Early Wednesday morning, FiveThirtyEight’s Harry Enten published a statement of defeat: “Bernie Sanders made folks like me eat a stack of humble pie on Tuesday night.” He was referring to, of course, the massive upset in the Michigan primary—no polls had Sanders anywhere near grasping distance of that win, and FiveThirtyEight itself had put Hillary Clinton’s odds of winning the state at greater than 99 percent. But here he was, somehow, the winner.
On Tuesday evening, Trump handily won primaries in Michigan, Mississippi, and Hawaii, Ted Cruz in Idaho, and Clinton in Mississippi’s Democratic primary. Wednesday morning’s takes, however, are by and large about the Michigan upset.
After every caucus and primary, every unpleasant data point, pundits take to their preferred medium to write their version of “Here’s Why the Iowa Polls Were So Off.” The exhausting mistake that everyone keeps making is that the polls had an isolated moment of unreliability. But the truth is that polls don’t really work all that reliably anymore; they are like the unidentified cord that you keep in a desk drawer even though it doesn’t plug into anything, or your grandfather’s penis.
In 2015, Nate Silver, founder of the aforementioned blog and statistician who has ostensibly devoted his entire career to the poll, wrote a short post entitled, “The World May Have a Polling Problem,” in which he explained the growing inaccuracy of once-revered data.
But there are lots of reasons to worry about the state of the polling industry. Voters are becoming harder to contact, especially on landline telephones. Online polls have become commonplace, but some eschew probability sampling, historically the bedrock of polling methodology. And in the U.S., some pollsters have been caught withholding results when they differ from other surveys,“herding” toward a false consensus about a race instead of behaving independently. There may be more difficult times ahead for the polling industry.
Beyond that, response rates are terrible—in the ‘90s, about 35 percent of those contacted would complete a survey, in 2014 that number had dropped to 10 percent (some of this thanks to caller ID); they are expensive to conduct; it is often illegal or expensive for pollsters to contact civilians on cell phones, meaning large portions of the population are excluded.
Still, we continue to pay attention to their results—often, it seems, in order to bolster our own bias. When a poll confirms accepted knowledge, polls are an amazing tool we have at our disposal! When a poll is off, it is an anomaly to be ignored.
One idiot powerfully proves this:
In the Wall Street Journal, political analyst Michael Barone writes that polls aren’t hard data—instead, they are clues:
“Particular poll numbers are like daubs of pigment on an Impressionist’s canvas, which by themselves don’t convey a sense of reality but which, taken together with many others, can give the aesthetically sensitive viewer a more vivid sense of the underlying reality than the most accurate photograph.”
He continues: “Technological change may be making polls less scientifically reliable, but reading polls has never been entirely a science; it has also been an art and seems to be getting more so.”
Still, polls are one of the very few bellwethers we have in a basically incomprehensible election; I’d imagine we all feel more comfortable following incorrect highway signs than nothing at all.
Contact the author at firstname.lastname@example.org.
Image via Getty.