It wasn’t quite “Dewey Beats Truman,” but voters who assumed the polls and predictions of a Hillary Clinton victory had a rude awakening on Nov. 9, 2016.

Clinton, who led most state- level polls throughout October, fell short in her quest for the White House. Donald Trump won the race for the presidency.

The Franklin & Marshall College Poll, one of the region’s oldest statewide polls, had her up by 11 points in the final week of October. A Quinnipiac University Poll had her up by five points in Pennsylvania a week before the election, while the Monmouth University poll had Clinton winning in Pennsylvania by 10 points in mid-October.

Ultimately, Clinton lost Pennsylvania to Trump by 44,000 votes — or less than 1 percentage point of the more than 6 million ballots cast in the state.

Similar results in Michigan and Wisconsin, where state polls had placed Clinton up by five or six points before the election, carved a large crack in the so-called Blue Wall that Democrats had counted on to win the White House.

“The state polls were way off in 2016, there’s no doubt, but the national polls had a good year. The final polling had Clinton ahead by 3%. She won the popular vote by about 2%,” said Courtney Kennedy, director of survey research for the Pew Research Center, based in Washington, D.C.

Ultimately, that would not matter. The U.S. president is elected by the Electoral College, a body of 538 electors who cast ballots based on their state’s election laws.

Narrow wins in Pennsylvania, Wisconsin and Michigan, which award Electoral College votes to the winner in their state election, handed Trump the win — despite Clinton getting more votes nationally.

Factors of difference

Kennedy sat on a post-election commission of experts who reviewed polling in the 2016 presidential election. She said several subplots played into the discrepancies between state-level polls and election results.

“There is strong evidence that there was a real late shift toward Trump among the undecideds,” she said. “They usually break along the lines with those who have decided. But they went with Trump by about 10%-15%, by double digits in places like Pennsylvania and Florida.”

Moreover, a larger than usual percentage of undecided voters late in the game made that difference count.

“And for 20 years we’ve known some groups are more likely to respond to polls — older folks, college graduates and whites are more likely to pick up and respond to a poll. So you must weight them to their ratio in the population,” she said. “In 2016, most state polls were not fixing the education imbalance. There were too many college graduates, so you had too many Clinton supporters.

“And then there was a third factor: Trump was highly successful in turning out voters we hadn’t seen at the polls before. He really increased the rural turnout in the upper Midwest.”

Although state polls consistently showed Clinton ahead in 2016, Berwood Yost, chief methodologist for the Franklin & Marshall College Poll, said there were suggestions in the data that change was afoot, even as forecasters proclaimed a Clinton win a foregone conclusion.

“Frankly, the polls from 2016 at the state level performed the way they usually do, when you looked at the poll numbers and all of the things they were showing. They showed two candidates who were not very well-liked, a lot of undecided voters, third-party voters and general dissatisfaction with the incumbent and the incumbent’s party,” he said. “There were a lot of things that were set for a Republican win.”

Like Kennedy, he said much of the discrepancy between poll numbers and the final count arose from things that happened at the very end of the campaign.

G. Terry Madonna, director of the Franklin & Marshall poll and a veteran of three decades of Pennsylvania polls, said the October 2016 poll had Clinton ahead of Trump 49% to 38% among likely voters a week before the election. That poll, however, didn’t reflect the impact of then-FBI Director James Comey’s Oct. 28 letter renewing questions about Clinton’s private email server.

So-called “shy Trump voters” — poll respondents who were hesitant to commit to the unconventional candidate — also may have skewed results, as did higher-than-usual turnout in areas that would come to be called Trump Country.

Yost and Madonna said pollsters have become ever more cautious about building their samples and ensuring they are weighted to the reality of voters.

Madonna noted the college’s most recent poll found former Vice President Joe Biden ahead of Trump by 6 points among likely voters. That put the poll in the middle of pack of national polls that saw Clinton’s margin vary from 4 to 9 points.

Although early polls typically rely on a random sample of registered voters adjusted for age, geographic area and party registration, final polls as Election Day draws near always focus on likely voters. Pollsters use a variety of questions, including interest level in the upcoming election and voting in the last three elections, to determine who should be considered likely voters. It’s yet another process designed to hew closer to reality.

So, are polls reliable?

Experts say it is important to realize that even the best poll is only a snapshot of public opinion at a given point in time. While they may suggest the way voters are moving, they are not predictive.

On the other hand, organizations that aggregate polling results sometimes rely on algorithms that take various polls into account in their predictions.

Kyle Kondik is managing editor of Sabato’s Crystal Ball at the University of Virginia Center for Politics. The Crystal Ball, an oft-cited political report, routinely issues projections as campaigns move toward Election Day. Although experts there follow the polls, Kondik said they don’t rely on any mathematical models for their projections.

“Our projections are qualitative. It’s up to us to decide how to call an election,” he said.

But the lessons of 2016 and concerns about the ability of pollsters to count hard-to-reach voters in areas such as the rural upper Midwest, even as Biden’s lead appears more stable than Clinton’s was four years ago, have given experts at his center pause this year.

“We’re trying to convey a certain amount of uncertainty in this election,” Kondik said.

But like others, he sees this as a more traditional year in terms of opinion polls. While Biden’s edge is about the same 5 points many polls gave Clinton over Trump at the same point in 2016, Kondik said there are fewer undecideds this year. And fewer voters are saying they support a third-party candidate this year. Jo Jorgensen, a Libertarian, is running, and Green Party candidate Howie Hawkins is not even on the Pennsylvania ballot.

Kennedy points to state polls from 2018. She said they were back on track for the 2018 congressional elections and appear to be doing a better job this year after the lessons of 2016.

“But all they’re telling us now is that the race is competitive,” she said.

Copyright ©2025— Trib Total Media, LLC (TribLIVE.com)