That New York Times Poll Is Such A ... New York Times Poll!
Keep your eye on 'likely' voters, folks.
Those dagnabbed data dinguses at the New York Times are at it again, with another 2024 election poll that appears to be Terrible News for Joe Biden: “Trump Leads in 5 Key States, as Young and Nonwhite Voters Express Discontent With Biden.” Oh no! The Times-Siena Poll shows Trump leading in five of six swing states, is it too late to replace Biden with a friendly labradoodle instead? Looks like Joe is DOOOMED!
Yes, again. And also again.
But the scary stuff is far less scary once you step away from the breathless headline and start looking at the actual poll results, to say nothing of potential issues with the poll’s methodology.
So yeah, let’s do that, shall we?
For starters, as Simon Rosenberg points out again, the poll results are quite different when you compare the responses of “registered voters” — the topline results used in the Times story — with the results for “likely voters,” which the Times dutifully includes near the top of the story and in one chart, but otherwise doesn’t say much about.
Here’s how the Times framed that, even including what might be mistaken for a very useful reminder if it didn’t appear in a story so full of negative conclusions based on the worser set of data for Biden:
The race was closer among likely voters. Mr. Trump led in five states as well, but Mr. Biden edged ahead in Michigan while trailing only narrowly in Wisconsin and Pennsylvania. While Mr. Biden won all six of those states in 2020, victories in Pennsylvania, Michigan and Wisconsin would be enough for him to win re-election, provided he won everywhere else he did four years ago.
But then, a mere two paragraphs later, the Times story proclaims that despite some good news for Joe Biden — big gains in the stock market, the start of Trump’s election interference trial, and huge ad buys by Biden’s campaign — the poll results (for registered voters) “offer little indication that any of these developments have helped Mr. Biden.”
Rosenberg notes further that in all three of those states, the likely-voter difference is well within the margin of error, too:
Let’s look at MI/PA/WI in the NYT poll (Biden-Trump, Likely Voters):
• MI 47-46
• PA 45-48
• WI 46-47 (Biden leads in WI among registered voters)
These finds are similar to the recent large sample CBS News polls in these three states which also found all three within margin of error/up for grabs
That CBS poll had Biden up by two points in Michigan (51-49), and down by one in Pennsylvania and Wisconsin (49-50 in both). All very much within the margin of error.
So if you consider the margin of error and the likely voters, Trump isn’t so much winning in five of six states as he is winning in three (Arizona, Georgia, and Nevada) and pretty much tied in the other three.
As Jay Kuo points out, the Times headline might just as well have been “Race Essentially Tied in the Electoral College,” which is almost exactly the gist of that one paragraph in the Times story about the likely-voter data. But how many top story slots would that get you on cable news?
When you dig into the poll’s methodology, Kuo notes, you find additional problems, like the fact that the base of respondents was actually more heavily Republican (49 percent) than Democratic (45 percent), an imbalance compared to the real world that pollsters had to try to correct for by “adding all kinds of mathematical ‘weights’ to try and account for the fact that there are just so many more Republicans in their data.”
The sample set also appears to be “heavily biased toward conservatives,” another factor that oughta be smoothed out by weighting of the results, but which might be better address by starting with a less biased poll population.
For that matter, Kuo says, even the Times’s “likely voter” criteria seem a bit loose, considering that, as the New Republic’s Greg Sargent confirmed with Times pollster Nate Cohn on Twitter, 20 percent of the sample who counted as “likely” voters
have either not voted in the last two midterms, have not voted in general or last two midterms, or are newly registered/never voted
Erm. It’s a slice of a slice of the poll respondents, but I’m not sure how you get “likely” from that.
The race is entirely too close, but we are going to once again yell at the Times that no, Trump is not literally winning young people, so get the entire fuck out.
Mr. Trump and Mr. Biden are essentially tied among 18-to-29-year-olds and Hispanic voters, even though each group gave Mr. Biden more than 60 percent of their vote in 2020. Mr. Trump also wins more than 20 percent of Black voters — a tally that would be the highest level of Black support for any Republican presidential candidate since the enactment of the Civil Rights Act of 1964.
Simply … no.
And what’s this, we guess the New York Times released another poll and according to Simon Rosenberg, it’s not as stupid.
There are other possible problems with the Times poll that I could probably nerd out about, but suddenly I am run over by an average truck, even though I was safely standing in the median. The driver was very mean.
[NYT / Hopium Chronicles / The Status Kuo]
Yr Wonkette is funded entirely by reader donations. If you can, please become a paid subscriber, or if a one-time donation is more up your alley, this button’s for you, cat.
Maybe they should make a category, "Voters likely to be pissed off that Roe was overturned and now they have to drag their asses to the polls lest these absolute nut jobs on the right wing do something similarly horrible in the next 2 years."
That might have some improved predictive power compared to the models they've been using.
Heck, I didn't even mention that whatever the poll numbers on the candidates themselves, looking at the candidates in isolation fails to account for other factors, like how mad lots of Arizona voters are going to be at the legislature and Supreme Court over that Civil War abortion law, and how the state's initiative to add abortion rights to the state constitution may drive turnout.