Dem 51
image description
GOP 49
image description

Who's Gonna Win This Thing?, Part I: The Siena Poll

We had a few e-mails this week proposing that we are in denial about the possibility that Donald Trump could win the 2024 presidential election, and that we are doing a disservice to readers by encouraging them to bury their heads in the sand.

We feel we have been clear about this, having written numerous times in the past month or two that our guess is that the election is a coin flip at this particular moment in time. That means that Trump most certainly could win—indeed, that he is as likely to win as Joe Biden is, as far as we can tell. Here are some arguments for why Trump might win:

Again, Trump could definitely win. Full stop.

That said, there are reasons we think it's roughly a coin flip, and that you have to take the polling with a few grains of salt. And those reasons go beyond "We just can't believe a meanie like Trump isn't losing by 30 points." To try to explain, let's take a quick look at the numbers from the recent Siena poll (which are pretty much in line with all Siena polls this cycle), as compared to the actual numbers from Trump's other two elections:

2024, Siena
2020, Actual
2016, Actual

The spread is the gap between the most extreme result in the Democrats' direction and the most extreme result in the Republicans' direction. In 2020, for example, the Democrats did best in Michigan (+2.8%) and worst in Georgia (+0.2%), resulting in a spread of 2.6%.

To believe that Siena has the right of it, you have to believe that the three "Southern route" states have gone from being competitive to being laughers for Trump, even as the "Northern route" states remain close. Similarly, the reason these six are the "swing" states is that they usually correlate with each other in terms of being close. But now, the spread between them is nearly double what it was in 2016 and nearly quintuple what it was in 2020. It is certainly possible there has been this much movement, but it would be very unusual. And, zooming in a bit more, it is particularly hard to accept that Nevada—which is similar to Arizona demographically but more Democratic—has suddenly become seven points more Trumpy than the Grand Canyon State.

There's another way to examine this, and that's by looking at the crosstabs, which are just full of things that stretch credulity to the breaking point. According to Siena:

We did not cherry-pick the most outlandish results; in choosing which two numbers to look at in each swing state, we actually were trying to give a broad representation of the various demographic categories. And note that not all the wackiness works in Biden's direction; Siena appears to be overstating his support with older voters in Michigan, for example. That said, the wackiness skews against Biden far more often than not, and the most extreme and head-scratching numbers almost always involve (apparently) understating Biden's support.

Let's put it this way. Steve Deace is a right-wing talker who does a show for Glenn Beck's The Blaze network. And as reader J.R. in Philadelphia, PA, brought to our attention, Deace looked over Siena's cross-tab numbers, just like we did above, and concluded the poll is very possibly... a Democratic trick. Seriously. He writes:

Trump's strong performance in the top-line of this poll is contrary to the cross-tabs that show him often under-performing where he was in 2020, and almost exclusively results from Biden's under-performance with core constituencies of the Democrat Party—blacks, Hispanics, and some demos of women. In other words, for this poll to be fulfilled 25 weeks from now would require Biden to perform worse with women than any Democrat since 1984, worse with Blacks than any Democrat since 1956, and worse with Hispanics than any Democrat ever.

This poll isn't a realistic path to victory. We all know Democrats will "fortify" their standing with these demos come Election Day/Week/Month/Festival.

Trump could win, but it won't look like this, because that would mean we're in an environment where Democrats would also be looking at a 40-seat loss in the House as well and nobody sees that.

At best this poll is problematic. At worse it's a psyop to convince you to stay drunk on numbing over-confidence and watching Fox News shilling all day, rather than the hard work of matching the Democrats' "ballot harvesting scheme."

We're not embracing the kooky elements here, but it's pretty good evidence that the numbers really don't add up when you look closely.

Exactly why they don't add up is a matter of conjecture. One possibility is pollster error, probably in the form of an inaccurate model of the electorate. Another possibility, and this is the drum we keep beating, is that a lot of voters are undecided, and considerably more probable Democrats are in that group than probable Republicans.

Now, if you want someone who will say these numbers are absolutely on target, and very meaningful, there are people out there happy to do it. That's what Nate Cohn is saying, although he kinda has to, because he runs the Times' polling operation. The other Nate (Silver) too, though we're not going to link to him because his stuff is paywalled. CNN's Harry Enten is another; he rarely sees a poll that he does not take at face value, it would seem.

As for us, we will continue to look at polls with a wary eye, probably until September or so. As we have noted, the conventions will be over by then, low-information voters will be paying (some amount of) attention, votes will start to become very real (as opposed to making a statement to pollsters), and that's when things start to settle in.

We'll have a "Part II" and a "Part III" tomorrow. We WERE going to have them today, but we're trying to keep the postings from getting too long, or from going live too late. So, we'll have to hold them. (Z)

This item appeared on Read it Monday through Friday for political and election news, Saturday for answers to reader's questions, and Sunday for letters from readers.                     State polls                     All Senate candidates