In the Cook group paper, the ‘authors’ measure the degree of acceptance of a ‘consensus’ in climate literature.
Remarkably enough, this is what they find:
From 1991 to 2011, the fractions of papers accepting the orthodox position decrease with time (Figure 1 & 2).
Of the papers said to have accepted a consensus position, the major fraction, declines from 33% to about 24% ( ‘implicit endorsers’, ‘3’ in Figure 1) (Figure 2).
Papers that explicitly support the consensus position (‘2’ in the graph) also decline.
Cook and co-authors say they identify ‘strengthening consensus’, among other increasing consensus trends. The underlying data however does not support their claims. Instead, there is a remarkable stability in the overall composition of the literature. There is a steady increase in the proportion of neutral papers (called ‘No position’). In other words, no partisan category increases (or decreases) at the cost of another (Figure 3A & 3B).
Strangely enough, Cook and co-authors take note of these findings. Their interpretation however reveals a major problem in their analytic approach.
Cook and co-authors rationalize the decrease in the proportion of papers supporting the consensus, via a convoluted theory, as evidence for a high degree of consensus. They contend the decrease implies more papers have accepted the consensus and therefore don’t need to talk about it. At the same time, they take the increase in absolute numbers of orthodox position papers as evidence for ‘increasing consensus’.
The fallacy in reasoning is shown easily. Consider, as an example, a prosperous county which shows 60 cases of pneumonia in 1993. The media raises a hue and cry. Stung by criticism, the county institutes rigorous public health and education measures. Twenty years pass and a survey is undertaken. The cases of pneumonia for 2012 is 80. The media goes on a rampage. Is this justified?
It turns out that it is not. The county experienced a population boom in the early 2000’s and the incidence of pneumonia/100,000 per year actually fell during the period.
Now, imagine a mayor loudly criticizing local health officials for the increase in pneumonia cases, and a while later, traveling to a conference to boast that his city had lowered pneumonia rates due to measures undertaken by him.
This is exactly what Cook and his co-authors do. They put a different spin on two facets of the same observation.
Finally, if the proportions of papers accepting the orthodox position is decreasing, by what way does their actual number go up? The explanation it turns out is deceptively simple.
Cook et al studied 11944 papers for acceptance (or rejection) of AGW. There are papers which explicitly state something with respect to this question, and those that do not. As noted above (in Figures 1, 3A, & 3B), the overall composition of the literature remains more or less constant. Yet total numbers of papers published in the climate field increases dramatically during this period (Figure 4), particularly after 2005.
Examine the composition when the two groups are broken apart. Shown in the graph below, the light blue line is papers that don’t say anything explicit and the red line is for papers that make an explicit statement (Figure 5). As can be seen, the rise in number of papers seen in Figure 4 is almost entirely made up by papers that say nothing explicitly about anthropogenic warming.
In their study, Cook and co-authors include a significant chunk of the rising group into the ‘endorse the consensus’ category. In a widely circulated draft, Tol reaches the same conclusion: “the apparent trend in endorsement is thus a trend in composition rather than in endorsement”.
The inclusion of papers into the consensus from a group of papers that is increasing over time, makes the consensus appear to increase over time.