Recommendations for Improving the Treatment of Risk and Uncertainty in Economic Estimates of Climate Impacts in the Sixth Intergovernmental Panel on Climate Change Assessment Report?

DSC_7504 A slightly odd paper in Review of Environmental Economics and Policy by Thomas Stoerk  Gernot Wagner  Robert E T Ward, via the Grantham Institute at LSE via Twatter. Abstract:
Large discrepancies persist between projections of the physical impacts of climate change and economic damage estimates. These discrepancies increase with increasing global average temperature projections. Based on this observation, we recommend that in its Sixth Assessment Report (AR6), the Intergovernmental Panel on Climate Change (IPCC) improve its approach to the management of the uncertainties inherent in climate policy decisions. In particular, we suggest that the IPCC (1) strengthen its focus on applications of decision making under risk, uncertainty, and outright ambiguity and (2) estimate how the uncertainty itself affects its economic and financial cost estimates of climate damage and, ultimately, the optimal price for each ton of carbon dioxide released. Our hope is that by adopting these recommendations, AR6 will be able to resolve some of the documented inconsistencies in estimates of the physical and economic impacts of climate change and more effectively fulfill the IPCC’s mission to provide policymakers with a robust and rigorous approach for assessing the potential future risks of climate change.
(my bold). Obviously, a "projection of physical impacts" and an "economic damage estimate" are two different things. For there to be a discrepancy you'd need, well, something obvious would be 10 m of SLR and very little economic damage. Obviously no paper making such a strong assertion would fail to support it, so let's go look.

In the Inconsistent Assessment of Risks in the IPCC Fifth Assessment Report (AR5) section they kinda make three arguments: (1) There is now a growing awareness of the limitations of the existing generation of IAMs; (2) these estimates largely ignore the potential for “tipping points”; and suggest (3) strengthening its focus on decision making under uncertainty and outright ambiguity. (1) isn't clearly supported by refs; (2) gets you the standard answer; and (3) isn't clear in what way that represents a discrepancy.

But never mind that, because the next section is called Evidence of Discrepancies Between Scientific and Economic Assessments of Climate Impacts; it's just bound to be in there. And the relevant text appears to be "DICE, for example, projects a loss of less than 10 percent of global economic output as a result of raising the global mean surface temperature by 6°C". But while that might be a discrepancy, it isn't obviously. And their reference for it is "Climate Shlock", a book wot one of 'em wrote. That's not a scholarly reference. And as apparently the key reference to their key point, it's just a little careless. They continue physical impacts are often not translated into monetary terms and they have largely been ignored by climate economists which is odd. The IAMs don't include everything, but to express it in this way is strange. Is this Pindyck again?

I don't think I'm trying to defend IAMs at this point. Just point out that this paper seems rather ill considered. A case for tighter refereeing I think.


The rich want to live in the Anglosphere.
What’s the damage (of that climate change cost-benefit model)? - ftalphaville


Everett F Sargent said...

Off topic, but ...

When I hoover (sic hover) over a timestamp for anyone's post I don't see a date associated with a post (the time of day is obviously there on all posts).

I'm kind of thinking that I might not want to post a reply to someone else, if say, their post was from a long time ago (e. g. months or years).

Your articles have a time/date stamp (Zulu). But, at my end at least, all posts/replies just show the time of day (no date).

Anonymous said...

"And the relevant text appears to be "DICE, for example, projects a loss of less than 10 percent of global economic output as a result of raising the global mean surface temperature by 6°C". But while that might be a discrepancy, it isn't obviously."

You're saying that the DICE outcome isn't obviously wrong?:

"There was an even closer fit at the end of the Permian, 251m years ago, when global temperatures rose by – yes – six degrees, and 95% of species were wiped out.

That episode was the worst ever endured by life on Earth, the closest the planet has come to ending up a dead and desolate rock in space.” On land, the only winners were fungi that flourished on dying trees and shrubs. At sea there were only losers. Warm water is a killer. Less oxygen can dissolve, so conditions become stagnant and anoxic. Oxygen-breathing water-dwellers – all the higher forms of life from plankton to sharks – face suffocation. Warm water also expands, and sea levels rose by 20 metres.” The resulting “super-hurricanes” hitting the coasts would have triggered flash floods that no living thing could have survived."


William M. Connolley said...

EFS: that turns out to be an option. I've now set it to full, so should be good.

Anon: I was going to answer that 10% for 6 oC seems on the low side, but then I realised I didn't even know what it meant.

Time is now. Economic output is X. We get to 2100, at which point CO2 has increased, so T has gone up +6 oC, and economic output is now Y.

Does he mean:

(a) DICE predicts that Y = Y_a = X * (1 - 0.1).
(b) DICE predicts that Y = Y_b = Z * (1 - 0.1), where Y is (significantly) larger than X, and is the economic output at 2100 that would have occurred if T had not increased?

I think he probably means (b) but I'm not sure.

David B. Benson said...

End Permian event was caused by the eruptions of the Siberian Trapps via penitrations through massive coal seams. The resulting explosions threw megatonnes of nickel into the atmosphere. The nickel enabled a methanogen to convert the shallow ocean to an anoxic state.

