“There are three kinds of lies,” British Prime Minister Benjamin Disraeli is purported to have said. “Lies, damned lies, and statistics.”
Disraeli would have enjoyed both the tone and the premise of economist Gary Smith’s often funny and frequently enlightening new book for a lay audience, Standard Deviations: Flawed Assumptions, Tortured Data, and Other Ways to Lie With Statistics. With the help of prognosticating octopi, dead salmon, a host of rigged charts and all manner of creative bookkeeping, Smith presents case study after case study on how to twist the facts, exploring how phenomena such as confirmation bias, regression towards the mean, and the law of averages truly inform our lives.
The author, who has long been collecting examples of statistical abuse in his day job is as a professor at Pomona College, initially conceived of his book as an updating of the classic 1954 tome on gaming the numbers, How to Lie With Statistics.
But the professor, who started teaching econ at Yale back in 1971, soon realized there was a bigger story about “a lot of the mischief that’s happened since the 1970s.” As we’ve all been told, we’re now living in the era of Big Data, that avalanche of information made possible by technology. And in its wake, data mining, or the practice of combing through these mountainous stats until patterns emerge and new insights are supposedly gained, has become a national pastime.
There’s just one problem with that, Smith says. We're going about the whole thing backwards.
“To be accused of data mining used to be a slur,” Smith said. “That’s been totally turned on its head. Now it’s a badge of honor.”
Indeed, the author devotes much of his book to debunking the vogue for putting the data cart before the theoretical horse, or what he calls “data without a theory.”
“Data are just data,” he writes in Standard Deviations. “Even if we see a clear and obvious pattern, we still need a logical reason for believing that the pattern is something more than mere coincidence.” Play the game the other way round and the results could be “treacherous.”
“All statistical theory is based on the assumption that you have the theory first,” Smith said.
As many social scientists have been trying to point out to us, it’s become disconcertingly easy to ransack data from every vantage point until something emerges: Poor housekeeping turns people racist, those given “positive” initials like ACE live 3 to 5 years longer, and the stock market goes up when an NFC team wins the Super Bowl (yes, these are real examples, all referenced in the book). Just keep fiddling with the numbers long enough, Smith says, and “you’ll always find correlations, even in random patterns.” Drawing conclusions from such becomes a rather dicy proposition, especially as there are “enormous social consequences” to this type of research.
Sometimes questionable studies mainly proffer food for thought — did you hear the one about how hurricanes with female names are more destructive? — but when it comes to medical procedures or public policy, the results can become dire. Luckily, as with those hurricanes, if readers wait a minute, a trained statistician will likely come along to poke holes it. Science, like breaking news, often lives or dies by a second opinion.
But in the meantime, common sense can be the best defense. If the claim that Asian-Americans that are more susceptible to heart attacks on the fourth day of the month gives you pause, well, you might be on to something.
“There is this notion that’s always been with us that a theory, no matter how strange, that comes out of the mouth of a PhD or out of the brain of a supercomputer must be true,” Smith said.
But we must also not be so quick to trust our own brains, which are innately hard-wired to spot patterns and create narratives around them. In other words, one element underlying this fascination with Big Data is our species' deep animosity to the fickleness of chance, and the concomitant idea that we can quantify away risks and flukes that are built into the system. The good news there is that no matter how many times you get onboard an airplane, your chances of coming down the wrong way never go up. Likewise, if you lose some you might or might not turn around and win some. The law of averages just doesn’t operate that way, no matter what story we tell ourselves.
“One ball hits the goal post and bounces off, one ball goes in, and afterwards people go around saying, ‘My country’s better than yours,’” Smith says with a laugh. “We’re not comfortable with the idea that our heroes win or the villains lose because of luck.
“But randomness makes life interesting,” he continues. “If you know everything that’s going to happen to you in your life, what’s the point of living it?”
This month Smith discusses Standard Deviations on WNYC.
Public Spectacle, L.A. Weekly's arts & culture blog, on Facebook and Twitter: