Is global warming going to be very costly? I certainly think so. But the key premise of Jim Manzi's argument against tackling carbon emissions (see his initial post, my reply, and his reply) is that it won't do too much damage. Indeed, Manzi's reading of the IPCC leads him to contend that future temperatures rises can be expected to reduce global GDP by only about 3 percent in 2100, while curbing carbon pollution will cost more than that.

But how confident should we be in those numbers? The blogger over at Things Break makes the case that this 3 percent figure is dubious:

1. There is only one SRES scenario that reasonably tracks real world emissions growth per observations and infrastructure legacy: A1FI.

2. The “likely” temperature range for A1FI is 2.4-6.4°C.

3. The high end cost of ~6°C warming in Manzi’s source is upwards of 11% global GDP, yielding a range/estimate of 1-11%/5.5% GDP, not 1-5%/3%.

4. Manzi cites a third party estimate of mitigation costs as ~6% GDP for stabilization at 450 ppm, while other analyses by experts in the field put the cost far lower- e.g. 2.5% at 350-400 ppm.

5. Being generous and more than splitting the difference- rather than using my own preferred analysis outright- gives a cost of 4.25% GDP for 450 ppm, compared to a non-stabilization cost of 5.5%GDP.

In short Manzi’s analysis depends on a suspiciously narrow reading and selection of source material that doesn’t hold up to even a cursory amount of scrutiny.

Further points: This simple debunking has ignored other problematic assumptions implicit in Manzi’s analysis—the absurdly conservative damages of the DICE model, for example. Notoriously, DICE shows a warming of 20°C resulting in a loss of only 50% of global GDP, a warming so extreme that it would exceed humans’ (and other mammals’) capacity to endure heat stress, resulting in mass extinction. It’s fairly safe to say that there wouldn’t even be a “Gross Domestic Product” much less one half as large as today’s. Ocean acidification, non-linear ice sheet collapse and resulting sea level rise, and other costly consequences of unchecked emissions are likewise ignored by DICE.

That last sentence, in particular, partly gets at why I'm leery of relying too heavily on numerical cost-benefit analyses. Are we really sure we comprehend the costs of acidifying the oceans—and risking the mass disruption of marine ecosystems? How confident are we that we can put a price on the consequences of destabilizing the ice sheets? (The answer to the latter is: Not very. As I noted earlier, even the IPCC admits it doesn't yet have a good handle on how the ice sheets will respond to higher temperatures, though James Hansen's research has suggested that, in prehistoric times, ice sheets have been known to destabilize very quickly, and once they get going, sea-level rises of a few meters per century can happen quite easily. None of the cost figures mentioned above account for this.)

On the flip side, we do have a pretty good sense of what will happen if we start putting a price on carbon emissions: Yes, taxes might be a drag on the global economy, but that's a phenomenon that's pretty well understood, and taxes and regulations can always be modified if they turn out to be unduly harmful. But we don't have nearly as good an understanding of what natural disasters await as we run this massive greenhouse experiment, and if things go horribly wrong, there's no going back.