As I mentioned in my last post, I went on a trip this weekend. It was mostly great. I didn't really get to sleep though, and I'm exhausted. I wasn't planning on writing a post today, but I had to because there was a remarkable development over the weekend which turns out to have been even more remarkable than I had realized.
As readers will know, some of my recent posts have talked about work by Richard Tol. That started because of an article I wrote about his work being published at DeSmog Blog. The article gave a brief overview of the history of Tol's work and the many errors in it. One of the most remarkable errors Tol had made was to claim papers which said global warming would be harmful would instead be beneficial.
It's difficult to imagine a more serious error than completely inverting people's conclusions. Tol's explanation was:
“Gremlins intervened in the preparation of my paper 'The Economic Effects of Climate Change' published in the Spring 2009 issue of this journal. In Table 1 of that paper, titled 'Estimates of the Welfare Impact of Climate Change,' minus signs were dropped from the two impact estimates.”
It's not clear why gremlins should be blamed rather than Tol himself, but as my article points out, Tol made the exact same mistake in a new paper he's written, only this time, he did it with results he had previously used correctly. Not only that, those previous results were ones he him himself had calculated. But somehow, after five years of using them, he suddenly turned them upside down.
What happened is originally he calculated results using data provided by the authors of a paper (Nordhaus and Yang 1996), finding the results showed global warming would be harmful. He used those results for five years. Then one day, he decided to throw out the data used by the authors, replace it with different data and recalculate the data.
Why? Nobody knows. Tol didn't explain why he did it. He didn't even acknowledge having done it. He did it in secret.
That seems pretty bad, right? It is, but guess what? It doesn't matter. The new results still said global warming will be harmful. The difference in results was to change one value from -1.7% to -1.4%. That's all. Tol secretly redid the calculations for basically nothing.
But then the "gremlins" came back and made him forget his results were in damages. Somehow, despite doing the calculations on the exact same page as the old ones, in almost the same format as the old ones (this can be seen in the spreadsheet he used), Tol somehow forgot his results were for damages, so he decided they were benefits and published them as such.
That's it. Tol just somehow completely inverted results he had been using for over five years by forgetting the calculations he did were for how much damage global warming would do, not how much benefit it would have. And he didn't notice this mistake even though the new results were completely at odds with the old ones. And he did this even though he had made the exact same mistake twice before.
It's mind-bogglingly stupid. I noticed the problem within moments of looking at Tol's new data table and figure. It was obvious because the results were so different than his previous ones. And even if I hadn't been familiar with his previous work, anyone just comparing the previous version of his table/figure to the new one to see what's changed would have spotted the issue immediately.
But still, people can be stupid. Tol made this mistake before. I guess he could have made it again, somehow. I guess he could have secretly redone a set of calculations and failed to notice the dramatic changes it caused. It'd take a mind-boggling level of stupidity, but perhaps not any dishonesty. Only, the paper also says:
There are three differences between this table and the IPCC one. First, the table here includes the estimates by d’Arge, Berz and Nordhaus 1982. Second, the Mendelsohn estimates are shown against the area-average temperature change, rather than the population-average, just like the other estimates in the current table. Third, the Maddison and Rehdanz estimate is shown in market exchange rate dollars, rather than in purchasing power parity dollars, just like the other estimates in the current table
You'll notice that paragraph doesn't mention Nordhaus and Yang (1996) anywhere. Tol says there "are three differences between" his current table and the previous one, but nowhere does he mention the difference which caused him to claim a paper which said global warming will be harmful actually said global warming will be beneficial.
Even if he didn't realize how much he changed the value for Nordhaus and Yang (1996), there's no way Tol was unaware of the fact he redid the calculations for it. Intentionally not disclosing the fact he redid them was dishonest, pure and simple.
But I'm sure some people would call it minor. As my last post mentions, a lot of people seem hesitant to criticize Richard Tol. I suspect they'd dismiss that dishonesty rather than call it fraud like they would have if it had been done by someone else. So lets look at some greater dishonesty and stupidity by Tol.
That's right. Tol's decided to double down on his shenanigans. You see, after my article was published, I e-mailed Tol to inform him of it. I also explained to him why it could be considered fraudulent to redo calculations to change their results for no legitimate reason, while claiming not to have changed them at all - even if one didn't completely invert the results in the process.
