A year ago this month, George Will wrote a howler of a column in the Washington Post about global warming, loaded with scientific errors and profoundly illogical arguments. It would not have survived even the most perfunctory fact-checking–despite claims from the Washington Post that his columns go through a “multi-layered fact checking process.” In subsequent months, Will has continued to offer new climate howlers, and this Sunday he provided us all with a dubious one-year birthday gift.
In Will’s latest piece, he yet again declares global warming a construction of hysterical climate scientists who, in his words, “compound their delusions of intellectual adequacy with messiah complexes.” This time, he claims that climate scientists themselves are finally confessing that it’s all been a whole lot of hooey.
Will backs up this claim with a link to a BBC interview last week with Phil Jones of the University of East Anglia. A BBC journalist asked Jones questions, some of which had been submitted by unnamed climate skeptics, including this one:
Q: Do you agree that from 1995 to the present there has been no statistically-significant global warming?
A: Yes, but only just. I also calculated the trend for the period 1995 to 2009. This trend (0.12C per decade) is positive, but not significant at the 95% significance level. The positive trend is quite close to the significance level. Achieving statistical significance in scientific terms is much more likely for longer periods, and much less likely for shorter periods.
This statement then got run through a sausage grinder run by journalists who are apparently both innumerate and illiterate. The Daily Mail declared,
“This week the unit’s former head Professor Phil Jones, performed a majot [sic] u-turn and admitted there had been no ‘statistically significant’ global warming in the last 15 years.”
This version of the story, which makes Jones sound like he was making a confession under enhanced interrogation techniques, ended up on the Wall Street Journal editorial page and today in George Will’s column:
Global warming skeptics, too, have erred. They have said there has been no statistically significant warming for 10 years. Phil Jones, former director of Britain’s Climatic Research Unit, source of the leaked documents, admits it has been 15 years.
Will doesn’t tell us exactly who these skeptics are who claimed there had been no “statistically significant warming” for 10 years. I have no way of knowing if they in fact exist. Will himself has been loudly beating the “no-warming-for-a-long-time” drum over the past year. But he has backed up this claim simply by searching for the hottest single year in recent history. “According to statistics published by the World Meteorological Organization, there has not been a warmer year on record than 1998,” he wrote in April. Will continued to claim that global warming has stopped since 1998 even after the secretary general of the World Meteorological Organization wrote into the Post to explain why Will was wrong.
In his latest column, Will added the fancy, shiny new term “statistically significant” to his claim that there has been no global warming. But in doing so, he misleads his readers about what statistical significance actually means.
To see why, take a look at this graph. NASA scientists have been building it for years now, using weather records from around the world. Other graphs built by other teams of scientists have produced similar patterns. If you only look at a small vertical slice of the graph, you’ll see the temperature jump up and down and up again. That’s the sort of pattern you’d expect from a system as big and noisy as the planet’s climate. There are lots of sources of variations in the average global temperature, such as El Nino, a natural oscillation in the movement of heat in the oceans.
Sometimes these hopping temperatures don’t seem to go anywhere in particular. In other cases, there appear to be trends lurking under the noise. To test a hypothesis like this, scientists estimate how likely it would be for an apparent trend to be nothing more than the noise in the climate system. They then set a threshold for those odds.
In many branches of science, researchers set that threshold at 5%. In other words, if there’s only a 5% chance that a particular pattern of temperatures was the result of pure noise, scientists will call the trend “statistically significant.” If, on the other hand, the probability turns out to be 5.1%, the trend is still likely not to be the result of noise, but it’s not officially statistically significant.
“The boundary of .05 should be seen as a guide to interpretation, not as a clear boundary between truth and fiction,” Michael Whitlock and Dolph Schluter write in their book, The Analysis of Biological Data.
Just because a trend over a particular stretch of time doesn’t quite meet the 5% cutoff doesn’t necessarily mean it’s not real. It just means that scientists cannot reject the null hypothesis that noise is the cause. One way scientists can deal with this challenge is to look at longer sets of records. In the case of climate, looking at longer stretches of time reveals that there is indeed a real trend of warming temperatures. Just because the BBC’s questioner arbitrarily set the cutoff for analyzing the climate at 1995 doesn’t change that fact. Jones openly addressed this fact, but George Will conveniently omitted it.
Significance is one of the basic concepts of statistics that everybody should learn about. We rely on these concepts to judge not just the state of the climate, but also the meaning of clinical trials of drugs, the conclusions of psychology experiments that help reveal the inner workings of the mind, and all manner of other discoveries. In today’s column, George Will isn’t just making misleading statements in the service of trying to foster doubt about climate change. He’s also helping to muddle our collective scientific literacy. Why the editors of the Washington Post’s editorial page want to be a party to this is a mystery to me.