It Goes On

I'm a "climate denier." Or a "climate blogger." Or something. Nobody seems to know. I'm not even sure what a "climate denier" would be, and I still cringe at the idea of being a "blogger." The reason I say this is there was recently a Press Complaints Commission (PCC) ruling which cited a couple of the posts I've written here. When it referred to me, it gave the footnote:

Brandon Shollenberger is described by some as both a “climate blogger” and a “climate denier”. He has published a number of online critiques including a negative critique of Dana Nuccitelli’s work.

This ruling bears on the long running controversy I've been somewhat involved in regarding Richard Tol and the changes he made to the latest IPCC report. People who follow my writing will remember this controversy built to the point I submitted a formal complaint with the IPCC in January. I still haven't received a response regarding it. I find that troubling, and I've tried following up on it, but so far I've had no luck.

In any event, the PCC ruling was made regarding complaints by Richard Tol about articles written by his critics. Tol claimed these articles were highly misleading, part of a "smear campaign" and amount to harassment. The PCC ruled against Tol on over a dozen specific claims, finding only one of his minor complaints had any merit.

Unfortunately, This hasn't resolved anything. There has been no resolution of the scientific issues underlying this controversy. Last week, Richard Tol published an article criticizing an article written by one of his critics, Bob Ward. In it, he says:

Mr Ward claims that there are "significant errors in [a] study suggesting global warming is good for the world". That study is a survey published by me in 2009. There were errors in that survey, now corrected, but the headline conclusions did not change. The latest survey contains four studies - by the late Ralph d'Arge, Robert Mendelsohn of Yale University, David Maddison of Birmingham University and myself - that conclude that climate change would improve human welfare. None of these studies have been found to be in error.

For some reason, this is a central issue in the controversy. Richard Tol received a significant amount of media attention and support from global warming "skeptics" after he published several papers claiming to show moderate global warming will be beneficial for the world. He got this result by looking at papers people had published on the issue and combining their results.

The problem is almost none of these results supported his claim. This wasn't apparent at first because there were many data errors in Tol's work, errors which made it appear more data supported his conclusions than actually did. This includes results from two papers which found global warming will be harmful which Tol somehow listed as finding global warming will be beneficial.

Tol refers to these errors and some corrections of them, saying:

Mr Ward highlights "shortcomings in the trend that [I] had fitted to the data". In fact, the original and corrected data are not materially different. There is no statistically significant difference between the trends fitted to the original data and to the corrected data.

This is somewhat true, but mostly misleading. Not all of the problems with the data were fixed, but it is true the "trends fitted" were not materially different. This can be seen in a figure comparing the two trend lines:

3-13-correction

However, this figure shows there is a significant problem: almost no data supports Tol's conclusions. The figure shows only one data point which suggests any sizable benefit from global warming. There's another data point, but it is so barely positive (0.1%) it is almost irrelevant.

To make matters worse, the one data point showing any sizable benefit from global warming was taken from Tol 2002, a paper written by Richard Tol himself. That means once a number of data errors were corrected, Tol's only support for his "headline conclusion[]" was a paper he himself published. His trend lines were unaffected by this, but trend lines dependent entirely upon a single outlier are obviously incorrect (and due to a simple statistical artifact). This is particularly noteworthy because Tol's original paper studying this data said:

it is striking that the estimates are in broad agreement on a number of points—indeed, the uncertainty analysis displayed in Figure 1 reveals that no estimate is an obvious outlier

In any event, as time passed, more data was added to the analysis. The first quote I provided from Tol in this post says there are now four studies which support his conclusions. One of them is Tol 2002. Another is by Robert Mendelsohn, which is the barely positive data point at 0.1%.

The other two are... I don't know. It appears they might be figments of Tol's imagination. Tol claims the "latest survey contains" studies by Ralph d'Arge and David Maddison "that conclude that climate change will improve human welfare." These studies are not present in any of the work which has been discussed. This is a table of the studies used by Tol for the IPCC report:

10-17-Figure10-Table

Maddison is listed as an author on three studies, all of which find negative effects from global warming (at -0.1%, -0.4% and -12.4%). The same studies are listed with negative effects in a recent paper by Tol (though strangely, the exact values are not the same). There is no apparent explanation for why Tol claims a study by Maddison finds global warming "will improve human welfare" when his own published data sets show the opposite.

There is also no apparent explanation for why Tol claims a study by Ralph d'Arge shows anything. No work by him was used in any of Tol's work that has been discussed. It wasn't even used in his more recent paper. Maybe some survey exists somewhere which shows what Tol claims, but if so, it has never been cited in any discussion so nobody reading Tol's article could possibly know what he's talking about. As far as anyone can tell, Tol may simply be making things up. He certainly is when he says:

Since 2009, however, more estimates of the economic impact of climate change have been published. These new results do affect the fitted trend, but not in the way suggested by Mr Ward. The new trend shows positive impacts for warming up to about two degrees global warming, just like the old trend did. The new trend, however, shows markedly less negative impacts for more profound warming than did the old trend. In other words, in the last five years, we have become less pessimistic about the impacts of climate change.

