Charles Franklin, director of the Marquette Law School Poll, speaks in a Wednesday virtual luncheon hosted by the Milwaukee Milwaukee Press Club. Franklin gave a postmortem of polling errors in Wisconsin for the 2020 presidential election and talked about why virtually every poll seems to miss a significant segment of Donald Trump's supporters. (Screenshot)
Charles Franklin, director of the Marquette Law School Poll, speaks in a Wednesday virtual luncheon hosted by the Milwaukee Milwaukee Press Club. Franklin gave a postmortem of polling errors in Wisconsin for the 2020 presidential election and talked about why virtually every poll seems to miss a significant segment of Donald Trump's supporters. (Screenshot)

Wisconsin’s most prominent pollster gives his theories on why presidential election polls were so wrong in 2016 and 2020.

Donald Trump broke countless presidential norms during his term in office: Firing top government officials via a tweet. Claiming a president is immune from conflicts of interest. Not reading presidential intelligence briefs. Saying there were “very fine people on both sides” of a white supremacist rally and counter-protest. Refusing to concede the election when it’s clear he’s lost.

Marquette Law School Poll Director Charles Franklin on Wednesday added one more shattered norm to the list: the accuracy of public polling.

“There’s more evidence now than there was four years ago that we’re missing a small, but not trivial, group of people who support Trump but are not strong Republicans and may not participate in elections outside of ones where President Trump is on the ballot,” Franklin said in a virtual luncheon hosted by the Milwaukee Press Club.

Just as they were in 2016, pollsters, including Franklin, were way off on estimating Trump’s electoral performance in states like Wisconsin. Pollsters across the spectrum had President-elect Joe Biden up by an average of 8.4% over Trump heading into the Nov. 3 election; Biden ended up winning the state by just 0.6%. In 2016, polls had Hillary Clinton ahead of Trump by an average of 5.9%; Trump carried the state by 0.7%.

While the polling errors are massive and glaring, what Franklin said frequently gets overlooked—at least as far as his poll is concerned—is that Biden and Clinton performed incredibly close to their predicted vote shares. Marquette’s final 2020 poll showed Biden getting 48% of the vote (he ended up with 49.4%), and the final 2016 poll showed Clinton getting 46% (she got 46.5%). 

Franklin said Trump’s pool of poll-blind supporters suggests there truly is something intangible about Trump and his followers that systematically tricks pollsters. 

“If a small segment of Trump voters are systematically declining to participate in the poll, that can account for why we are systematically understating Trump’s vote,” Franklin said.

Some major Wisconsin polls had Biden with a gaudy lead. The most widely cited example is a Washington Post-ABC News poll released Oct. 28 that showed Biden leading Wisconsin by a 17-point margin, 58% to 41%; another from the New York Times and Siena College released on Nov. 1 put Biden ahead by 11 points, 52% to 41%. 

Trump himself called out the Washington Post poll in a Wednesday morning tweet as “a possibly illegal suppression poll,” a made-up term that does not exist. Franklin said the Washington Post poll was “so obviously an outlier” that it can be completely disregarded, but it remains “a little bit of a puzzle” as to why most Wisconsin polls were so far off the mark.

Franklin pushed back on the “suppression poll” assertion and said that, if anything, those polls may have motivated Trump’s base to turn out, as evidenced by the much-tighter-than-expected margin.

Already alarms are sounding about whether the polling industry is in its death throes after two straight presidential misses. Franklin said that while pollsters’ reputation has taken a collective hit, public polling remains valuable so people can know what their fellow Americans are thinking.

“If we didn’t have public polling, whether done by universities or done by media organizations or other independent-but-public polling, then the only people that had it would be the parties and the campaigns and the interest groups,” Franklin said. “And those organizations are intrinsically self-interested, with no reason to release information that’s not favorable to their interests.”