Think political polls are junk science? Here’s what you need to know

You’re probably like a lot of people. You look at political polls and think, “I can’t trust them.”

But maybe you’re the problem. Maybe you don’t understand how they work and how much stock to put in them.

Maybe a poll doesn’t match your perception of where a race stands, so you dismiss it (I’ve been guilty of this). Maybe you look back at 2016 and wonder how Donald Trump won the presidency when so many surveys suggested that Hillary Clinton would win. (In reality, the 2016 polls weren’t that far off.)

Or maybe you’ve heard that polls these days don’t include cellphones (they do, actually) or that pollsters wind up skipping big chunks of the electorate (they can account for this, too).

Patrick Miller, the astute University of Kansas political scientist, is a defender of polling. He says too many people don’t understand it.

He and others point to 2016 when 12 of 13 national polls conducted shortly before the election had Clinton ahead. In hindsight, that’s a catastrophe, right?

Maybe not. The polls survey the popular vote, and Clinton won the popular vote by 2.1 points. An average of the 13 final national polls showed Clinton ahead by 3.1 points. So the polls were pretty accurate.

Miller points out that final campaign polls might miss a late surge, and there’s evidence that Trump, in fact, surged.

Listen, I’m not immune to the skepticism either. I remember a mid-June poll from state Sen. Laura Kelly that had her easily winning a three-way race for the Kansas Democratic gubernatorial nomination. In that survey, rival candidate Josh Svaty was in distant third place behind Kelly and Carl Brewer. Svaty called foul, and I thought he was right. Svaty was too good a candidate to be ensconced so far back. Chalk it up to another lousy poll.

Six weeks later, the three Democrats finished in precisely that order.

To be wise polling consumers, voters need to consider the margin of error. Not every Democrat in Kansas was surveyed. That’s virtually impossible. So pollsters select what they hope is a representative sample. They attach to that an error margin to reflect that the sample could differ from the whole population.

In Missouri’s U.S. Senate race, Fox News conducted one of the most recent polls, concluding the race was deadlocked at 46 percent with a margin of error of 3.5 percent. That means Claire McCaskill could have as much as 49.5 percent of the vote or as little as 42.5. The same with Josh Hawley.

It’s a mistake to overlook those margins of error.

“Polls are supposed to be an approximation,” Miller said.

The wild card in any survey is its weighting where the pollster adjusts the demographics of those surveyed to reflect the turnout the pollster expects on Election Day. The approximation of the number of Democrats and Republicans likely to show up can vary significantly from poll to poll, and that can have a significant impact on a survey’s bottom line.

A poll is nothing more than a snapshot in time with a wider possible range than most people acknowledge. Given that, the reputable polls are still worth your time.