The Grey Lady has decided to offer bullshit for readers who don’t like science. (More)
“Much else that passes as accepted fact is really a matter of probabilities”
So claims former Wall Street Journal columnist Bret Stephens in his first column at the New York Times:
Anyone who has read the 2014 report of the Intergovernmental Panel on Climate Change knows that, while the modest (0.85 degrees Celsius, or about 1.5 degrees Fahrenheit) warming of the Northern Hemisphere since 1880 is indisputable, as is the human influence on that warming, much else that passes as accepted fact is really a matter of probabilities. That’s especially true of the sophisticated but fallible models and simulations by which scientists attempt to peer into the climate future. To say this isn’t to deny science. It’s to acknowledge it honestly.
While the Times hired him as a conservative anti-Trumper, Stephens introduces his climate denialism with an anecdote that might have been written by the God-King himself:
In the final stretch of last year’s presidential race, Hillary Clinton and her team thought they were, if not 100 percent right, then very close.
Right on the merits. Confident in their methods. Sure of their chances. When Bill Clinton suggested to his wife’s advisers that, considering Brexit, they might be underestimating the strength of the populist tide, the campaign manager, Robby Mook, had a bulletproof answer: The data run counter to your anecdotes.
That detail comes from Shattered, Jonathan Allen and Amie Parnes’s compulsively readable account of Clinton’s 2016 train wreck. Mook belonged to a new breed of political technologists with little time for retail campaigning and limitless faith in the power of models and algorithms to minimize uncertainty and all but predict the future.
“Mook and his ‘Moneyball’ approach to politics rankled the old order of political operatives and consultants because it made some of their work obsolete,” Allen and Parnes write about the campaign’s final days. “The memo that one Hillary adviser had sent months earlier warning that they should add three or four points to Trump’s poll position was a distant memory.”
His argument: Because 100% certain poll-based projections were wrong about the 2016 Electoral College outcome, 100% certain science-based projections of climate change may be equally wrong.
Okay, let’s unpack that.
First, no poll-based projections gave the Clinton campaign a 100% probability of winning. FiveThirtyEight’s final projection had her a 70% favorite, and the Times’ Upshot model had her an 85% favorite. The Huffington Post and other poll-based projections had her win probabilities in the mid-80s to low-90s. The differences were mostly a matter of how models correlated individual state polls’ errors. Models that treated states as more independent, so sampling errors would cancel each other out, gave higher win probabilities. Models that assumed the states would be more correlated, so sampling errors would run the same way rather than canceling each other, gave lower win probabilities.
Because Clinton lost, should we assume the most accurate model was the one that gave her the lowest win probability? That’s an easy conclusion, and it could be correct. But not necessarily. It’s equally possible that the median projection – about 85% – was accurate … and “the other 15%” happened.
Fact is, we’ve had only a dozen or so presidential elections since the advent of scientific polling. That’s too small a sample set to draw firm conclusions on poll-based probability modeling, and especially when the underlying conditions for sample-based polling – voter participation rates among different demographics – have changed dramatically since the 1950s. Even so, the vote percentages in every state were within the polls’ published margins of error … and every pollster publishes a margin of error with every poll. Not a single pollster said “We 100% guarantee these results are exact.”
So there were no “100% certain poll-based projections.” Not one.
Second, and following from the first, the Clinton campaign never believed they were 100% certain to win. They kept campaigning right through Election Day, with appearances, ads, canvassing, and phone-banking. They may have been 100% certain their strategy gave them the best chance to win. And if the poll-based probabilities were accurate, they were probably correct. That strategy had made them at least a 70% favorite to win, and perhaps an 85% favorite to win.
Like a football coach who calls safe plays to secure a last-minute 70-85% field goal opportunity rather than risking a game-ending turnover to go for a touchdown, the Clinton campaign probably gave themselves the best possible chance to win. And “the other 15-30%” happened. That doesn’t prove the field goal was ‘really’ a 0% chance, or that a high-risk long pass would have been certain to win. It merely proves that a 70-85% favorite can still lose … and will, 15-30% of the time.
So the whole “Because 100% certain poll-based projections were wrong about the 2016 Electoral College outcome” premise of Stephens’ argument is false. The God-King’s unlikely victory does not prove that data-based planning is a suckers’ bet.
