Understanding the Science of Polling | Opinion

At this time in the 2016 campaign, just about every serious political writer—I count myself as a notable exception—was absolutely certain Hillary Clinton would become the first woman to be elected president of the United States. Why? Because the polls said so.

There's been a lot said and written over the past few weeks about how wrong those polls proved to be, and by inference, how wrong they remain. There's plenty of evidence that their predictive value has declined over recent decades but there's more to it. Many people, including a lot who cover elections, don't understand what goes into the science of polling and are at the mercy of others to interpret the data.

Ultimately, it's the electorate that loses. There's no example worth pointing to of a pollster whose results didn't attract notice in subsequent election cycles because they'd been so far off the mark in previous ones. Unlike the commercial marketplace, the political universe doesn't punish people who got it wrong.

This matters because so much political reporting, especially at the national level, focuses on polls. Sometimes that's because—and this is something I consistently have to warn myself to guard against—political writers often have a rooting interest in the outcomes.

Remember the bumper sticker popular in 1992: "Annoy the Media. Re-elect Bush." Voters get it, at least some of them. Analysts with excellent reputations including the Washington Examiner's Michael Barone—likely the best among us—have written recently about how many polls going back many years overestimated the strength of Democrats running for president, U.S. Senate, and the House while undercounting Republican support.

That has had a considerable impact on campaigns and strategy. Many donors are more likely to give to candidates they believe are winning than losers. Campaign volunteer attitudes are shaped. Many are inspired to work harder for candidates who are ahead, while those working for those perceived to be running behind begin to question the use of their time. It may even affect turnout because of what it does to voter enthusiasm.

Too many political writers cover elections from the same perspective sportswriters cover football. There's color and nuance in both kinds of reporting to be sure, but the numbers drive the narrative.

So, can you believe a reporter, for example, who writes that Democratic Pennsylvania Lt. Gov. John Fetterman is going to beat the GOP's Dr. Oz because all the polls have him ahead? No, because the race has been so close and because it's getting closer. Remember, compared with how people voted, the Keystone State is one of those places where Democrats consistently outpoll Republicans and then lose. A Fetterman win is by no means a certainty.

There are plenty of races this cycle where that holds. Published polls work to the benefit of the Democrats against the Republicans. America remains a center-right nation, as Gallup has consistently shown. Polls regarding issues, parenthetically, tend to be more reliable than polls testing electoral strength, even as partisan voting patterns strengthen. The number of "independents" may continue to be on the rise but there are consistently fewer "ticket-splitters" at the polls.

Intent and intensity matter. Many analysts, again because of a rooting interest, have written volumes about how the U.S. Supreme Court ruling in Dobbs that overturned Roe v. Wade will boost Democratic turnout. Perhaps, but it's a safe bet the 1970s-style inflation Joe Biden brought back is giving plenty of Republicans and independents reason to go to the polls intending to put a check on Bidenomics. Which issue has more power? Well, and yes, you can use the polls to predict whatever you want—more Americans, more registered voters, and the voters who are believed to be most likely to vote in the next election are all more concerned about inflation than abortion.

Pieces of data like that are too often dismissed. It's how people vote that counts, not what they tell pollsters. One more bit of data to consider: In states with contested races on both sides of the aisle, Republican candidates got many more votes over what they did in 2018—the last midterm election—and the Democrats got less. That ought to tell you something next time you look at a poll.

Newsweek contributing editor Peter Roff is a veteran Washington journalist and media fellow at the Trans-Atlantic Leadership Network. He is a former writer and columnist for U.S. News & World Report and United Press International. Reach him by email at RoffColumns@GMAIL.com. Follow him on Twitter @TheRoffDraft

The views expressed in this article are the writer's own.