I focused my critique on the disaster impacts because it seems common wisdom that the attribution of malaria, diarrhea and malnutrition deaths to greenhouse gas emissions needs no further critique. I was wrong about the common wisdom.
Roger finally admits he was only talking about less than 5% of the deaths. As to his claim that he omitted critiquing the 95% of the study because he was being nice, decide for yourself if you think he's telling the truth.
The attribution of health impacts to greenhouse gas emissions relies on work done by the WHO. Here is how the WHO describes its own analyses:
In 2002 the WHO explained (at p. 26 in this PDF):And by quantitative modeling, they mean a model with assumptions included for the effects of GHG-driven climate change. To be clear, these assumptions are not based on observations or measurements.
Climate exhibits natural variability, and its effects on health are mediated by many other determinants. There are currently insufficient high-quality, long-term data on climate-sensitive diseases to provide a direct measurement of the health impact of anthropogenic climate change, particularly in the most vulnerable populations. Quantitative modelling is therefore the only practical route for estimating the health impacts of climate change.
So we're back, yet again, to Roger's shibboleth, his saying you can't start with an estimate of known climate impacts and use that to attribute a percentage of deaths and economic costs. Why? He says it's hard to do it accurately, that a slight change in the assumptions will give large changes in the numbers.
What he calls nonempirical is what normal humans would call an estimate. If you can't estimate perfectly, then you try to do it conservatively, and make clear that you're being conservative.
Roger thinks that if you can't get a perfect number, then don't do it at all, and give polluters a free ride instead.
UPDATE: Roger says in his post that tweaking the assumptions could result in an analysis where climate change actually reduces deaths instead of increasing them. That sounds to me like a great way for Roger to falsify this method - show that different but equally-reasonable assumptions lead to radically different conclusions. Incidentally, that's also how the fossil fuel companies could have backed up their assertion that climate models are useless - they could build their own with defensible parameters and assumptions and get different outputs. The fact that they never did it (publicly, anyway) suggests to me that any defensible model leads to a similar conclusion.
UPDATE 2: See William's post - he thinks the study's flawed, and makes the good point that my attack on Roger's critique sounds like a defense of the study. It's not meant to be.