Is it true that there’s no such thing as bad publicity? If so, we’re in luck. The paper that Elizabeth A. Stanton and I wrote on the social cost of carbon has been discussed on the Bishop Hill blog, a leading forum for British climate skeptics – and in comments on that blog and on Twitter by Richard Tol.
Bishop Hill cites us as estimating that the social cost of carbon – the monetary value of the present and future damage caused by emitting one ton of carbon dioxide – could be $1,000 or more. Tol calls this estimate “complete nonsense,” and Bishop Hill refers to the increase from the U.S. government’s $21 estimate to $1,000 and higher as “fairly jawdropping.”
Feel free to pick your jaw back up; we never said that the social cost of carbon is $1,000. We did say that the value should reflect important climate uncertainties, and that our modeling of those uncertainties produced a range of possible values from $28 to almost $900 for emissions today, or from $64 to about $1,500 for emissions in 2050.
A wide range of possible values is the only reasonable economic representation of scientific uncertainty about climate outcomes. Since the science says that outcomes are uncertain over a wide range, but catastrophic risks cannot be ruled out, then the corresponding economic evaluation should say that the social cost of carbon (SCC) is uncertain over a wide range, but catastrophically high costs cannot be ruled out. The reduction of the range of potential outcomes to a single, precise value such as $21 (or $1,000) slips in a radical change in the structure of information; it implausibly asserts that economists can find certainty where scientists cannot.
What are the uncertainties we considered? Using William Nordhaus’ DICE model, we examined median vs. 95th percentile climate sensitivity, high vs. low discount rates, and high vs. low estimates of damages at low temperatures, and at high temperatures. The multiple combinations of high/low estimates on four parameters gave us 16 variants of the SCC, spanning that vast range.
Richard Tol, in a comment on Bishop Hill, suggests that his forthcoming literature review of SCC estimates should be used instead of our analysis. In that article, updating his similar, earlier review, he includes 311 estimates, of which 184 (59%) come from his own publications. Those who want a Tol-centric review of the SCC literature should certainly consult his periodic updates, although readers should realize that these articles are self-referential to an extent that is unusual in academic literature reviews.
In debates on Twitter sparked by the Bishop Hill discussion, Tol has raised a number of other complaints. I am accused of having received funding from Friends of the Earth. Guilty as charged. I have also done work funded by the European Commission, various United Nations agencies, national and state governments, and charitable foundations, as well as other environmental NGOs. Tol has, on the other hand, cited obscure legal grounds for failing to reveal anything about his own funding. But really, we should judge one another’s work on the basis of its content, not its funding.
Since the Bishop Hill discussion Tol has tweeted, more than once, his belief that Liz Stanton and I are “mediocre” economists – a very weak substitute for substantive comment on our work. Come on, Richard: hurling hostile epithets at those we disagree with does nothing for the quality of debate.
And Tol has tweeted quite inaccurately about my critique of his FUND model. In an article with another co-author, I found that:
- FUND estimates that the worst economic impact of climate change will be the increased cost of air conditioning.
- FUND’s analysis of agriculture assumes a large net benefit from the first several degrees of warming, based entirely on research published in 1996 or earlier (the field has changed dramatically since then).
- Equation A.3 in the agricultural module (see the FUND documentation) of FUND versions 3.5 and earlier contains a serious risk of division by zero, for a plausible (relatively high-probability) value of one of the variables.
FUND version 3.5, containing that software error, was one of three models used in developing the U.S. government’s estimate of the SCC. When I recalculated FUND’s SCC estimate after attempting to correct the divide-by-zero error, the number more than doubled. If you’re interested in this, read my article, and the FUND 3.5 documentation – or if you’re truly fearless, the FUND 3.5 source code.
You won’t be surprised to learn that Tol disagrees. I’m sure you’re about to have a chance to read his response.
This article was also posted on the Real Climate Economics blog.
1. Discussion is good.
2. Ackerman and Stanton are mediocre economists by objective standards (http://ideas.repec.org/top/top.person.all.html).
3. Mediocre researchers are, of course, capable of useful research. I only raised the issue because Andy Haines exclusively referred to Ackerman and Stanton’s unpublished work. If you cite one study only, then pick a published, highly cited one. If you cite one person only, then pick the highest pedigree. On those grounds, Haines should have cited Nordhaus.
4. My funding is audited annually by the Comptroller and Auditor General of Ireland. No irregularities were ever found. Sources include Ireland EPA, US EPA, Asian Development Bank, Ireland Department of Energy, Science Foundation Ireland, and EU DG Research. There is corporate money, from energy companies with interests in peat, coal, oil, gas, and wind. No NGOs.
5. The latest meta-analysis is published (http://www.annualreviews.org/doi/abs/10.1146/annurev-resource-083110-120028). A preprint is available here http://ideas.repec.org/p/esr/wpaper/wp377.html). The data are here (http://dvn.iq.harvard.edu/dvn/dv/rtol). The results are corrected for author bias, and the data and algorithms are available for all to do their favorite sensitivity analysis.
6. Similarly, the FUND model is fully documented. The input data and source code are freely available. If you do not like our assumptions, you’re free to change them. Mr Ackerman forgot to include the link to the Excel version (http://dvn.iq.harvard.edu/dvn/dv/rtol) for those who do not appreciate C#.
7. Mr Ackerman correctly characterizes the energy and agriculture impacts in FUND. This material has gone through peer-review again and again. That does not make it right, of course. Mr Ackerman is free to publish his own impact estimates, but has yet to do so.
8. The division by zero is a non-issue. This is done in a piece of model code that was never used by us, and therefore never properly tested. It affects Mr Ackerman’s results from our model, but it does not affect our results.
See my detailed response here
http://bishophill.squarespace.com/blog/2011/10/26/for-whom-the-blog-tols.html
Most importantly, the division by zero is a non-issue. It is a bug in a piece of code that was never intended to be used in this way, and therefore never tested. It affects Mr Ackerman’s results, but it did not affect ours.
We repeatedly alerted Mr Ackerman to the fact that he introduced the error that he now blames on us.
I reproduced Tol’s results with the FUND model, then made a one-line correction to the code to eliminate the division-by-zero error. That one-line correction caused a big change in the model’s results. This doesn’t sound much like “a piece of code that was never used and therefore never tested”. If it was never used, how could a correction to it change the results?
The division by zero problem occurs in the core of FUND’s calculation of agricultural impacts – it is present in equation A.3, page 6, of the documentation for FUND version 3.5, available on the FUND model website at http://www.fund-model.org/. Equation A.3 calculates one of the three principal components of FUND’s estimate of the effects of climate on agriculture; it’s used every time the model runs.
Technical note: the divide-by-zero problem occurs in the fraction in equation A.3 when the Monte Carlo variable T-opt in the denominator = 1.6. This is not far from the mean of the probability distribution for T-opt, so values very close to 1.6 (and hence denominators very close to zero) are likely to occur quite often in the thousands of runs of FUND’s Monte Carlo analysis.
Mr Ackerman: In your email of March 17, 2011, 9:29 PM, you mention that you made more extensive changes to the model. Do I have your permission to release that email?
[…] a comment to yet another one of his postings, Mr Ackerman admits that he “made a one-line correction to the code”. A single […]
Whoa Nikki. I hope ur ok! Nasty fall, probably! It was pretty cool how u thought of Christmas wrapping paper right on the spot. And how Chloe made cookies! Awesome start!