Home Subscriber Only MU’s Franklin says polling is not ‘broken as a methodology’

MU’s Franklin says polling is not ‘broken as a methodology’

Charles Franklin
Charles Franklin

In its final polling average of Wisconsin before the election, the website FiveThirtyEight gave President-elect Joe Biden a 52.1% to 43.7% lead over President Donald Trump. The unofficial results following Election Day showed a much tighter race in Wisconsin. Biden led Trump 49.6% to 48.9%. A little more than 20,000 votes separated the two candidates.

Already a subscriber? Log in

To continue reading this article ...

Subscribe to BizTimes today and get immediate access to our Insider-only content and much more.

Learn More and Subscribe Now
Arthur covers banking and finance and the economy at BizTimes while also leading special projects as an associate editor. He also spent five years covering manufacturing at BizTimes. He previously was managing editor at The Waukesha Freeman. He is a graduate of Carroll University and did graduate coursework at Marquette. A native of southeastern Wisconsin, he is also a nationally certified gymnastics judge and enjoys golf on the weekends.

In its final polling average of Wisconsin before the election, the website FiveThirtyEight gave President-elect Joe Biden a 52.1% to 43.7% lead over President Donald Trump.

The unofficial results following Election Day showed a much tighter race in Wisconsin. Biden led Trump 49.6% to 48.9%. A little more than 20,000 votes separated the two candidates.

“The elephant in the room is the polling outcomes,” Charles Franklin told a monthly meeting of the Greater Milwaukee Committee almost a week after the election.

Franklin, director of the Marquette University Law School poll, said by his estimates, state polling averages around the country were off by around 6 percentage points, although he noted that could change as vote totals are finalized.

The miss was even worse in Wisconsin. Franklin estimated it at 9 points while FiveThirtyEight’s average was off 7.7 points.

By comparison, Franklin’s Marquette poll performed better. Its last release estimated a 50-45 lead for Biden after accounting for undecided voters leaning toward a candidate and allocating those who wouldn’t say based on their favorability views of each candidate.

Franklin noted that there was some amount of bifurcation in Wisconsin polls with a group consistently in the 8- to 11-point Biden lead range while Marquette was consistently around a 5-point Biden lead.

“Exactly why there was that gap between our polling and theirs, and what accounts for that, we don’t know yet,” Franklin said, adding that it is something for those other pollsters to address.

Marquette Law School poll track record

[caption id="attachment_516085" align="alignnone" width="1280"] Source: MU Law poll, Wisconsin Election Commission, AP unofficial results[/caption]

But the 4.3-point miss stands out in what is generally a strong record for the Marquette poll since it started in 2012. The problem is compounded by the fact that the poll’s other major miss came in the 2016 presidential election. That year, the poll gave Hillary Clinton a 6-point lead heading into the election only to see Trump win by 0.7 points.

“I think that the bottom line is that for the second (presidential) election in a row, polling did have substantial misses, but this time, the misses were across the country, whereas in 2016 they were more focused here in the Midwest,” Franklin said.

“What we still see is a persistent understatement of the support for President Trump,” he added.

In both 2016 and 2020, the poll was fairly close to the Democratic candidate’s vote share, but underestimated Trump by 7.2 and 3.9 points, respectively.

Mike Gousha, a distinguished fellow in law and public policy at Marquette Law School, joined Franklin for the GMC meeting and said his conversations with Democratic strategists suggest that the support missing from polls comes from a group of voters who are quite motivated to vote for Trump and have a deep distrust of the nation’s institutions, including polls and many of them decline to participate.

Franklin said his conversations with Democratic and Republican pollsters suggest their internal polling was off too.

“One of the big things about this is that we’ve missed these people despite looking for them,” Franklin said. “It’s not that polling took place this year completely ignoring the possibility that we were missing people, but that as we tried to locate them, tried to find indicators that we were missing them, by and large we didn’t see those indicators.”

Franklin was asked during the meeting to comment on Trafalgar Group’s polling, which showed the Wisconsin race as even in its final poll.

“Trafalgar is, forgive me, not very transparent about exactly what they’re doing and that’s annoying in the polling industry. We generally share our methods relatively openly,” Franklin said.

He acknowledged Trafalgar had managed to reach Trump supporters, but also said it was important to look at the firm’s entire record. Its final polls had Trump up 2 points in Pennsylvania and 3 points in Michigan. As of Nov. 16, Biden was leading those states by 1.0 and 2.6 points, respectively.

Despite misses in two straight presidential elections, the Marquette poll did perform well in 2018. It showed a 47-47 tie between Tony Evers and Scott Walker. Evers won by 1.1 percentage points. The poll also had an 11-point lead for Democrat Tammy Baldwin, who ended up beating Republican Leah Vukmir by 10.9 points.

“That’s why I don’t think polling is broken as a methodology, but it sure does look like we have a big problem of reaching, whether we want to think of them as President Trump supporters specifically or this group of people within that realm that are … sort of low-propensity voters but very dedicated to President Trump,” Franklin said.

Stay up-to-date with our free email newsletter

Keep up with the issues, companies and people that matter most to business in the Milwaukee metro area.

By subscribing you agree to our privacy policy.

No, thank you.
Exit mobile version