“People are getting wet, right now”
But then Stephens uses that false premise to ‘prove’ an unrelated phenomenon – that science-based projections about climate change may be equally wrong – as if the equations of thermodynamics were as novel and untested as poll-based electoral projections. And they’re not. The equations of thermodynamics have been tested and used to make reliable predictions for about 160 years. Those equations explain why Stephens’ wine stays chilled in his fridge, and why the temperature in his attic varies more than the temperature in his basement.
What’s more, you won’t find a single reputable scientist who claims to be 100% certain about climate change projections. They won’t give you a “100% certain” prediction about global temperature rise as a function of greenhouse gas emission. They won’t give you a “100% certain” prediction about sea level rise as a function of rising temperatures. They won’t give you a “100% certain” prediction about the frequency or severity of hurricanes, tornadoes, floods, droughts, coastal and sinkhole erosion, forestation loss and desertification rates, crop failures, or any of the other dangers they predict as a consequence of rising temperatures.
Instead, reputable scientists will run models and announce they’re 95% confident that exceeding 350ppm of greenhouse gases will lead to a 1.5- to 3-degree Celsius rise in mean global temperatures over 20-40 years, and 95% confident that the median of that temperature rise will cause a 2- to 5-meter sea level rise.
If you look at their papers, you’ll see they concede a small probability that 350ppm concentration of greenhouse gases could cause only a 0.5-degree temperature rise, and an equally small probability that concentration could cause a 4-degree rise. You’ll see they concede a small probability that a 2.25-degree temperature rise could cause only a trivially small sea level rise, and an equally small probability that it could cause a 6- or even 10-meter sea level rise.
So those 95% confidence-interval predictions are not “100% certain.” The actual science – rather than Stephens’ strawman – says there’s a small probability that outcomes might be less severe than projected …and an equally small probability they’ll be more severe than projected.
And if that theoretical analysis weren’t enough, there’s happening-right-now data to consider, as the Washington Post’s Eric Wemple explains:
Now to familiarize Stephens with the work of one of his news-side colleagues, Justin Gillis. Back in September, Gillis wrote a big Sunday piece seeking to shift the debate away from some fuzzy future, and plop it right in the present. He focused on flooding along the East Coast, with a dateline of Norfolk, Va. After noting that scientists had long predicted that warming trends caused by humans would cause ice melt and rising oceans, the impact is no longer a matter of projection:
Now, those warnings are no longer theoretical: The inundation of the coast has begun. The sea has crept up to the point that a high tide and a brisk wind are all it takes to send water pouring into streets and homes.
Need evidence? Miami Beach, as Gillis reported, had hatched a “$400 million plan that includes raising streets, installing pumps and elevating sea walls.” Norfolk has a $1.2 billion “wish list” to fund anti-flooding measures; and scientists have documented a boost in so-called “sunny-day flooding” on the East Coast and Gulf Coast.
People are getting wet, right now.
Scientists’ policy recommendations are based on their best mathematical models of what we can do to reduce and mitigate the effects of climate change. To suggest they’re “ideologically motivated” – as Stephens does in his column – tacitly argues that, because conservatives oppose those policy changes for ideological reasons, then scientists must be proposing those changes for ideological reasons. There’s a term for that kind of argument, and it ain’t “logic.” It’s psychological projection.
“The millions of people who agree with him on a range of issues”
So okay, Stephens is a crackpot masquerading as an intellectual. The abysmal ‘logic’ of his column proves that. But there are plenty of wingnut crackpots masquerading as intellectuals.
The charge that Stephens is a “climate denialist” is “terribly unfair,” [Times editorial page editor James] Bennet said. “There’s more than one kind of denial,” he continued. “And to pretend like the views of a thinker like Bret, and the millions of people who agree with him on a range of issues, should simply be ignored, that they’re outside the bounds of reasonable debate, is a really dangerous form of delusion.”
In other words, if “millions of people” agree with an argument that is 95% likely to be catastrophically wrong … then that argument deserves space in the Times, because “reasonable debate.”
That, again, is something the God-King might say.
Photo Credit: Wally Gobetz (Flickr)
Good day and good nuts