Rethinking the Social Cost of Carbon

by Judith Curry
The Social Cost of Carbon is emerging as a major source of contention in the Trump Administration.

Andy Revkin has an article summarizing the issue:  Will Trump’s climate team accept any social cost of carbon?  Excerpts:
But there’s probably no more consequential and contentious a target for the incoming administration than an arcane metric called the “social cost of carbon.”
This value is the government’s best estimate of how much society gains over the long haul by cutting each ton of the heat-trapping carbon-dioxide emissions scientists have linked to global warming.
The contention arises because the social cost of carbon underpins justifications for policies dealing with everything from power plants to car mileage to refrigerator efficiency. The carbon valuation has already helped shape 79 regulations.
The strongest sign of a coming challenge to the social cost calculation came in a post-election memorandum from Thomas Pyle, who was then president of the industry-funded American Energy Alliance and Institute for Energy Research and who now leads the Trump transition team for the Department of Energy. In the memo, he predicted policies resulting in “ending the use of the social cost of carbon in federal rule makings.”
In 2013, an economist from Pyle’s energy institute testified in a Senate hearing that under a proper calculation, the social cost of carbon “would probably be close to zero, or possibly even negative.”
A deep cut would be both dangerous and unjustified, given the basics of both climate science and economics, said Gernot Wagner, a Harvard economist focused on climate risk and policy.
Some illumination on the controversies and uncertainties is provided by a new draft report from the National Academies of Sciences Valuing Climate Damages:
Updating Estimation of the Social Cost of Carbon Dioxide.
The NAS Report is a fascinating read, and describes in detail all of the sausage making that goes into evaluating the Social Cost of Carbon (SCC).  Andy Revkin provides the following summary:
The main recommendation is to “unbundle” the mix of models behind that seemingly simple dollar figure. The models, melding climate science, demographic change and economics, project harms by looking at possible shifts in human populations, technologies, economies and the climate in coming decades.
The assumptions and uncertainties for each step could then be more clearly laid out in transparent ways that might constrain misinterpretations and boost societal, and political, acceptance.
Myles R. Allen, an author of the report and a climate scientist at Oxford University, said in an interview that such a structure could help clarify where data ends and societal and political choices begin.
“There are obviously political decisions which need to be made in any calculation like the social cost of carbon,” he said. “On the other hand, the way the climate system responds to greenhouse gas emission levels is not really up for political discussion.”
I’m not going to make any attempt to summarize the report here; it is valuable for highlighting many areas of uncertainties.  My comments on this subject are based on stepping outside of the SCC frame of the NAS report.
Scope and magnitude of the uncertainties
After reading the NAS report, I was overwhelmed by the HUGE uncertainties associated with the estimates of the Social Cost of Carbon.  However, these HUGE uncertainties are overwhelmed by MASSIVE meta-uncertainties not even considered in the NAS report.
For purposes of discussing these uncertainties, I will adopt the terminology Risbey and Kandlikar (2007):

The physical climate system is regarded as the least uncertain of all the various modules in the SCC determination (which I agree with). However, with regards to the physical climate module proposed in the NAS report, there are the following meta-uncertainties that are not acknowledged in the report:

  • Sensitivity to doubling CO2 remains highly uncertain, much more uncertain even than the IPCC AR5 (which could not determine a median or ‘best’ value). Recent assessments that aerosol forcing is much less than assumed by the AR5 and refinements to observational methods for determining climate sensitivity considerably lower the estimates of climate sensitivity, beyond what is considered in the NAS report (e.g. Nic Lewis’ work) RK score: 2.5
  • The carbon cycle (both land and ocean) is poorly understood in quantitative terms, so it is not a simple task to translate emissions into atmospheric CO2 concentrations with any kind of accuracy or confidence RK score: 3
  • The other requirement is for additional climatic variables related to damage, such as extreme weather events, regional changes, and sea level rise — variables that the climate models currently do not predict with any accuracy at all. RK score: 3

These meta uncertainties in our understanding of the physical climate system do not even allow us to bound these, let alone produce a pdf that makes any kind of sense.
With regards to the Damages Module, this one might be the most uncertain of all (although it is rather a toss up with the Socioeconomic Module and the Discounting Module).  Apart from the obvious demographic and socioeconomic uncertainties:
Assessing damages from CO2 requires three steps:  1) determine that warming is ‘bad’ (costly, dangerous, whatever) RK score: 5; 2) link the warming to extreme weather events or other factors that are associated with costs RK score: 5; and 3) attribute the warming to CO2 RK score: 2.8.
With regards to attributing the warming to CO2, that is an unresolved problem in my opinion.  If only 51% of the recent warming is caused by humans (which is technically within the scope of the IPCC AR5 attribution statement), that would halve any benefits of attempting to reduce warming via eliminating CO2 emissions.  And given what we don’t know about natural climate variability, there is plenty of scope for human causes to have contributed less than 50%. Climate models effectively find that  100% of the recent warming is caused by humans, which is the implicit assumption going into the SCC estimates.
Linking extreme weather, sea level rise, etc to human-caused CO2 emissions has remained elusive, as per the IPCC AR5 WG2 report [link].  Whether or not sea level rise is accelerating remains a subject of debate (I have had many previous posts on this).  It is very difficult to attribute any extreme weather events to warming, let alone human caused warming.
And finally, it is not at all clear to me that on balance (with all the complexities, feedbacks, adaptations, etc.), that a warm climate is more costly than a cold one, on the timescale of the 21st century.  I am not convinced by the arguments that I’ve seen.
The Socioeconomic Module requires making projections of population, technologies, economic development, etc., 300 years into the future! RK score: 4.5 Voodoo.
And then we have the Discount Rate RK score: 5.  To me, the whole concept of discounting over several hundred years makes no sense.  What you assume for discounting can change the sign of the SCC outcome.  This tells me that the whole exercise does not rest on a  robust foundation.
So, does this SCC exercise make any sense?  Well, I think that it is an interesting thing to pursue academically — all these interactions are rather fascinating.
However, these huge to massive uncertainties render SCC as largely useless for cost/benefit  analyses, since the number and magnitude of these uncertainties violate the premises behind cost/benefit analysis:  uncertainty should be well characterized (RK score 1 or 2), and model structure should be well known.  You can disagree with my RK scores above, but it is very difficult to argue any of them into 1 to 2 territory.
A paper by Andrea Saltelli et al. also makes this point: Climate Models as Economic Guides: Scientific Challenge or Quixotic Quest?. Their concluding statement:
The uncertainties associated with mathematical models that assess the costs and benefits of climate change policy options are unknowable. Such models can be valuable guides to scientific inquiry, but they should not be used to guide climate policy decisions.
See also the paper by Robert Pindyck: Climate Change Policy:  What Do the Models Tell Us?  Short answer:  “Very little.
As a basis for policy, I will argue that the MASSIVE meta-uncertainties (or even just the HUGE uncertainties identified by the NAS Panel) put this whole topic well into the territory of  ‘deep uncertainty’.  I’ve written and spoken about deep uncertainty many times [link].
Deep uncertainty is characterized by situations in which:

  • phenomena are characterized by high levels of ignorance and are poorly understood scientifically
  • modelling and subjective judgments must substitute extensively for estimates based upon experience with actual events and outcomes
  • ethical rules must be formulated to substitute for risk-based decisions.

Different decision making frameworks are more useful than cost/benefit analysis once you are in deep uncertainty territory.
And there is little to no evidence that climate change is a ‘ruin’ problem on the timescale of the 21st century [link].
Where do we go from here?  
Since I’ve argued that the cost/benefit approach doesn’t really make sense for such a wicked problem with massive uncertainties, does this mean I think we should ignore the problem?
NO, we should not ignore the problem, but we should reframe it in ways that put some realistic bounds on what we are dealing with– not just climate change, but also population increase and concentration of wealth in vulnerable coastal regions.  Not to mention the growing needs of this increasing population for energy, water and food.
I would challenge the policy making community and the science-policy interface communities to consider the following questions and proposed analyses:
1.How many different combinations of assumptions in the SCC models can produce a SCC value that is not significantly different from zero, or within some ‘tolerable’ limit?
2.Imagine the worst plausible future climate outcome on the time scale of the 21st century (consistent with the AR5), and estimate the damages in the 21st century.  Assess whether any conceivable path of CO2 emissions reductions in the 21st century would make a significant dent in those damages.
3.Estimate the costs of extreme weather in the 21st century, based on weather statistics from the 20th century while accounting for 21st century changes in population, demographic, property, GDP, etc.   Then compare the impact of socioeconomic changes on 21st century costs relative to the hypothetical delta of extreme weather events as derived from climate models.  I wouldn’t be surprised if population increase and concentration of wealth in coastal regions is a much bigger factor here than climate change.
4.Estimate the costs of sea level rise in the 21st century based on three different assumptions: 1) applying the average sea level rise rate over the 20th century; 2) applying the average rate of sea level rise for the past 50 years for each coastal location, which also includes land use, geologic factors, groundwater withdrawal, etc.; 3) apply the average rate of sea level rise from the IPCC AR5.  I suspect that #2 will be associated with the most damage, since the most vulnerable locations have local sea level rise rates that far exceed anything that can be explained by warming.  Apart from geologic and land use effects on sea level rise, the increase of population and concentration of wealth in coastal regions may also be a bigger factor than sea level rise associated with warming.
5. Estimate regional per capita water needs (globally), using population and socioeconomic projections for the 21st century.  Compare that with 20th century water availability (total and per capita). Estimate the per capita water availability in the 21st century using climate models.  Is the decline in 21st century per capita water availability caused by population increase or climate change?  (Hint: whether or not you find them convincing, climate models predict overall MORE rainfall in a warmer climate; melting glaciers will help at least in the short term.) Assess the costs of meeting per capita water needs using 20th century rainfall versus 21st century projected rainfall.
6.Based on estimates from #3, #4, #5, decide on how much resilience we can afford, in terms of infrastructure, and work on other clever ways to reduce your vulnerability through land use policies, advance warning of severe weather, etc.
This list is by no means exhaustive; once you think about reframing the climate problem and the solutions, lots of new ideas pop up. Such analyses would provide the basis for a pragmatic climate policy that puts people first in the 21st century, which is a reasonable thing to do given the deep uncertainties surrounding the wicked climate change problem.  Any rationale that supports rapid reductions of CO2 emissions needs to provide pathways for improved technologies for energy, transportation, agriculture, etc.  Not to mention supporting human development in regions that currently do not have access to grid electricity.
The bottom line is:  water, food, energy.  Heck, even the folks attending Davos get it [link]. People need it and large numbers of people want more of it.  And there are more and more people all the time.  A single minded focus on reducing CO2 emissions neglects a lot of real problems facing many nations across the globe.
Climate variability and change impacts water, food and energy.  But there isn’t much we can do to influence the climate on the timescale of the 21st century — however much we have impacted the climate over the past 70 years or so, those impacts (large or small) will work their way through climate system over the next centuries as the oceans act as a big flywheel on the climate system.
Back to the question posed by Revkin: Will Trump’s climate team accept any social cost of carbon? Well, I hope not.  Here’s to hoping for a more pragmatic approach to all this in the Trump administration.
 Filed under: Economics, Policy, Sensitivity & feedbacks

Source