A unique event.

crandles said...

If we have to rebuild cities further inland, would this create extra GDP with all that building but it is all just to keep facilities we would have had without SLR. Is this a big loss but GDP increases? How is this considered by models?

PaulS said...


That would seem to cast use of GDP as a "broken window fallacy", which I don't think is correct since it inherently does take into account opportunity costs in the long run. Yes, the replacement building activity would constitute extra GDP, but so would whatever else those resources would otherwise have been used for. And presumably replacing existing facilities results in little change in economic capacity, whereas the other thing which could otherwise have been built using those resources may have increased economic capacity, allowing greater future GDP growth. In other words, spending money on new things is economically preferable to spending money replacing broken things.

Loss of facilities before replacements can be built should produce a fall in GDP due to the loss of economic activity, but if the replacement pops up ready to go as soon as it's needed there would be no direct loss. Viewing the facilities themselves purely as means to generate economic production there is no direct loss from abandoning one building and moving to a new one. What is lost from having to use resources to build a replacement is the opportunity to do something more productive with those resources. That loss is (or should be) reflected in GDP.

At least, that's my understanding of things.

crandles said...

Thanks PaulS, I think that makes a lot of sense.


6C would presumably turn large area into poor scrub land which wouldn't support very much, likely causing lots of migration and/or deaths. Not sure to what extent irrigation, air conditioning and other innovation and development etc counters such effects but not having to is surely better. Devoting resources to such things as well as moving cities which might not otherwise be necessary and could be used for other things does seem likely to have significant effect on GDP.

William M. Connolley said...

> 6C would presumably turn large area into poor scrub land...

There's a connection there but I think it's largely backwards: lack of rainfall leads to particularly large temperature increases, and lack of rainfall causes land degradation. Rainforests are hot but wet. Being hot, in and of itself, tends to increased productivity (including of weeds and pests).

So (as I've said before) *probably* (but not certainly) the largest effects on food production will come from shifts in rainfall patterns. Which are harder to predict than temperature changes; but we can certainly predict that changes will occur.

Anonymous said...

WMC: "(b) DICE predicts that Y = Y_b = Z * (1 - 0.1), where Y is (significantly) larger than X, and is the economic output at 2100 that would have occurred if T had not increased?" is the general way that IAMs calculate damages. I.e., we expect GDP at the end of the century to be, say, 5 times larger than GDP today, so a "10%" damage from 6 degrees of warming means that GDP will instead be only 4.5 larger than today...

I think if IAMs started adopting Ramsey-type discounting, and then included large uncertainty in terms of future damages resulting from a given warming, we'd see the net-present-value impacts of 6 degrees warming increase substantially. This is because in some small fractions of futures, 6 degrees warming is truly cataclysmic, reducing GDP to below today's levels... by the magic of Ramsey discounting (in which the discount rate is a function of the rate of GDP growth), those futures will be weighted much more heavily than those futures in which GDP keeps growing at a fast clip.

DICE in particular tends to assume that damages (as a percent of GDP) grow as the square of temperature*. Weitzman et al. suggests that damages should actually be a multi-polynomial function, such that for temperature changes of 1-3 degrees, damages grow as the square, but that a second polynomial (say, to the 6th power) will eventually take over, such that for temperatures above 4 or 5 degrees, damages grow much more quickly (https://scholar.harvard.edu/files/weitzman/files/ghgtargetsinsuranceagainst.pdf).

Another approach is to assume not only that warming causes damages at any given time period, but that damages in one time period slow growth to the next time period: I don't have a good reference, but this tends to increase the net impact of warming on GDP.


*https://onlinelibrary.wiley.com/doi/pdf/10.1111/ecoj.12188 notes that, "The DICE default is quadratic and it is well known that with the standard values of the functions’ coefficients an implausible 18°C or so of warming is required in order to reduce global output by 50%."

crandles said...

>"There's a connection there but I think it's largely backwards: lack of rainfall leads to particularly large temperature increases, and lack of rainfall causes land degradation. Rainforests are hot but wet."

Yes re rainfall distribution not the temp change itself. I was thinking there was suggestion that cc could cause amazon to convert from rainforest to scrub land as rainforest becomes unable to cope with frequent droughts.

"Perz et al. (2008) examine the dynamics surrounding planned and unplanned road expansion and land-use change. Nepstad et al. (2008) provide an overview of the dynamics where the intensive use of fire in preparing land for agriculture and the leakage of this fire frequently degrades any surrounding remaining forests, potentially pushing the whole region through an eco-climatic ‘tipping point’ into a more degraded scrub system."

seems to suggest it is more fire and road expansion related than just climate change.

Still if expecting more floods and more droughts from temperature rise, does that lead to more scrub land? forest -> scrub re droughts and also dryish (not flat) areas -> scrub as intense rain washes away topsoil.

Phil said...

Standard answer?


I probably should link in some snowball Earth stuff as well. Almost for sure not in our future, as long as we have burnt enough fossil fuels. Like we did before... oh about 1930. Having carbon release subsidies once made sense... Too bad we can't get rid of them.

William M. Connolley said...

Incidentally, I said stuff abut 6 oC some while ago. That also talks about long tails.