A year ago, I was a witness in the Mann v Steyn-Simberg libel trial. It was an extremely frustrating experience. Mann’s lawyers tried to block McKitrick and I from giving evidence against Mann. The judge ruled against them, saying that our evidence was relevant to the defense. However, the judge then prevented either of us from presenting evidence on Mann’s concealment of failed verification statistics or even on the verification failure of Mann’s statistical model. The judge didn’t even allow the presentation of a table published in Geophysical Research Letters. In mid-trial, the judge also reduced the time available for the defense by about 40% from the original allocation; the time available for McKitrick and myself was almost chewed up by defense objections.
Because Steyn was so weak, nearly all of the defense was taken up by Simberg’s lawyers. They were highly professional, but their strategy was focused almost entirely on the lack of damages to Mann, and, in particular, to Mann’s claims about lost grants. In my opinion, the issues about, for example, Mann’s concealment of adverse verification statistics were issues that ought to have been raised in cross-examination of Mann (rather than late in the day in direct examination of McKitrick or me), but none of this took place. Instead, the cross-examination went on and on about Mann’s grants - an issue which seemed far less important to me than putting Mann on the spot about his concealment of adverse verification statistics,
However, the defense focus on Mann’s grants was vindicated by the scathing comments of the judge in his recent sanctions order against Mann and his lawyers (link; link), including the following:
"Dr. Mann's assertion that there was no falsehood or misrepresentation in his testimony or his counsel's conduct borders on frivolity."
"the record plainly shows the deliberate and knowing misconduct of Dr. Mann's counsel in eliciting false testimony from Dr. Mann and misrepresenting his grant funding."
"Dr. Mann's counsel's bad faith misconduct is an affront to the Court's authority and an attack on the integrity of the proceedings warranting sanctions."
The judge focused on two components of Mann’s testimony on supposedly lost grants following the blog articles: overstatement of non-funded grants; and under-statement of funded grants.
But neither the judge nor the defense lawyers had deep background in Mann’s “statistical” techniques and, as a result, the judge’s disparagement of the conduct of Mann and his lawyers, scathing as it was, merely scratched the surface. For today’s article, I’ve done a fresh analysis of Mann’s presentation and shown that there was much more to “Mann’s DC Trick” than discussed in the sanctions order. (I’ll do a separate article elucidating the sanctions order, but, in this article, will focus on issues that were overlooked in the trial and order.)
The centerpiece of Mann’s claim for lost damages was his assertion that his grant successes went from from “just under a million a year” in the four years ($3.3 million total) prior to the Simberg-Steyn blogs to “a little more than 100,000 a year” in the four years after ($500,000 total). They illustrated this claim with the “demonstrative” shown below right. This claim was asserted in the opening statement, in Mann’s direct evidence and the closing statement. An excerpt from Mann’s direct is below left. In the sanctions order, J Irving observed a significant exaggeration in the statements and demonstrative: according to Mann’s own data, the total value of grants after the blogs was $895,000, not $500,000; and reduced the before-and-after “disparity” from $2.8 million to $2.4 million. The judge was very troubled by this exaggeration.
But there were several other issues that go much deeper, issues that call into question the very existence of the mid-2012 breakpoint asserted by Mann and his lawyers and which strongly suggest that Mann’s grant claims were the product of data manipulation - the exact subject of the underlying science controversy.
It’s nearly always a good idea in statistical analysis to begin by plotting the data.
In the figure below, I’ve plotted Mann’s annual grants from 1996 to 2021 by US fiscal year ending September 301 using information in Mann’s CV circa 2022 and his amended interrogatory responses (March 2023). In the figure below, the 2008-2016 period selected into Mann’s demonstrative is highlighted in yellow background. The step function in the highlighted area illustrates Mann’s evidence: $1 million per year in the four years prior and $100,000 in the four years subsequent. As a reality check: does this step function fit the data? Obviously not. It’s not even close.
