We’ve just had another consensus paper published. The paper is called Does It Matter if the Consensus on Anthropogenic Global Warming Is 97% or 99.99%? (pre-print here). The lead author is Andy Skuce, who has a nice post about it. Dana Nuccitelli is also one of the authors and has written a Guardian article.
the consensus on AGW among publishing scientists is above 99.99%, verging on unanimity.
Powell’s paper also claims that the Cook et al. (2013) 97% was obtained by ignoring a large number of papers/abstracts (those that took no position) which Powell argues should be rated as endorsing the consensus. His paper also claims that the Cook et al. method would return a nonsensical result if applied to a topic for which their is actual unanimity (plate tectonics, for examples) – it would lead to, Powell claims, dividing zero by zero.
Dana and Andy’s articles (linked to above) provide very clear explanations of our response, so I don’t need to say much. One issue is simply related to defining what’s being determined. Powell determined what fraction of abstracts/papers rejected anthropogenic global warming (AGW) and concluded that the rest endorsed it. Cook et al., however, established what fraction of abstracts that took a position with respect to AGW, endorsed that humans are causing global warming. Powell, therefore, gets that 99.99% of abstracts/papers don’t reject AGW, while Cook et al. conclude that just over 97% of abstracts/papers endorse that humans are causing global warming. These are not quite measuring the same thing. In fact, if you apply Powell’s method to the Cook et al. data, you also get that more the 99% do not explicitly reject AGW.
Powell’s claim that Cook et als method would return a nonsensical results for a topic about which there is unanimity is simply wrong. Cook et al. rated papers as to whether they took a position, or not, and – if they took a position – if it was explicit or implicit. The consensus was then the fraction of papers that took a position, that endorsed (implicitly, or explicitly) the consensus. Even if explicit endorsements are unlikely for topics about which there is unanimity, a reasonable fraction of papers/abstracts still implicitly endorse the consensus. We tested this in our paper using plate tectonics as the topic, and did indeed find that a reasonable fraction of abstracts do indeed implicitly endorse the plate tectonic consensus. Of course, none reject the consensus (explicitly, or implicitly) and so the computed consensus is – as expected – 100%.
One could argue that maybe it doesn’t make much difference if the consensus is 97%, or 99.99%; it’s still almost all. Personally, I think it is important to be clear about what is actually being determined; the 97% and 99.99% are determining slightly different things. Also, that 99.99% of abstracts/papers do not reject AGW does not tell you what fraction endorse the stronger position that humans are causing global warming. The result does depend on what position is actually being tested – and whether you assess papers, abstracts, or people – but if you consider relevant experts, or the relevant literature, you find that more than 90% agree that humans are causing global warming. Showing that more than 90% (and typically around 97%) endorse that humans are the dominant cause of global warming, seems stronger than showing that virtually all accept some human contribution. Of course, the latter is still a very useful metric as it does show that explicitly rejecting AGW is now extremely rare.