Both Tom Tiffany and David Crowley have something to cheer about in Wednesday’s Marquette University Law School Poll. Tiffany has pulled dead even with Crowley among registered voters, while Democrat enthusiasm has Crowley up five percentage points among likely voters.
Already, though, the prevailing narrative in the race is that Crowley is pulling ahead and his lead will only grow from here as voters get to know him. History suggests that isn’t likely. In fact, this may well be the best Crowley polls.
Since its inception, Marquette’s August poll has not just been wildly unsuccessful at predicting the ultimate result of a November race, it has also overestimate Democrats’ support by an average of 3.5%. Even when taking out the anomalous 2016 presidential election (when every poll right up until Election Day) was enormously wrong, the average overestimation of Democrat support by 2.1%.
Just once did it get a race exactly right in August—its 2024 presidential election poll showing President Trump beating Kamala Harris by a point—and in five of the nine August polls of presidential and gubernatorial races it has conducted over the past decade, it has been off by six or more points. In every single one of those five massive misses, it dramatically overestimated the support of the Democrat.
Just once, in that August 2024 presidential poll, did it ever show the Republican candidate leading even though Republican candidates went on to win their respective races four times in nine elections.
Just three times in nine polls did the August Marquette poll overestimate Republican support: In 2018, when it showed incumbent governor Scott Walker and challenger Tony Evers tied (Evers went on to win the race by a point) and incumbent Senator Tammy Baldwin up three on challenger Leah Vukmir (Baldwin won by 11), and in 2022, when it had Evers leading Republican Tim Michels by two points (Evers won by three).
Nearly all the biggest misses showed Democrats with seemingly insurmountable leads in August. Hillary Clinton led Trump by 10 points in the August 2016 Marquette poll while Democrat Senator led Republican challenger Ron Johnson by six. Less than three months later, Trump won Wisconsin by about a point and Johnson beat Feingold by 3.3.
In 2020, Joe Biden led Trump by six points in August but won by 0.6%. Two years later, Marquette’s August poll had Mandela Barnes crushing Johnson by seven points. He lost by a point that November. In 2024, Baldwin led Republican Eric Hovde by six points in August but won the race by just a point.
Five polls showing big leads for Democrats in August, five polls that were proven to be wildly incorrect come November.
This is not to say that the Marquette University Law School poll is inherently biased. Quite the opposite; it’s been the gold standard in polling Wisconsin for years. But it is notoriously difficult to poll Wisconsin, especially in August.
Just look at the most recent Marquette poll, which showed socialist Francesca Hong leading Crowley by 31 points less than two weeks before this month’s Democratic primary. The day after that poll was released, though, Mandela Barnes dropped out of the race—essentially rendering the poll meaningless.
That’s the point, though: So much can happen in a race, especially between August and November, that early polls generally don’t carry a whole lot of predictive value. With so much happening in the most recent race, even a poll conducted while early votes were being cast couldn’t even get it close.
Early polls showing Democrats with huge leads do serve an important purpose, however: They shape the narrative of the race and create a “bandwagon effect” for the Democrat. Truly undecided voters tend to gravitate toward the candidate they believe to be winning, so any poll at any point showing a Democrat with a big lead is a huge edge.
Why do Democrats consistently get this advantage? It’s not that most pollsters are in the tank for them (although some most assuredly are); it’s that Democrat voters are far more likely to take part in polls. They are easier to reach on the phone, and they aren’t as likely to immediately hang up when a pollster calls. This is especially true in August, when so many people are on vacation and generally checked out of politics.
Polling is indeed an important part of politics, but in August, it’s just not worth assigning to much importance to polls—especially ones that show Democrats with big leads.