Some comments:
there is NO - repeat NO - evidence of a July 2012 breakpoint in the data. Any such claim is completely false. This is immediately visible in the above figure.
the claimed amount of Mann’s grants in 2009 was uniquely high. It was more than 10 times greater than the median value of Mann’s annual grants. Without 2009, the values for the three fiscal years prior (2010-2012) had no statistically significant difference to values for the three fiscal years subsequent (2013-2016). This is immediately visible in the above figure.
Because the 2009 value is so anomalously high, the timing of the step in Mann’s step function depends on the arbitrary length selected for the step. If a comparison period of two years is chosen beginning in 2008.5, the step will be 2010.5; for three years, the step will be 2011.5; for four years, the step will be 2012.5; for five years, the step will be 2013.5. One can reasonably surmise that Mann’s choice of four years prior as the comparison period was selected in order to place the “step” at the target date of 2012.5,
as an overall point on grants as a statistical distribution, grants were zero approximately 40% of the time over the 26 years, and between $100K and $500K about 54% of the time, with two outliers (2006, 2009)2. The distribution has peculiar “tail” behavior: almost 50% of Mann’s total grants over 26 years were obtained in these two outlier years.
The 2009 Outlier
Because the 2009 grant claim is such an extreme outlier compared to the other grants and because Mann’s “loss” claim is so dependent on this outlier, the outlier needs to be specifically examined3. There are three large issues.
ARRA
Mann’s 2009 grant claim consisted of two grants funded under the 2009 Obama “stimulus” bill - the so-called American Recovery and Reinvestment Act (ARRA): one grant ($541,000) in which he was the Principal Investigator and one much larger grant related to vector-borne disease ($1,885,000) in which he is listed as a Co-PI. The figure re-states the 1996-2021 grant figure by adding color coding by sponsor type, ARRA grants in red and magenta. The other anomalously large grant was a USAID grant in 2006 in which Mann was a Co-PI.
The American Recovery and Reinvestment Act (ARRA) was passed on February 17, 2009, in the first month of the Obama presidency, and had a total budget of $831 billion - about the same, allowing for inflation, as the $893 billion budget of Biden’s so-called Inflation Reduction Act in 2022. Approximately $3.1 billion of ARRA funding was allocated to the National Science Foundation (NSF).
In an interesting recent Jon Stewart podcast (link at 44 minutes), Ezra Klein noted the total failure of the ARRA program to deliver anything on its signature promises: high-speed rail, “smart” grid or interoperable electronic health care records, although, in fairness, others have pointed out (link) that it did succeed in building thousands of “ADA corner crosswalk things that didn't actually connect to anything”. It also funded research by Michael Mann (link) that may or may not “actually connect to anything”. In retrospect (and probably in real time), one would have thought that the purpose of the stimulus would have been better served elsewhere than academic researchers4.
In any event, by 2012, ARRA funding had almost entirely run its course. The implication of Mann’s claim to a “$2.8 million” shortfall is that Mann was entitled to a second helping of the ARRA bonanza during the second Obama term, even though the program had expired - a notion so unpalatable that neither Mann nor his lawyers expressed it in those terms.
In Steyn’s cross-examination of Mann, Steyn repeatedly asked Mann whether there were other factors besides the Simberg-Steyn blog articles which could have contributed to reduced grant funding. Obviously the expiry of the ARRA program removed the very source of funding which had funded Mann’s outlier 2009 grants. But Mann, needless to say, didn’t mention ARRA.
Contract or Subcontract?
The 2009 claim related to the vector-borne disease disease ($1,884,991) was much larger than other grants in Mann’s CV (most of which appear to have funded a couple of grad students). Mann’s claim in the interrogatories was equal to the full amount of the grant as reported in the NSF summary report (see link). The NSF summary report listed 29 articles attributable to this grant, but only one lists Mann as a coauthor. This strongly suggests that Mann was a subcontractor to this project.
In other interrogatory cases where Mann was a subcontractor, Mann’s amended interrogatory only claimed the amount of the subcontract, which, in an important example cited in J Irving’s sanctions order, was a tiny fraction of the total contract.
