Like the New York Giants trying to fix their offensive line issues, among other things, Democratic pollsters have been working for years trying to make sure they can nail their projections. It hasn’t been successful. Donald J. Trump wrecked the ship in more ways than one, with two presidential elections where these firms had underestimated the former president’s support. Joe Biden barely won the 2020 race, now an open secret since Democratic donors and party bosses rolled that old man off a cliff in July. At this point in the race four years ago, Biden was ahead of Trump by seven points. Kamala Harris, who was coronated last week after the Biden coup, is trailing Trump by the same amount.
Harris’ takeover has injected more enthusiasm into the Democratic Party base, and the cash flows back into the war chests. Still, Democratic pollsters are offering a fair warning that Harris’ position, while better than Biden’s, isn’t good. In their models, she’s tied with Trump in the swing states, though methods used to gauge voter preferences show that it’s still Trump’s election to lose. Yet, some of the reasons for their panicking are somewhat entertaining. In short, Democrats still struggle with the low-propensity voters, the folks they view as idiots, coming out of the woodwork during presidential cycles (via Politico):
So now that Kamala Harris has caught Trump in the polls in her first month as a candidate, it’s left Democrats wondering: How real is her surge?
Here at the Democratic convention this week, some in the party’s professional class are trying to tamp down the exuberance. Officials with the top pro-Harris super PAC said their polling “is much less rosy” than public surveys. Other Democratic pollsters noted that — even if their polling is right — Trump still maintains a lot of advantages.
“It’s still a very tough race, and that feels consistent with everything we know,” said Margie Omero, a partner at the Democratic polling firm GBAO Strategies.
There are plenty of warning signs hidden in the data: A poll commissioned by the Democratic messaging firm Navigator Research and unveiled during the convention showed Harris and Trump essentially tied across the swing-state map. And the candidate characteristics that are best correlated with voters’ preferences — whether a candidate is up to the job, has the right vision and is a strong leader — generally favored Trump in the survey.
[…]
One major part of the effort was a lengthy experiment in the swing state of Wisconsin. The goal wasn’t to predict the result of an election; it was to see which voters could be captured by a monthslong survey using multiple ways of finding people, including a door-to-door component — and how that group differs from the voters reached in the typical phone or web surveys conducted over the course of a few days.
The main finding: Standard polls capture voters who are more engaged with politics and consider it more important to their identity. That kind of bias wouldn’t necessarily cause problems in a low-turnout election, like an off year or midterm, because those are exactly the kind of voters who show up.
But in a presidential race, when lower-propensity voters also turn out, that could be an issue. And the kind of time, effort and expense that went into getting those voters to participate isn’t scalable in a fast-changing election.
And these pollsters conceded that the “Wisconsin project” cannot be replicated on the timeline we have right now in the 2024 cycle. Liberals know a tight race breaks for the GOP because this is all about the Electoral College. Second, let’s look at some of these surveys now that RFK Jr. has quit the 2024 race. Mr. Kennedy dropped out last Friday and endorsed Trump; his supporters are likelier to support the former president. We shall see, but the polling issue circles back to something that the political class fails to accept: the Trump coalition is diverse, it's dispersed efficiently geographically, and our system favors candidates who can string together coalitions of this nature.
Join the conversation as a VIP Member