A correction Tol was forced to publish to his work shows this is untrue. This is the figure Tol published after correcting (some) data errors and adding new data:

3-13-update

There is no benefit from global warming shown in this figure. Richard Tol himself even acknowledged this in his correction, saying:

I nonetheless highlight two differences between the old and the new results. First, unlike the original curve (Tol 2009, Figure 1) in which there were net benefits of climate change associated with warming below about 2°C, in the corrected and updated curve (Figure 2), impacts are always negative, at least in expectation.

Tol's recent article claims a critic of his, Bob Ward, is wrong for saying the same thing Tol himself published in a scientific journal not even a year ago. How can Tol claim saying the same thing as he says makes you wrong? I don't know. Tol doesn't offer any sort of reference or explanation for his claims in his article. Apparently readers are supposed to just take him at his word... even if his word contradicts his other words.

Finally, Tol's recent article says:

Mr Ward claims that "mistakes had been corrected" in the Fifth Assessment Report of Working Group II of the Intergovernmental Panel on Climate Change. In fact, we replaced the vague "may be beneficial" with the precise "17 out of 20 are negative", in line with the IPCC style which frowns on ambiguous wording and emphasizes the more likely outcomes.

This is incredibly misleading. Tol acts as though the IPCC only made one change to the report in response to Bob Ward's complaints. The reality is there were over a dozen changes to the data displayed and listed by the IPCC. Tol ignores them by pretending the only thing which matters are changes to the text of the IPCC report, but you can find a list of additional changes in the formal complaint I submitted to the IPCC (linked to earlier in this post).


This is not an exhaustive list of what is wrong with Tol's article, much less his published work. There is far more I could comment on. The reality is Tol's work is completely and utterly wrong, relying upon inappropriate statistical models to compare imcompatible data, data Tol consistently mangles and misrepresents.

I've discussed this in some detail before so I'm not going to rehash it here. If anyone wants information, I'd be happy to provide it.

March 17, 2015 10:20 Edit: It turns out the ruling I mentioned in this was not by the PCC. Details can be found in this comment.

12 comments

  1. Russ R., quite a few, actually. I had planned to write a post about it before, but I decided against it due to the disinterest I've received on this topic. So far it seems everyone either agrees what I say is obviously right or refuses to even look at the matters. Putting more effort into examining the details won't help with either group.

  2. There's no reason to believe that the economic impact function follows any particular curve, so trying to estimate impacts for 1C of warming by fitting a curve to various impact estimates for 2.5C or more seems pretty silly.

    When you look at all the impact estimates for 1C of warming, there are only two and of those, only one shows a slightly negative economic impact. Unfortunately, it seems to be based on statistical measures of "self-reported happiness". I have no idea whether that even deserves to be taken seriously.

    The only other study happens to be Tol (2002), so I'd be interested to hear if this study is, in your opinion, valid or invalid?

  3. Russ R., I'm not sure how I'd judge the study on its own, but in regard to its use in the analysis we're discussing, I'd say it is definitely invalid. This is from the abstract:

    Global estimates depend on the aggregation rule. Using a simple sum, world impact of a 1C warming would be a positive 2% of GDP, with a standard deviation of 1%. Using globally averaged values, world impact would be a negative 3% (standard deviation: 1%). Using equity weighting, world impact would amount to 0% (standard deviation: 1%).

    The 2.3% value this all hinges upon arbitrarily choosing which calculations from the paper to use. The paper proposes three different ways of coming up with a global estimate, and it never concludes one is better than the rest. Here is more detail on the matter from the paper:

    Table VIII displays the impact on the world as a whole. Simply aggregating estimated impacts across regions leads to a positive impact (i.e., a benefit) of about $448 billion per year, equal to 2.3% of total world income. The standard deviation is a little less than half of that, at $197 billion or 1.0% of income.

    The interpretation of simple aggregation is not obvious. In fact, the estimate is a potential Pareto improvement, but compensation is unlikely. A global impact estimate is useful to a (non-existent) global decision maker, or a group of cooperating regional decision makers. In either case, the sum of regional estimates ignores the wide disparity between these regional estimates. Also, different monetary values are used for similar impacts, notably statistical lifes are valued differently.

    One solution is to use globally averaged prices to value non-market goods and services. Table VIII displays the result. World impacts are estimated at a negative $522 billion, or 2.7% of income, with a standard deviation of $150 billion, or 0.8% of income.

    Showing that +2.3% value could have just as easily been given as -2.7%. Or alternatively:

    Another solution is advocated by Fankhauser et al. (1997). When added, different regions’ impact estimates should be given weights. These ‘equity weights’ reflect the regions’ risk aversion and the world inequality aversion. A mild version is to use the ratio of global to regional per capita income as an equity weight. Table VIII displays the result. World impact is again positive, at $40 billion or 0.2% of income. The standard deviation is substantially larger, at $257 billion or 1.3% of income.

