YUGEIn the world of politics, polls are a big deal—some might even say “yuge.” Polls can give voters clues about elections months before they get underway, and they often define the way the media talk about candidates and races.

Yet the results of the recent Iowa caucus show us that polls aren’t always predictive. Donald Trump was polling at an average of nearly five points in the lead up to the day of the election, but he ended coming in second to Ted Cruz by 3 points. Six months ago Hillary Clinton was leading Bernie Sanders in Iowa by over 50 points, only to win by a final count of less than two tenths of one percent.

This leads us to consider two important trends in presidential polling:

On the one hand, “election polling is in a near crisis” even according to those in the industry. Increasingly, polling methodologies are failing to accurately represent the modern electorate, largely due to rapid changes in technology that pollsters have been unable to keep up with. Fewer people respond to surveys, skewing the data and leading to unreliable results.

At the same time, polling has steadily become a defining force in how the media cover presidential elections. With the 24-hour news cycle and the ubiquity of social media, election news is increasingly dominated by “horse-race” coverage of whoever is “winning” via polls. Tight competition for airtime and page clicks can lead journalists to exaggerate or manufacture narratives. And when there’s a new poll out every other day, narratives swirling around today’s winners and losers can obscure broader trends and bigger pictures.

As polling has become less reliable its prominence has become more significant.

Old-school polling in a digital world

Four years ago, Gallup called the presidential election for Mitt Romney. After getting it astonishingly, badly wrong, the agency conducted an internal report finding fault not only with their own methods, but with the entire polling industry at large. The report claimed the industry “was due for an overhaul, with some of the leading firms using analog, black-and-white methods in a digital, multicultural world. Case in point: the rise of the cell phone and the fall of public engagement in opinion surveys.”

Because you can’t poll everyone, polls are only as accurate as they are representative. But getting that proper representation is becoming increasingly difficult.

Cliff Zukin, former president of the American Association for Public Opinion Research, noted in a recent New York Times op-ed that when he started working in the 1970s an 80 percent response rate was considered to be “acceptable,” but not ideal. Today, response rates are often below 10 percent.

The decreasing number of Americans using traditional landlines and the skyrocketing numbers of citizens relying on cell phones and Caller ID make it harder for pollsters to reach the broad audience they were once able to.

What makes this election especially interesting when it comes to polling is that voter and party behavior is largely bucking established trends. Anti-establishment candidates are drawing crowds by the tens of thousands and surprising the media at every turn. And this matters for pollsters because it means that traditional metrics for voter turnout could well turn out to be wrong.

For example: In the 2008 presidential election, voter turnout for 18-24 year olds was 44%. Young people flocked to the polls to vote for Obama, but there’s evidence that Bernie Sanders is drawing even more youth support than the current president did. And since youngsters are among the least reliable “likely voters,” when pollsters use that metric to make predictions, they can’t be too sure if all the Millennials “Feeling the Bern” will show up on election day—or if they won’t.

“The polls drive the media narrative” – and vice versa?

There’s been a lot of harrumphing this election season about the dominance of polls in the media coverage. Is Trump leading in the polls because he’s getting so much free airtime? Or do the media have to cover him, since he’s up so high in the polls? Whether media attention drives up poll numbers or high numbers mean more coverage is this election season’s very own chicken-or-egg predicament.150812161517-cnn-orc-poll-08-12-king-magic-wall-large-169

This predicament is important because polling can alter perceptions of electability. For many voters, it matters whether they believe a candidate has a chance to win. More media focus on polls helps legitimize certain candidacies, while suppressing others. Momentum in polling can thus be a self-fulfilling prophecy.

This problem is compounded by the effect polls have had on voters’ exposure to candidates in national debates: On the Republican side, who gets to stand on the debate stage, and even where the candidates stand, has been determined by poll placement. So there’s no doubt that those numbers matter when it comes to raising and sustaining a candidate’s national profile.

In 2016, polls are as influential as ever in our political process. Whether their prestige and importance will wane or hold firm after this year will likely depend on how many correct calls the pollsters make, or don’t. In any case, it’s inarguable that traditional polling methods are up against some seismic incoming shifts, demographically, technologically and politically.