate change in reducing cold-related deaths in temperate regions will be outweighed by the increase in heat-related deaths in warmer regions.

On the most extreme (but unlikely) warming scenarios, that trend could slightly reverse. But just as an average of 4.5 million annual cold-related deaths, while a great human tragedy, does not amount to an existential catastrophe, so too most projections for future heat-related deaths, while appalling, are better classified as catastrophic risks than as existential risks.

5. Looking ahead

So far, we have looked at two impacts of climate change: agricultural shifts and heat stress. We saw that on the most likely, moderate scenarios for climate change, effects will be manageable and could even be outweighed by benefits of technological innovation or climate change. We also saw that even on the most extreme scenarios, there is no credible pathway from heat stress or agricultural change to irreversible existential catastrophe by 2100.

There is, I am afraid, much more to come, for there are many other ways in which climate change will threaten humanity in this century. I will look at some other scenarios in the next post of this series. In the meantime, are there other reasons for thinking that heat stress or agricultural change could lead to irreversible existential catastrophe by 2100? And are there other climate risks that we should be worried about? Let me know.


4 responses to “Exaggerating the risks (Part 3: Climate change, lessons from Halstead)”

  1. Ramiro Avatar

    Thanks for this series. Do you intend to review Kemp et al. (2022) (doi:10.1073/pnas.2108146119)? I guess the case for a c-risk is way better than for x-risk? After all, climate change seems to be correlated with many changes in human trajectory, at least since the neolithic revolution, and with extinctions in general.

    1. David Thorstad Avatar

      Thanks Ramiro! I’m a big fan of Kemp et al. I don’t plan to review it simply because (with a few exceptions) I have nothing but praise for it. I was very happy to see this come out in PNAS.

      I am quite sympathetic to the idea that climate change poses a catastrophic risk to humanity, and that as a result of this, we should do a great deal to address climate change. One of the most off-putting features of recent longtermist theorizing for many non-EAs is that the turn to existential risk (rather than catastrophic risk) has left climate change rather neglected. Even Ord, who I think inflates existential risk from climate change a great deal, still thinks that many other cause areas are far more important. And to the extent that many EA authors pay lip service to the idea that climate change is an important cause area, I think that this is often done as much for PR reasons as from a place of genuine belief.

      One of the themes of this blog will be that catastrophic risks are important, real and scientifically tractable. Once we see that existential risk has been exaggerated (and note other problems, for example with the valuation of existential risk reduction efforts), we can turn our efforts back to catastrophic risks, as well as back towards other tractable problems such as global health and poverty. I think that once we do this, fighting climate change will emerge as a central moral priority of our time.

      1. Ramiro Avatar

        Thanks. Now I’m even more eager to see the next chapters 😉

      2. yellowgreen1689 Avatar

        This seems wrong to me. The argument has to be that the tractability of reducing existential risks has an elasticity greater than 1 (i.e is elastic rather than inelastic) with respect to the level of existential risk, otherwise reductions in the level of existential risk make it more work working on. Lower climate and nuclear x-risk I think should increase our efforts in bio and ai.

Leave a Reply