    It could have been given as +0.2%. Richard Tol has never explained why he chose to use 2.3% when referring to this work even though it's become clear that choice is the only thing which allowed him to garner media attention by telling everyone moderate global warming will beneficial. It is the only reason "skeptics" have been able to advance this talking point based on his work.

    Is that +2.3% the "best" estimate from Tol 2002? Maybe. I don't know. What I do know is it is completely inexcusable to cherry-pick the most favorable set of calculations and pretend the rest don't exist. If that estimate was the "best," Tol should have shown all estimates from Tol 2002 and explained why we should prefer one over the rest.

  4. Of the three calculation methods described in Tol (2002), none is strictly "correct" in economic terms (i.e. no matter what, you're still trying to add apples and ostriches) but if the choice of any of three arbitrary methods changes the sign of the impact, then the conclusion can only be that the results are inconclusive. (Which is not the same as saying the study is invalid.)

    But I agree that it's irresponible to promote an outlier result and discard the others.

    My conclusion... warming will be beneficial for some and detrimental for others, but there's no conclusive evidence that, on aggregate, the net global impact of a modest amount of warming (ie 1C) is either positive or negative.

  5. Oh, by the way Russ R., I'm not sure if you've seen some of my previous commentary on the figure in question. In case you haven't, one of the central problems with this whole thing is the ~20 studies all seek to measure different things. The huge outlier at the bottom uses PPP GDP (as opposed to nominal GDP which is used by the rest). One paper, I think Bosello, estimates damage for a rise in temperature since pre-industrial times. Its numbers were treated the same as numbers from papers estimating damage for a rise in temperature since ~2010. Those are obviously not measuring the same thing. Then there's a paper which estimates damages for a rise in temperature from a different baseline for each country. I have no idea how that's supposed to be compared to the others given its not even working with a single rise in temperature.

    Given the baseline temperatures for each paper are different, the chart is all sorts of wrong. Even if we ignore that though, the estimates are still not very comparable. You have estimates based upon ~1995 economies which would be significantly different if they were updated to reflect the ~2005 economies used for other papers, including a 1996 paper. Yes, that's right. Tol used 2005 economic data with formulas from a 1996 paper to come up with one estimate yet took an estimate from a 1995 paper which was based upon ~1995 economic data. Why? I don't know. Just like I don't know why he replaced the data used in one paper with a different data set than did his own calculations rather than just using the results published in the paper. (See this post for the details of those examples.)

    Then there were papers whose results weren't even based on actual economic data, but rather, relied upon projections of what economies would be like in the future. These projections are all different, and the ones we can check against data collected since they were made haven't been very accurate. Even worse, some of these estimates also depend upon GCM projections. That means we have results based upon GCM models which skeptics are quick to criticize and economic models which are notoriously inaccurate, and which (if any) models are used depend upon the paper.

    There is no reason to believe all these results are comparable. Even if they were, there is no reason to think they should all be treated the same way. One paper did nothing more than ask some people how much damage they thought global warming would cause, and that was nearly 20 years ago. Are we really supposed to believe the opinion of some people ~20 years ago should be compared to numbers taken from the self-reported happiness of people (like you, I'm surprised by that one), which should be compared to modeled results specifically designed not to reflect the patterns of global warming we might see?

    I find that baffling, but there's actually something even worse. These ~20 papers did not just come up with single values for their conclusions. They contain far more information than that. If we do want to compare them to one another, we have far more information we can use than Richard Tol did with these figures. I have no idea why he would ignore it. He's never explained why we should only take single-point estimates from papers which provide far more than just single-point estimates.

  6. Brandon, one minor correction, right at the top, regarding what you describe as a "Press Complaints Commission (PCC) ruling".

    The Press Complaints Commission no longer exists. It was closed down in September 2014. It was replaced by something called the Independent Press Standards Organisation (IPSO). But membership of IPSO is voluntary for news organisations, and the Guardian has not signed up to it (as I found out when I tried to complain to IPSO about a Guardian article). This means that the Guardian has no independent oversight.

    So what is the document linked to at the top? It's from something called the Scott Trust, which sounds like something independent and sounds ("Trust") like it might be a kind of charity. But in fact, it's a company, and it's the company that owns the Guardian. As you might say, "I find that troubling", or even "incredibly misleading" 🙂

  7. Thanks for pointing that out. I saw Richard Tol had filed a complaint with the PCC, then I saw a link to that ruling, and I assumed the ruling was by the PCC. Apparently the ruling arose from the complaint with the PCC, but it was made by a different group (one which I guess owns the Guardian) since the PCC shut down. It's a relative minor thing for this post, but it is definitely the sort of detail I should make sure to get right in the future.

    I'll try to get it corrected by this afternoon (editing a post from my phone would be a huge pain).

  8. That is a fascinating comment Richard Tol. I am sure it will resolve all the outstanding issues highlighted here, like you apparently making things up, criticizing people for saying the same thing you said, and oh, even basing a significant amount of your work upon a blatant cherry-pick of your own results.

    Or it won't, because you're apparently incapable of addressing criticisms in anything resembling a reasonable manner.

Leave a Reply

Your email address will not be published. Required fields are marked *