How do the new estimates of transient climate response affect the social cost of CO2?

A new paper by Otto et al in Nature Geoscience has made an updated estimate of the transient climate response (TCR) implied by the climate change we have observed from 1970 to 2009. It finds the TCR to lie in the range 0.7–2.5 °C, with a best estimate of 1.4 °C. This is somewhat lower than an estimate based on temperatures just from the 1990s  which had a range of  0.9–3.1 °C, with a best estimate of 1.6 °C. The finding has been used by some commentators such as Matt Ridley to suggest that our policy on global warming is hopelessly misguided. But is this a reasonable conclusion?

The social cost of CO2 (SCCO2) measures the Net Present Value (NPV) of the extra damage caused by the emission of one more tonne of CO2 today.  So the SCCO2 represents the level at which we should set a tax on CO2 to internalise the impacts of climate change. So the relevant question for climate policy is, how does the new estimate of TCR affect the SCCO2?

The only way in which the  SCCO2 can be calculated is by the use of an integrated assessment model (IAM). Fortunately, I have one of the leading IAMs, PAGE09, in my possession, and I have used it with previously accepted estimates of the TCR to calculate that the mean value of the SCCO2 is about $100 per tonne of CO2.

What happens if we run PAGE09 with the updated estimate of TCR from Otto et al? The mean value of the SCCO2 goes down from about $100 to about $80 per tonne of CO2. So the price we should be charging energy users on each and every tonne of CO2 emissions is $80, about £50. As the UK’s carbon floor price is currently £16 per tonne, and that is only paid by the most energy intensive firms, we can indeed say that the new estimates of the TCR do show that our climate policy is misguided, but not in the direction that Matt Ridley believes; the new TCR estimates support a much tougher policy than the one we currently have in place.

Wonky details: The transient climate response is defined as the warming expected at the time when carbon dioxide in the atmosphere reaches double its original concentration, having increased at 1% per year, ie after 70 years. The default version of PAGE09 uses a triangular distribution for TCR with a minimum value of 1.0, a mean value of 1.3 and a maximum value of 2.8 to give a mean SCCO2 of $106 in 2010. The calculations reported here use a  triangular distribution for TCR with a minimum value of 0.7, a mean value of 1.4 and a maximum value of 2.5 and give a mean SCCO2 of $81 in 2010. All results are from 100,000 runs of the model which give a standard error of the mean SCCO2 of about $1.

Update added 10 June 2013: I have now had a chance to run PAGE09 with the TCR distribution (most likely value 1.3 °C with a 5-95% confidence interval of 0.9 to 2.0 °C) suggested in the comment below by Nic Lewis. It gives a mean SCCO2 value of $65 per tonne of CO2. Slightly lower, as one would expect, but still well above the UK’s carbon floor price.

Please consider leaving a comment or subscribing to the RSS feed to have future posts delivered to your feed reader.

8 Responses to How do the new estimates of transient climate response affect the social cost of CO2?

  1. Dr Hope
    Being one of the authors of the Nature Geoscience paper, I am interested in your PAGE09 results.

    However, you have not presented results using our paper’s best estimate of TCR based on observations of the most recent dedade, being 1.3 C with a 5-95% confidence interval of 0.9 to 2.0 C. Would you mind stating what the mean value of SCCO2 comes out as using that TCR estimmate?

    The Nature Geoscience paper presents arguments as to why the TCR estimate based on data for the 2000s should be the most reliable, rather than that based on the 1990s (a decade affected by the eruption of Mount Pinatubo in 1991 and then a once-in-a-century huge El Nino event in 1997-98) or the 1970-2009 period as a whole.

    FYI, the posterior probability density for the TCR estimate based on data from the 2000s is not far from being Gaussian, save for having a longer tail on the upper side.

    Thank you

    • Dear Nic,

      Thank you for your comment. You will see that I have done the runs you requested and added an update to the blog post.

  2. Dr Hope

    do you have a link to the latest description of your Page model and might be even to the source code? I am a climate modeller with a rising interest in economics.

    Best regards

    Georg Hoffmann

    PS My first impression of your ssco2 computation was that there are extremely sensitive to TCS changes by just 0.1C. Can that be understood easily and is this a common behaviour of all IAMs?

    • My policy is to make the model freely available on a case-by-case basis to those who wish to use it for research.

      Best wishes

      Chris

  3. ‘Fraid I have “wonky questions”, but I’m sure you’ll understand if I “wonk away”.

    First, is there any estimate of variability in the SCCO2 that comes out of the revised Nic Lewis 0.9C to 2.0C estimate? Presumably there are uncertainties in PAGE09’s calculations as well as uncertainties in estimates, and, in general even, say, a Gaussian uncertainty envelope on an input won’t produce a Gaussian envelope on outputs. I’d even bet it’s multimodal, unless very simple, like linear.

    I ask this not to be punctilious, but because policy might be better advised by such a range. If we (collectively) were trying to control mean-squared-error of social cost, that’d be one thing, but policy might prefer to control to something like maximum expected cost.

    Second, what’s the technical difference between the TCR, the equilibrium response at, say, 1000 years, and some kind of impulse response? I mean, presumably, for the equilibrium, that’ll be what we’re stuck with, no matter, based upon the 30%-50% residue of CO2 after 300 years, because scrubbing depends upon carbonate formation. And as far as impulse goes, the question would be for the increment of CO2 that’s emitted THIS YEAR, how much warming is done?

Leave a reply