In Mann’s June 2020 interrogatory responses, he had claimed an amount of $9,713,924 in non-funded grants in relation to a grant application entitled “WAter VariabilitiEs Stressors and Sensitivities (WAVESS)”, dated 9/1/2014. (He had listed this amount in his CV at the time.) In his amended March 2023 response, Mann reduced the claim to $112,000! The lesser amount was the amount attributable to his Penn State group; the larger amount was the amount attributable to the entire project.
This became an issue in the trial because Mann’s team presented the incorrect June 2020 information to the jury in an exhibit. When the defense objected to the exhibit, Mann’s team falsely reassured the judge that there were no significant differences between the exhibit and the final numbers. In the sanctions order, J Irving excoriated Mann and his team both for presenting the false number to the jury and for falsely reassuring the judge about the validity of the numbers.
The purpose of this example is to show the difference between the value of the full contract and the value attributable to a subcontractor. The question then is whether Mann’s calculation should have shown the full value of the grant (most of which went to a different department) or the amount attributable to Mann’s putative subcontract. If the latter, then the appropriate amount is likely more like $125,000-150,000, as opposed to $1,885,000.
Mann similarly appears to have been a subcontractor on the 2006 USAID grant that comprised the majority of the 2006 outlier. It was listed in his CV at its total face value, rather than the value of Mann’s subcontract.
Climategate, November 2009
The third factor against inclusion of 2009 as a comparandum is, of course, that the Climategate emails were released in November 2009, about 6 months after the extravagant 2009 ARRA “stimulus” grants.
On January 20, 2010, during the midst of Climategate controversy. Mann’s receipt of 2009 “stimulus” funds became the subject of a Wall Street Journal opinion editorial (archive; link) entitled “Michael Mann’s Climate Stimulus: A case study in one job ‘saved’”.
The editorial observed that they had contacted NSF about the awards to Mann, but were told by a representative that she was "unaware of any discussion regarding suspending or changing the awards made to Michael Mann." The op ed concluded with the observation that “your tax dollars will continue to fund a climate scientist whose main contribution to the field has been to discredit climate science”.
The NSF made these awards prior to last year's climate email scandal, but a member of its Office of Legislative and Public Affairs told us she was "unaware of any discussion regarding suspending or changing the awards made to Michael Mann." So your tax dollars will continue to fund a climate scientist whose main contribution to the field has been to discredit climate science.
Mann didn’t mention this editorial as a contributing factor nor did the defense ask him about it.
Conclusion
J Irving was scathing at Mann’s understatement of funded grants after the Simberg-Steyn blogs which reduced the supposed “disparity” from $2.8 million to $2.4 million. But when the 2009 outlier is excluded (for any one of the three reasons cited above), the so-called disparity between 2010-2012 and 2013-2016 grants is eliminated, as shown below. (In the figure below, estimated value of 2006 and 2009 subcontracts are shown, but these are outside the highlighted period anyway.) Readers who are familiar with Mann’s “scientific” work will be unsurprised at Mann’s “research” into his grants.
In the interrogatories, Mann collated the grants according to calendar year of the start date of the grant. In his evidence, he observed that the decision date for a grant with start date of September 1, 2012 was prior to July 31. The date of the blogs was July 31, 2012. For practical purposes, classification by the Sept 30 fiscal year of the start dates will closely approximate classification by the July 31 fiscal year of decision dates. In the CV, grants are shown by calendar year of start dates; for the purposes of this graph, the fiscal year and calendar year totals are the same. Any discrepancies will not change the conclusions,
Here is a histogram
The only other year in which Mann’s grants exceeded $500,000 was in 2006, during which Mann was a subcontractor to a USAID grant entitled “Climate Change Collective Learning and Observatory Network in Ghana”. See link for a profile of this program.
Scientific American (link) asked this exact question on February 17, 2010: Is the Recovery Act Stimulating Science and the Economy? They observed that most of the money was sitting in university bank accounts and had failed to provide the quick economic stimulus which had been the ostensible purpose of the legislation.
Brilliantly presented. Thank you, Steve.
Henceforth, I propose that the Streisand effect be updated. Now, "the Mann-Streisand Effect."