FYI, Michael Greenstone (one of the authors of this study, the co-director of the Climate Impact Lab, and the Milton Friedman Distinguished Service Professor in Economics at University of Chicago) testified at a hearing in Congress on the health impacts of climate change a few weeks ago: https://youtu.be/N8nCZC0_yxU His opening statement is available in written form here: http://www.impactlab.org/news-insights/michael-greenstone-testifies-on-the-health-impacts-of-climate-change/
This is a linkpost for the article Global death rate from rising temperatures projected to surpass the current death rate of all infectious diseases combined, published by Climate Impacts Lab, based on this study.
Bill Gates summarised this as follows:
In other words, by 2060, climate change could be just as deadly as COVID-19, and by 2100 it could be five times as deadly.
From the summary article:
New study from the Climate Impact Lab finds people in poor parts of the world are disproportionately vulnerable to the risk of death associated with increased heat.
...the study projects that climate change’s effect on temperatures could raise global mortality rates by 73 deaths per 100,000 people in 2100 under a continued high emissions scenario, compared to a world with no warming. That level is roughly equal to the current death rate for all infectious diseases—including tuberculosis, HIV/AIDS, malaria, dengue, yellow fever, and diseases transmitted by ticks, mosquitos, and parasites—combined (approximately 74 deaths per 100,000 globally).
As far as I'm aware, this modelling also excludes potential exacerbating effects that climate change has on infectious diseases, e.g through the expansion of tropical regions leading to vastly more deaths from malaria.
This latest modelling could impact conclusions on the relative importance of climate change, e.g. examined here, though some estimates indicate that work to mitigate climate change could be more effective than many global health interventions, depending on our assumptions.
Thanks for the link, this is interesting!
A couple of quick thoughts from glancing at the paper (more hopefully later):
1) It is a first draft / working paper, not peer-reviewed (though the website frames it as if it was a finished study).
2) Some of the central assumptions seem quite selected for finding maximum impact, e.g. the 73 death per 100,000 people comes from a model run with RCP 8.5 which is currently seen as an extreme case, not business as usual but considerably worse, https://thebreakthrough.org/issues/energy/3c-world). The combination with SSP3 as the socio-economic scenario also seems to point in the direction of worst case assumption as this is a scenario of low/difficult adaptation, https://www.sciencedirect.com/science/article/pii/S0959378016300838 ). So, yes, 73 deaths per 100,000 from heat is possible but it is probably in the top 5% of the distribution of worst outcomes based on what we now think is realistic.
3) Something else that made me a bit worried about the bias in the direction of finding high impact was this statement "This projection accounts for adaptations to climate that populations are likely to make, given historical patterns of adaptation." One of the key features of expectable changes in the world to 2100, especially in high emissions scenarios, is that currently poor countries get a fair deal richer and use a lot more energy (in RCP 8.5 we could all burn lots of coal). So, it seems that the accounting for adaptation seems minimal compared to what one might expect, more adaptation beyond historical patterns as those countries get richer and have more resources available.
4) I am not saying it is a bad study and I am not really qualified to assess that on the deeper details, but I often find a lot of the climate-impact related literature to work with assumptions that seem very focused on finding maximal impacts (or with article titles exaggerating what the paper actually says, even in journals like Nature/Science etc.), a kind of publication bias that has made me quite skeptical of any single study.
1) This hasn't been through peer review yet, but it's a project they've been working on for years, and this is at least the third iteration (first two iterations: https://epic.uchicago.edu/wp-content/uploads/2019/07/Working-Paper-2.pdf https://bfi.uchicago.edu/wp-content/uploads/BFI_WP_201851.pdf). They've presented this paper at many academic conferences where they get criticism and feedback from other experts (including one I've been to). Unfortunately, publication timelines are very long in economics, so NBER working papers are where much of the new interesting research is presented. By the time a paper ultimately comes out in an economics journal, it's oftentimes old news. I'd expect this to publish in a good journal, and I would be surprised if the results change much.
