How Did the Marquette Poll Misread Trump’s Support, Again?

In 2016, the poll suggested Wisconsin’s electoral votes would go to Hillary Clinton. This year, it got the candidate right, but the margin was still off.

A highly respected political poll missed the mark again in predicting the margin of victory for this year’s presidential race in Wisconsin and the battle for the state’s crucial 10 electoral votes. Despite being off for a second time, the poll’s nationally recognized director Charles Franklin told Milwaukee Magazine that he sees improvement over the frustrating polling from the previous presidential election four years ago. 

The Marquette University Law School polls repeatedly showed former Vice President Joe Biden holding a 5-point edge over President Donald Trump heading into Tuesday’s election. Biden likely only wished he had a margin that comfortable as his campaign had to sweat out a Wisconsin race that finished with a razor-thin margin. The election results weren’t known until the early morning hours on Wednesday. 

Biden’s margin of victory over Trump turned out to be a mere 20,517 votes of the 3.24 million votes cast, or 0.6 percentage points, according to the unofficial count. The tally was so close that Trump has already requested a recount, which is allowed when there is a margin of 1 point or less.

The final Marquette Law School poll released just before Tuesday’s election showed Biden as the choice of 48% of likely voters in Wisconsin, with Trump supported by 43% and Libertarian Jo Jorgensen receiving 2%. The poll had a margin of error of plus or minus 4.4 points. 

Franklin offered a potential explanation for the differences between the polling and the election results. 

“One is that some supporters of President Trump are now less willing to do an interview for a poll than they may have been in the past. There’s obvious reason for that,” Franklin explained. “The president complains about fake polls, about polls that are used for partisan purposes in his view and, of course, he also talks about fake news. So, his public rhetoric may be acting to give some of his supporters reasons not to participate in polling. The irony of that is it actually makes him look a bit worse in the polls than he is actually doing. It’s ultimately damaging him in the public perception of the polling before an election.”

Franklin suggested another possible reason stems from Trump’s constant criticism of absentee and early voting, which persuaded more Republicans to vote in person on Election Day.

“It’s possible that there was a bit of a surge in turnout on Election Day that we may have missed in the polling,” Franklin said. 

The 5-point poll margin in favor of Biden came about through the use of a normal “likely voter” model, which assumed a voter turnout in Wisconsin of about 3.1 million people, which would have been a record but wasn’t considered ‘extraordinarily’ high,” Franklin said. 

The poll also examined a scenario for a high-turnout election with 3.2 million voters, which approximated the number of Wisconsinites who actually cast ballots. Under that method, the margin for Biden dropped to 4 points. 

“It’s not a lot of movement, but it captured the fact that higher turnout in this election was producing a little bit better outcome for Trump,” Franklin said. “I think we captured a little more of the sensitivity to turnout.”

Franklin said the Marquette poll made changes after the 2016 presidential race that featured Trump and Democrat Hillary Clinton, including taking respondents who declined to reveal for whom they planned to vote and “allocating” them to either Trump or Biden for this election if they had somehow indicated their preference for one candidate or the other.

As it turned out, the poll came close to forecasting the percentage of voters who cast ballots for Biden, but was too low – by 3 or 4 points – in predicting support for Trump. 

In the 2016 election, the Marquette poll came “within a decimal point” of forecasting the percentage of voter support for Clinton but was nearly 8 points off on Trump’s vote percentage. 

“The good news, from my point of view, is that we cut that error at least in half between 2016 and now on Trump, but we’re still seeing this persistent understatement of what Trump is actually getting,” Franklin said. “In the coming days, we will take a further dive into that.”

Franklin noted that other polls showed gaps of 8 to 11 points in favor of Biden among Wisconsin voters. 

“So, I think we did pretty well,” Franklin said. “I would love to be perfect, but I don’t think we were badly off like the way we were in 2016. There’s no getting around the fact that 2016 was not a good result, but I think we did pretty well when compared with other national and state pollsters that were active here. Does that mean that I’m going to quit and stick with everything we are doing? No, we’re going to innovate all the time.”

Chicago-based American Association for Public Opinion Research cautioned that sweeping judgments on the polls’ overall performance before all ballots are counted are premature. A task force will perform a comprehensive evaluation of this year’s general election polls. 

“The task force will conduct a comprehensive evaluation of how the polls performed and will explore recommendations for improvement,” the organization said in a statement.



Rich Rovito is a freelance writer for Milwaukee Magazine.