Pollsters have made a few monumental misses in 2016.
In March, when polls completely missed Bernie Sanders’ upset win over Hillary Clinton in the Democratic primary in Michigan, one political editor called it “among the greatest polling errors in primary history.”
Then in June, the world woke up to the shock of Britain’s decision to leave the European Union after polls had overwhelmingly predicted it would remain.
Now as the presidential race enters the home stretch and new polls are released daily, voters are left to wonder – how accurate are they?
WCPO spoke to political and polling experts about the accuracy of polls, how to read them and the best way to predict who will win the presidential contest between Clinton and Republican nominee Donald Trump.
Here’s their advice:
Focus on swing state polls
“Keep in mind, the presidential election is not a national election; it is a state-by-state election,” said University of Akron political science professor Dave Cohen. “It is won or lost in a dozen or so swing states. You need to focus on those swing states if you want to figure out who will win that election.”
Watch polls among likely voters in Ohio, Pennsylvania, North Carolina and Florida.
Don’t focus on individual polls
Experts watch trend lines over multiple polls to get a better prediction. Each poll company uses a different methodology (who they poll and how) so results can vary widely. Cohen prefers websites that offer poll averages and trend lines. His favorites: Nate Silver's FiveThirtyEight and RealClearPolitics.
Be skeptical of polls immediately after a debate
Clinton may have gotten a bump after the first debate on September 26, but it may even out in a week.
“There is an immediate reaction, but then there will be a settling effect,” said Xavier University’s director of government relations Sean Comer.
Voters are changing their minds very quickly this election, making them harder to poll, said Michelle Henry, president of the Center for Marketing and Opinion Research in Akron.
“These polls are just … a snapshot in time done in day or two; they do not take into account what’s happening in society during a larger frame of time,” Henry said.
But those polls might matter this year
With both parties pushing voters to cast their ballots ahead of Election Day this year, Miami University political science professor Chris Kelley predicts polls taken after events, such as a debate, will be more telling.
“I’m more and more a believer that early voting is going to throw our models in disarray,” Kelley said. “When you have early voting, people are inclined to vote based on some impressionable moment.”
A poll released soon after a major gaffe or poor debate performance could influence how people vote. If voters see a poll that reflects their feelings, they may be more likely to send in an absentee ballot.
“They’re going to say: ‘Wow – that was striking’ and then they’re going to see polling that shows up in the next day’s news that shows they’re not alone,” Kelley said.
Finding likely voters
Polls of likely voters are considered the most accurate (more so than polls of registered voters). Yet finding likely voters is a big concern of pollsters this year. With two historically unpopular presidential candidates, voter enthusiasm is hard to measure because it constantly changes.
“People are more likely to vote the more engaged they are," Henry said. "If a candidate fails to impress me, I might not be as certain anymore. Typically we don’t see so many ups and downs … it’s just so hard to gage.”
Polls are less volatile closer to the election. Many undecided voters, and those considering a third party candidate, reach a final decision as the election nears. Many want to cast a vote of consequence, so even though they might dislike Trump or Clinton, they could vote for them anyway. And once that decision is made, it is hard to get voters to change their minds.
“The closer it gets to an election, the more people move out of the undecided column,” Comer said.
Millennials are hard to reach
Especially if they’re men, said Lauren Copeland, associate director of the Community Research Institute at Baldwin Wallace University.
“For whatever reason, the population is really hard to reach,” Copeland said of men between the ages of 18 to 29.
The university is partnering with the Cleveland Plain Dealer to conduct polls this fall. Because of the newspaper’s popular sports section, she’s been surprised by how many men are willing to take the polls.
But that’s unusual. She’s found older women are more likely to pick up phone calls from pollsters and share their opinions. Meanwhile millennials often don't have time for pollsters because they might be working multiple jobs, going to school or caring for young children, Henry said.
Copleland pins faulty polling during the Michigan Democratic primary on pollsters’ failure to reach young men.
“What happened in those polls is that millennials were under surveyed – particularly young white men ages 18 to 29,” Copeland said.
Beware of online polls
Telephone polls are still the best way to reach a good sample audience, Henry said. Online polls might only capture opinions of people who find or seek them out.
The biggest change for pollsters in the last 20 years has been the advent of cell phones. Nowadays half of households don’t use land lines, especially younger voters, she said. Polling cell phone users is more expensive and time consuming because auto-dialers are not allowed, so pollsters have to manually call each number. So relying on reputable polling firms, who pay for cell phone calls, is key.
“Industry research show us, if you do not include cell phones in a sampling frame, the results will not be accurate or reliable,” Henry said.