Knowingly Promoting False Conclusions(?)

Richard Tol has a blog post responding to an article in the Guardian discussing one of the undisclosed changes to IPCC report I mentioned in my last post. There are a variety of things worth commenting on in it, but one stands out to me more than the rest. The Guardian article focused on the fact this sentence had been removed from the IPCC report:

Climate change may be beneficial for moderate climate change but turn negative for greater warming.

Tol’s post explains:

Here is the story. The old data (the blue circles) roughly fit a parabola: first up, then down, and ever faster down.

The new data do not fit a parabola: The initial impacts are positive, but the progression to negative impacts is linear rather than quadratic.

If you fit a parabola to this data, you will find that the mildly negative estimate at 5.5K dominates the positive estimate at 1.0K and the sharply negative estimate at 3.2K. The parabola become essentially a straight line through the origin and the right-most observation.

I think the appropriate conclusion from this is to fit a bi-linear relationship to the data, rather than stick with a parabolic one. This was not yet in the peer-reviewed literature when the window for AR5 closed, so we decided to just show the data.

To better understand what Tol is saying, we can look at the old and new curves he refers to. Here’s the image he tweeted of them:

As we can see, the old curve says some amounts of global warming will have (net) benefits. The new curve says no amount of global warming will have (net) benefits. That’s a significant change.

Tol’s post explains that change is not due to the multitude of data errors in his earlier work as claimed by some. We can verify that by looking at the effect of the corrections:


The corrections clearly are not the cause of the change in his conclusions. That means, as he says, the change is due entirely to the newer, more up-to-date, data. That would be the data he showed as diamonds in this figure he added to the IPCC report:


But think about that. The figure he added has the new data, meaning the curve for it would show no benefits. Tol says that means a different regression should be used to generate the curve, but that hadn’t been done at the time. The only regression which had been published would show no benefits. Given that, why would Tol try to make the IPCC say:

Climate change may be beneficial for moderate climate change but turn negative for greater warming.

When he knew there was no published work to support that conclusion for the data he was showing? Doesn’t that mean he knowingly added a conclusion to the IPCC report which couldn’t be supported by any published literature? And doesn’t that mean the IPCC lets authors add conclusions to its report not supported by the data shown in the report, much less any published literature?

Or am I missing something?

Side note, Tol’s claim to have fit a parabola to the data is false. He actually used a linear + parabolic fit. That’s why the new curve was so linear – he used a model with a linear component.



  1. A response to your side note: Every parabolic fit has a linear term when viewed from a perspective other than its line of symmetry. The general form of a parabola (assuming a horizontal directrix) is
    y = A(x-x0)^2 + C.
    But if viewed as an expansion from x=0, it’s y=a x^2 + b x + c, with a non-zero b term (if x0 is non-zero).

  2. HaroldW, I’m afraid I have to disagree. What you describe is a quadratic fit. It has a parabolic and linear component. Parabolas are, by definition, absent the linear (x^1) component. If you include a linear component, you’re no longer using a parabolic fit.

    I probably wouldn’t have cared enough to leave a note, only Richard Tol said things like, “The new data do not fit a parabola.” That statement is only true when interpreted as written. If we interpret it as referring to a quadratic fit, it is no longer true. A quadratic function does fit the data. It has a couple outliers, but nothing too severe.

    When Tol’s entire point rests upon his fit being too linear instead of parabolic, I think it’s appropriate to point out that’s only true because he used a fit with a linear component.

  3. Richard Tol, you said this about that paper in the post I discuss:

    This was not yet in the peer-reviewed literature (it is now: paper open access pre-print) when the window for AR5 closed, so we decided to just show the data.

    You yourself state the paper the results of that paper were not published in literature yet, thus they did not qualify for the IPCC report. If you’re claiming those results are the basis for the removed sentence, you’re openly admitting to bypassing the IPCC’s guidelines to sneak your conclusions in.

    Incidentally, it’s cute you assume I’m not aware of that paper. I had planned to write a post about that paper and another of yours months ago. I decided not to because I realized nobody had any interest in the work.

  4. Brandon: You’re again assuming things, not having been in the room. In an intermediate draft, we had included the kernel regression. I would argue that applying a standard method to data does not count as new research. In the end, we decided to leave that out, because Nadaraya-Watson needs a twist to make it work in this case.

  5. Richard Tol, I’m not assuming anything. The text drew a conclusion which was not based upon anything present in the report. That you, at some point, thought about showing unpublished work which supported the text is irrelevant. Also, this is just stupid:

    I would argue that applying a standard method to data does not count as new research.

    There is nothing about the regression you performed which is “standard” for this problem. I’d argue the methodology in question isn’t even appropriate, but I don’t need to. The reality is there are dozens of “standard method[s]” you could have used. Arbitrarily picking one which gives the answers you like is not okay, especially not in the IPCC report which prohibits the creation of new research.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s