While Tol did respond to me in a nonsensical manner, over the weekend, I was informed of Tol's further response - to try to cover up his mistake. This is from his paper, as I originally saw it:
This is from his paper, as you'll see it if you download it now:
You're looking at the first part of one of the tables in his figure. It's been changed. Nordhaus and Yang 1996 got its value switched from 1.4 to -1.4. that change isn't mentioned anywhere. Tol secretly changed his paper to cover up his error so people wouldn't know about it.
That means anyone reading my article about Tol's work who follows its link to his paper will likely be confused as to why the paper doesn't say what I claim it says. They may well be tricked into thinking my article is wrong because they don't realize Tol has deceived them. It's ridiculous.
This is completely and utterly dishonest. If Tol wanted to correct his error, that would have been fine. There's no need to leave mistakes in papers. But if he wanted to correct his error, he should have acknowledged the error and acknowledged he was correcting it. And he should have acknowledged what effect the error had on his results. Because it had one. This is a table from the first version of his paper:
This shows the mathematical model Tol fit to the data to estimate what effect global warming would have on the economy as compared to that used in other papers. Here is the same table after Tol's secret update:
The changes are significant. I won't go into all the details right now since it'd require discussing math, but what matters the most is the changes to the model used for "This paper." That model called a "piecewise linear function." In simpler terms, it's two straight lines. You take two straight lines, of whatever lengths you want, and try to arrange them so that their ends touch and they overlap the data as best as possible.
You can see a pretty significant change in Tol's model. The first line he uses, was originally used for warming of up to 1.24 degrees (T<1.24) and went up at a rate of 1.24% of GDP per degree. This can be seen by where it meets up with the second line, which has changed from 0.92 to 0.83. Or of course, we could see it by comparing the figures in the two papers. Here is the figure from the first paper:
As you can see, there are two data points well above zero in this. That should have raised red flags since there weren't two data points well above zero in the previous version of this data set. I don't know why Tol failed to be curious about this sudden change.
There's also another issue. If you look, a lot of the data points seem to be plotted at strange temperature values. Somehow, most of the values have been shifted to the right a few tenths of a degree. I thought that was weird, but I wasn't sure if maybe it was somehow a display issue on my end or what. It turns out they weren't, and they were fixed along with the data error.
So that means we have one data point moved from well above the zero line to well below the zero line, many data points shifted slightly to the left, and the piecewise linear regression significantly altered. All in secret. The result is this:
Wow. Would you look at that? Tol was so secretive about his change to his model, you can't even see the change!
That's right. Tol fixed the data error my article pointed out, and he updated his model to account for the change. And he did this in secret so nobody reading his paper would know about it. But because that apparently wasn't enough, he also decided to update his figure to reflect the change in his data but not the change in his model.
Is it possible that was a mistake? I mean, could he have honestly failed to notice the lines in his figures were exactly the same even though the data was different? I guess. I mean, at this point, I wouldn't put anything past him. I don't really care though. Whether he failed to show the impact his error on purpose or not doesn't matter. Because remember:
These changes were all secret.
Maybe Tol meant to update the figure to reflect the change in his model while making his secret changes but somehow failed to. He was trying to deceive people either way. It doesn't matter if he stumbled into this deception in a fit of idiocy while engaging in other deception or if he had actually planned it out in the first place.
Now, at this point I'd normally show you the effect Tol's deception had. Unfortunately, I can't really do that. You see, it turns out after writing the above section, I decided to plot a before and after version of Tol's model. When I went to, I realized I had made a mistake. When I saw the models listed in Tol's paper had changed, I assumed that was because of the data error he had corrected.
I know, crazy me, right? Tol secretly corrected a data error, and at the same time, he secretly updated some results. Why would I ever think the two were related?
Well, it turns out they weren't. Or, rather they were, but they weren't at the same time. It's complicated. Or at least, too complicated to bury at the end of this post, especially since it involves some math. So rather than make this post even longer than it already is, I'll leave this as a teaser:
Not only does Richard Tol make secret changes to try to hide his errors from the world, lie to claim he hasn't changed data and blame "gremlins" for somehow claiming papers which say global warming will be harmful say global warming will be beneficial... he can't do math. As in, he can't do basic arithmetic.
Stay tuned for a post showing this on Thursday!