How we discovered the 97% scientific consensus on man-made global warming

As part of our work on "The Consensus Project" study, I said:

"We want our scientists to answer questions for us, and there are lots of exciting questions in climate science. One of them is: are we causing global warming? We found over 4000 studies written by 10 000 scientists that stated a position on this, and 97 per cent said that recent warming is mostly man made.”

The journal we published in, Environmental Research Letters has a word limit and despite publishing loads of extra info since then, we haven't talked through some of the conversations we had on the way. There have been lots of comments asking why we considered the scientists we did, and how we judged whether they endorsed that most of the recent global warming was man-made or not.

A brain surgeon for brain surgery, a climate scientist for climate science

Our first thought was: who should count? We decided that researchers who work and publish on climate science are the right group to ask. Just like you wouldn't ask a dentist to do brain surgery we thought that it was better to ask those who research climate change rather than the Big Bang.

These are the guys and girls who spend all their working lives studying the climate and who stay up-to-date with new research. If something is discovered then they will read about it, put it in context, and judge how strong the evidence is.

To us, we were looking at the experts who could spend enough time and would know enough to judge all of the evidence for themselves. Not everyone has done new research showing that the world ain't flat, but scientists who need to know about the shape of the Earth have checked all the evidence and they're convinced that it ain't.

Second: how does research show whether it endorses or rejects man-made global warming? We found that many studies don't put in their summary an explicit statement that they've found that Earth is round rather than flat, that the atmosphere is made of air rather than cheese, that they determined that 1+1=2 or that they've calculated that climate change is man-made.

Let's say that some scientists are working on extreme weather in Japanese cities, and they say that the number of hot days are changing, and that they are interested in looking at what are the 'probable changes' in future due to 'predicted global warming due to an increase in atmospheric greenhouse gases'.

We reckoned that scientists like this were implying that man-made global warming is real when they wrote that they expect future warming from greenhouse gases to have an impact.

This isn't as 'explicit' as saying something like 'we find that since the nineteenth century, greenhouse gases, not solar irradiance variations, have been the dominant contributor to the observed temperature changes'. This is why we had different categories for how strong the endorsement or rejection was.

The atmosphere isn't made of cheese

One argument we've noticed is that the opinions of those Japanese scientists should be ignored and shouldn't be counted when trying to work out whether there's a scientific consensus on man-made global warming, because they don't explicitly calculate the cause.

But what if you apply that to other ideas, like whether the world is round or flat, or whether the atmosphere is made of air rather than cheese? A quick search through google scholar for studies on the atmosphere finds nothing on the first page where the authors explicitly say that they measured that the atmosphere isn't made of cheese. But they implicitly reject that idea when they work as if it's made of air. We reckoned that if scientists judge the evidence on a subject to be very strong, if there is a consensus, then they will work as if it's true and their work should be counted as an implicit endorsement.

Among those who want to try to hide the consensus, one trick is to try to rubbish or hide away as many studies as possible. They claim to make the consensus disappear by saying that scientists who say that climate has warmed and greenhouse gases will cause future warming do not think that recent global warming is man-made. This is an example of impossible expectations, one of five common features of denial movements as explained by Dana Nuccitelli.

The same tricks could be used to 'disappear' any scientific consensus: using their techniques, the consensus that the atmosphere isn't made of cheese could also be disappeared.

Who knows better about a study: the scientist who did the study, or Joe Blogger?

We only looked at the summaries (or 'abstracts', in journal-speak) because we had thousands of them to look at. They have limited space, so they don't always report everything that a study has found, but there was no way we had time to read thousands of full papers.

So we asked the scientists who wrote them to tell us what their full paper said about the cause of global warming. 1,189 scientists were kind enough to answer for 2,142 of their papers.

It turns out that our study of the summaries was very conservative. We thought that there were 18 papers, or just 0.8%, showing the most explicit endorsement of man-made global warming, although 36% had less strong endorsements. It turns out that the authors reckoned that more like 10% of their papers represented the most explicit type of endorsement possible once you went beyond the summary we looked at: more than 10 times as many as we reported.

The overall consensus remained about the same: 97% whether you look at just the summaries or at the whole paper, but often the scientists said that the full paper was a stronger endorsement than we'd found in the summary.

Commentators who've said that we should ignore the 'implicit' endorsements like those Japanese scientists have been curiously quiet on what the scientists themselves said.

Do you think that gravity is real, or is it 7% real?

Finally, where did we get the 10,000 scientists from? Our recipe was simple: if a scientist published a paper that endorsed or rejected, and hadn't published any paper saying the opposite, then they were added to the endorsement or rejection pile.

Let's take two examples: Professor Kevin Trenberth and Dr Roy Spencer.

Professor Kevin Trenberth has published over 200 papers in climate science, and our search terms found 14 studies, of which one is the highest level of explicit endorsement, and a total of 7 are endorsements of any level. The other 7 are 'no position'. In our ratings, that puts him as an endorsing scientist, and he's added to the pile of 10,188 endorsing authors.

There is an argument that he is only 50% endorsing, or if you believe that only the most explicit endorsements count, then only 7% of him should count as an endorsing scientist. We think that's the wrong way of going about it. Just because he wrote a paper saying that satellites were underestimating global warming but didn't talk about whether it was man-made doesn't mean that 7% of him is convinced that we don't know whether man-made global warming is real!

The example of Dr Spencer is the same, but in the opposite direction. We found 12 studies by him including 11 'no position' and 1 'rejection'. This meant we added him to the pile of 128 rejecting authors, even though only 7% of his papers were involved in that judgment. This is opposite to what he said in testimony to the US congress.


Testing scientific opinion isn't as easy as you'd think, but we're convinced we found a solid way of doing it. If a study assumes that man-made global warming is true, then the authors have obviously judged the evidence and should be counted as an implicit endorsement. Their judgment shouldn't just be thrown out: if we did that then we'd end up with claims that there is no consensus over whether Earth is round or the atmosphere is made of air rather than cheese!

If a scientist has published work saying that man-made global warming is real, but they also published a study that says nothing about the cause of global warming, then we should only count the bits where they talk about the cause, whether endorsing or rejecting. Otherwise you end up with the idea that someone is only 8% or 50% endorsing of an idea, which probably doesn't reflect them very well.

Anyway we didn't want you to just trust us: in the paper and our data we showed what the scientists who wrote the papers said. According to them, we were too conservative. We reckon 97% consensus and the scientists themselves reckon a 97% consensus, but their endorsements were on average stronger and more explicit than we estimated.

Posted by MarkR on Friday, 1 November, 2013

Creative Commons License The Skeptical Science website by Skeptical Science is licensed under a Creative Commons Attribution 3.0 Unported License.