2) They run projections for different emissions scenarios. Figure 9 shows RCP 8.5 and RCP 4.5. You can also go on their website and use their Impact Map to look at different scenarios: http://www.impactlab.org/map/#usmeas=absolute&usyear=2080-2099&gmeas=change-from-hist&gyear=2020-2039&tab=global&gvar=mortality
3) They account for the effects of both income growth and having experience living in a certain climate (what they call adaptation) on mitigating the mortality risk from climate change. Income growth substantially mitigates the mortality effect of climate change as you can see in figure 9a. Their main result of a 73/100,000 increase in the mortality rate in 2100 in RCP 8.5 accounts for the mitigating effect from income. Without higher incomes, it would be ~200/100,000 as figure 9a shows.
4) They use a reduced-form econometric strategy that exploits historical variations in temperatures on an annual basis to find a relationship between mortality and temperatures. Their approach allows them to account for climate-mortality effects that are driven by direct changes in the short-run distribution of temperatures such as the net mortality effect of more hot days and fewer cold days, the mortality effect of increased surface ozone formation, and even the effect of hot days on murders and suicides. However, their approach arguably does not fully capture climate-mortality channels that are driven in part by longer-term pathways that are not econometrically identified from shorter-term temperature fluctuations such as some diseases (like Malaria), flooding, and some causes of undernutrition.
2) I don't think this refutes Johannes point, which is that the headline figures claimed in the write-up on impact lab seem selected to get eye-catching figures. Although they run RCP4.5, they report the effects of RCP8.5 on the website and in the abstract. The mean effect is about a sixth smaller on RCP 4.5.
To put RCP8.5 in context, energy demand nearly quadruples, driven mainly by coal.
I do worry that this sort of work underestimates our ability to adapt. If energy demand does quadruple, there would be a lot more air conditioning to go round, and burning of coal would have driven a lot of income growth
3) From the copy I see, I think you are reporting Figure 7a, not 9a?
The copy you have is their 2019 version of the paper. The figure 9 I am referring to is their most recent 2020 NBER Working Paper version of the paper linked in the original post.
I agree that the RCPs, which were made in 2011, are outdated at this point. This is in large part because of the strong performance of renewable energy over the last decade. The RCPs at this point are still the standard emissions scenarios that are used in scientific papers, although I expect them to be updated in the near future when the next IPCC report comes out. Somewhere between RCP 6.0 (~3.2 degrees C in 2100) and RCP 8.5 (~4.8 degrees C) is probably what you can call a baseline emissions scenario (source: http://live.magicc.org/). The baseline emissions scenario in DICE-2016 for instance -- which involves significant reduction in emissions per unit of GDP, continued economic growth, and a small amount of carbon emissions abatement -- results in 4.1 degrees C warming in 2100. Still, the Climate Impact Lab results are pretty significant in any of these scenarios as you can see in figure 9a. In RCP 8.5, you get to 3.2 degrees C in about 2065 and you get to 4.1 degrees C around 2085. At both of those dates, there are significant increases in mortality. Just eyballing it, it looks like ~1/2 and ~4/5 of the mortality increase at 4.8 degrees C in 2100 respectively.
Also, if you wanted to look at the net mortality effect of a high emissions scenario with a lot of coal burning, you would also need to consider the effect of this on particulate matter pollution in addition to the affect from changing the climate. The particulate matter effect is very large and the climate impact paper does not account for this. This paper, for instance, does: https://www.nature.com/articles/s41467-019-09499-x/tables/2.
For the mitigating effect of income on mortality, I'd emphasize that there is significant uncertainty in these projections as shown in figure 9b, in large part driven by uncertainty around adaptation. For instance, with 80% confidence, they can't rule out the mortality effect of climate change being on net positive. Though from the perspective of decision theory, if you have a lottery across the possible outcomes in 2100 that include both sanguine and higher than expected damages and you have risk aversion (which most people do), this would cause you to want to undertake stricter climate policy than if you just look at the central estimate. I'd also emphasize that this is a very active area of research (e.g. this conference is happening next month: https://climateadaptationresearch.com/agenda/) so we will continue to get better at estimating climate impacts net of adaptation.
Very interesting, thanks!