For more than a decade, Wisconsin has been at the white-hot center of American politics. After a long stretch of Republican dominance there — capped off by a shock Trump win in 2016 — Democrats have mounted a considerable comeback in recent years. They ousted Scott Walker in 2018, tilted the state to Joe Biden — barely — in 2020, and gained a majority on the state supreme court, breaking Republicans’ suffocating statewide gerrymander. Polls show Kamala Harris with a small but persistent edge over Donald Trump, but for two straight presidential cycles, polls have dramatically overestimated Democratic strength in Wisconsin.
Nobody knows more about Wisconsin voting trends — or is more aware of those polling issues — than Charles Franklin. A professor of law and public policy at Marquette University, he has run the widely respected Marquette Law School survey since 2012. (FiveThirtyEight rates it as the third best in the country.) I spoke with him about where he thinks Harris vs. Trump stands in the Badger State, how political demographics in Wisconsin have flipped over the years, and why he isn’t sleeping that well right now.
The New York Times recently ran a piece with the headline, ‘Harris has a polling edge in Wisconsin, but Democrats don’t trust it.’ And you were quoted in there saying, “My numbers are my numbers, but I think in terms of putting it into context, four points would be a surprisingly strong finish for Harris.” That was referring to Marquette’s most recent poll, which had Harris up 52 and 48 over Trump. Do you have a different sense of the race than the numbers that you actually put out?
It’s the paradox of polling that the polling represents the best data that I’m able to produce. We do the methodology as best we can, but there’s always the subjective side about whether this is a four-point race at this stage. A reason for reticence is that four of the last six elections were decided here by less than one percentage-point margins, so a four point win for either side would be unusually large for recent years. Governor Evers was reelected in 2022 with a 3.4 percent margin, and he’s the most popular politician in the state.
The other reason, of course, is polling errors. We were off in 2016 by over six points, and we were off by about four points in 2020. At least in 2020 we got the winner right, which was an improvement. So those are just reasonable cautions to anyone reading polls — to look at the history of races in the state and look at previous polling errors and come to your own judgment.
People just always want to know who’s going to win. And there’s no good answer for that.
Yeah, and my standard answer this year is “Give me a quarter and I’ll flip it for you,” which is not entirely tongue-in-cheek. I think we are so close in so many battleground states that forecasts from now to November are perilous.
In late 2020, as you said, you were off by four points near the end of the Biden-Trump race. That was still quite a bit better than some — ABC News and the Washington Post memorably had Biden up 17 in late October.
Yes. That gets repeated every time, with good reason.
I think I was skeptical of that one at the time.
Well, Gary Langer, who conducted it, was skeptical.
Good for him.
Seriously, this isn’t really your focus, but within the polling community, it is a mark of some integrity that you publish your results as they stand and not hide the ones you don’t believe in. So in that sense, kudos to Gary and to ABC for putting that out. But we saw a range of polling in 2020 up in the nine, 11, 17 range.
Yeah, the Times had him up 11, so they were far from alone.
One of the paradoxes for us is our long-term average error across all of the races we’ve done is 2.2 percentage points. So we were off by four in 2020, which was double our average error, but we still look pretty good compared to all the other pollsters. It’s sort of cold comfort to be the least inaccurate poll, or one of the least inaccurate, but it does show that a lot of pollsters have come through Wisconsin and left with much bigger overestimates of the vote than we have.
The common story told about 2016 is that Trump was reaching all these white working-class people who may have been infrequent voters, and who were undersampled in polls. But do you have as good a sense of what went amiss in 2020 as you do in 2016? And how have you tried to repair whatever problems you’ve found?
My view of ‘16 and ‘20 is more or less the way you stated it, that Trump has done a remarkable job of activating people who hadn’t participated before, who are disgruntled with the political system. And the profile of those folks are exactly the profile of folks that don’t want to do polls. They distrust politics, they distrust the media, and they distrust pollsters. The net effect is not huge, but it’s three or four or five percent that were understating Trump’s support in ‘16 and in ‘20. And the challenge for pollsters is: Is there any way to reach those folks who are supporting Trump and probably will go to vote this time when they are so skeptical of mainstream media and pollsters and so on?
So we’re doing what we can, and we’ll know in November whether we’ve succeeded in that or not. What we’ve changed is that in the past, we stratified the state into five regions: Milwaukee, the rest of the........