A new report out by an international consortium of scientists and policymakers makes the somewhat surprising claim that climate change is already killing about 400,000 people a year, through a combination of its effects on disease and hunger. This certainly sounds like a big number, and would account for almost 1% of the roughly 50 million people that die around the world each year.
Reading up on the report's methodology a bit, it looks like they arrive at this number by thinking of all the things that can kill you, assembling estimates from the literature of how effective these things are in doing so, and then combining these with estimates of how much these things might be affected by climate. They do this at the country or grid level and then add everything up to get a global estimate.
This approach at getting at climate impacts is not unlike that taken by Integrated Assessment Models in the broader climate impact debate: you try to directly account for all the stuff that might be going on that affects your outcome of interest. The appeal is that you can get really down and dirty in the details and try to model everything that is going on. But this complexity is also a source of difficulty, since we don't always have good estimates of how climate affects each intermediate variable, and it's hard to know how to add things up without under- or over-counting (e.g. is a person who dies from malaria the same person who's going to die from hunger?).
You could imagine more "reduced form" ways at getting at the same thing: e.g. a study that looked at the direct effect of past climate variation on mortality. The disadvantage with this approach is that it is hard to illuminate the intermediate variables that link climate and mortality, but the advantage is that you avoid the adding-up problem of the more "integrated" approach. Put simply, your headline impact estimate might be more trustworthy, but you're probably going to understand it less.
I don't know of a good paper that has directly looked at the relationship between climate and overall mortality (although a working paper by Kudamatsu et al is a step in that direction), but there is a nice paper from Baird, Friedman, and Schady (undated version here) that tells us how much infant mortality goes up when aggregate economic productivity declines in poor countries. We can combine this with another nice recent estimate from Dell, Jones, and Olken (ungated here) of how aggregate economic productivity responds to temperature change to derive an estimate of the effect of climate on infant mortality via economic productivity. Economists are perhaps overly-comfortable assuming that aggregate economic measures such as GDP capture most of what we care about in terms of a country's well-being, but it's not a bad place to start.
So, some simple math:
- Dell Jones Olken estimate that a 1C increase in temperature reduced GDP growth rates in poor countries by 1.3 percentage points.
- Baird, Friedman, and Schady estimate that a 1% reduction in GDP increases infant mortality rate in poor countries between 0.24 and 0.4 deaths per 1000 births (call this 0.32 per 1000)
- IPCC AR4 estimates average surface warming of 0.13C per decade since 1950 - so 0.013C per year.
- The UN estimates that there have been 6.4 billion births in less developed countries between 1950-2010. Infant mortality rates over the period average 91 per 1000 births in these regions.
Combine these and (if I did the math right) you get an estimate that warming since 1950 has resulted in an extra 2 million infant deaths in the developing world. This gives you a per-year mortality estimate of a little less than 40,000, or about an order of magnitude smaller than the estimate in the new report.
What's going on here? Some possible explanations:
1. The lion's share of climate's effects on mortality are not captured in climate's effects on GDP or its correlates
2. The reduced form estimate only considers effects on infant mortality, and perhaps a lot of adults are dying and driving up the "integrated" estimate.
3. There is an adding-up problem in the 400,000 number.
Some part of (1) is undoubtedly true, but personally I find it hard to believe that GDP would not pick up a lot of climate's effect on a country's mortality outcomes - at least more than 10%. (2) seems unlikely, and the report notes that most of the 400,000 deaths are children. So I'm worried at the culprit might be in part (3): it's just hard to know how climate impacts every possible way people die, and it's maybe even harder to add up these impacts.
It would be nice if someone wanted to repeat the Baird et al paper but just run climate as the independent variable instead of growth, which would give us a direct estimate of climate on infant mortality. Until that time, while it seems almost certain that climate change is already killing people and will kill a lot more in the future, we might be a little skeptical that the effects are already anything like 400,000 climate deaths per year.