Tuesday, November 7, 2023

BTRTN: Why Biden’s 2024 Prospects Are Not Quite as Bad as the NY Times/Siena Poll Would Indicate

Tom questions the headlines of the recent poll shocker. 

Many friends and readers have been freaking out about the NY Times/Siena poll released earlier this week.  It paints a very grim picture of Biden’s reelection prospects based on polling in six swing states.  Their attention-grabbing conclusion is as follows:  If the results in the poll were the same next November, Mr. Trump would be poised to win more than 300 Electoral College votes, far above the 270 needed to take the White House.  They based that on their data that showed Trump ahead of Biden by 4 points or more, among registered voters, in five of those six swing states (Arizona, Georgia, Michigan, Nevada and Pennsylvania;  Biden was slightly ahead in Wisconsin) .  They are correct in their math:  if those results did not change by Election Day, 2024, then Trump would emerge the winner by a 302-236 electoral vote margin. 

But I don’t agree with that conclusion, based on all the recent polls, and here’s why, state by state: 

·        Michigan.  By the NYT/Siena poll’s own data, Trump may be ahead by +5 in MIchigan among “registered voters,” but the race is a dead heat among “likely voters.”  While it can be a little squishy on how a pollster defines a “likely voter,” there is little doubt that, in any form, it is a better measure than “registered voters,” many of whom do not vote, even in presidential years.  The “dead heat” result in Michigan is also consistent with that of other recent polling (by Morning Consult).   So, Michigan is really a toss-up as of now, not a Trump state. 

·        Pennsylvania.  The NYT/Siena poll also had Trump up in the Keystone State by +5 points.  But there are two other recent polls in Pennsylvania, by Morning Consult, which had Trump ahead by only +1, and another poll by Franklin & Marshall, which had Biden ahead by +2.  If you “aggregate” the three polls, you have a dead heat (Trump +1, within the margin of error).  So, Pennsylvania is also really a toss-up as of now. 

·        Nevada.  Similar story.  Morning Consult has Biden +3, and a CNN poll has Biden +1.  NYT/Siena had Trump ahead by a whopping +11.  That result is so dramatic that it looks like an outlier to me.  This is a state Biden won in 2020.  Before reporting such a number, or even using it in an aggregation, I would wait for a confirming poll.  But even if I use it in an aggregation with the other two polls, that takes Trump’s “lead” down from +11 to +3.  Given the aggregated number’s reliance on the NYT/Siena results, I would call Nevada a toss-up at this stage as well. 

So, summarizing the true electoral picture based on all recent polling: 

·        Trump is ahead in Arizona and Georgia (where other polls confirm the NYT/Siena outcome), and also North Carolina and Florida, which all the other polls would indicate (though NYT/Siena did not include them in their survey), as well as all the other red states he won in 2020.  Trump, of course, won North Carolina and Florida in 2020, so there is no net pickup for him in those states by virtue of being ahead. 

·        Biden is not meaningfully ahead at this point in any swing states, so he gets only the blue states he won in 2020. 

·        Michigan, Pennsylvania, Wisconsin and Nevada are toss-ups (the NYT/Siena poll had Biden +2 in Wisconsin, as did Morning Consult). 

When you do that math, it leads to the following electoral outlook based on recent polling: 

This is not good news for Biden, there is no mistaking that.  But my main conclusion, based on all of the recent polling is:  it is going to be another close election.  Which is not exactly a breakthrough insight.   To conclude, as the Times does, that Biden would lose if the polls don't change is simply not an accurate read of today’s polling environment.  The swing states could very well stay with Biden; he won them all in 2020.  And in that scenario he would win again in 2024, 276-262.

The NY Times is committing the error of relying on a single poll.  They are doing this because they have invested in the NYT/Siena poll and want to highlight their own results.  It is a way to make breaking news on which you cannot be scooped by other media competitors.   But in doing so, they are ignoring the lesson that Nate Silver (and BTRTN) taught everyone over a decade ago:  you get a more accurate picture when you aggregate polls. 

Just for summary purposes, the chart below captures all the various data in recent polling:

You can see the sea of red in the first data column, the NYT/Siena "registered voter" results, moving to a much purpler set of outcomes in the far right column if you average all the recent polling. 

Plus, in my opinion, to talk about any polling result today “holding until Election Day, 2024” is absurd.  The major variables in this (and any) election will proceed in unknown trajectories that could help or hurt the candidates.  No one can guess the direction, status or impact of the various Trump trials, the Israeli-Hamas war, the economy and Biden’s (or Trump’s) health, and that’s just among (as Donald Rumsfeld would say) the “known unknowns.”  

The Times knows this and they should have focused less on the numbers, and more on the underlying reasons for Biden's weakness.  The rest of their data and analysis is quite good, including key findings (confirmed elsewhere) that Trump is gaining strength among young, Black and Latino voters, and that Trump is viewed as better than Biden on the economy.  Biden would do well to heed those conclusions, and doubtless his team is agonizing over them, as they should. 

But Democrats, again, this is going to be a close race and could very well be determined by turnout and thus by get out the vote efforts.  Don’t be discouraged – be angry.  Work as hard in 2024 as you have in all recent elections, if not harder.  This one is the most meaningful of them all.





  1. This is overly convoluted….people under 45 do not have land lines and don’t answer strange phone calls. The end. It’s not generation Z, it still Z and X. So good luck ever trying to get a real sample distribution and I’m not sure you can conclude anything from a poll anymore….maybe older people vote more conservative and younger people don’t participate in polls?

    1. My friend, thanks for this comment, but it is wrong. Pollsters conquered the cell phone issue long ago. That's not a problem. The polls are generally very good: BTRTN accurately predicted the outcome of 34 out of 35 Senate races in 2022, and 35 and out 36 governor races. We could not do that without quite accurate polling on the part of the pollsters. The only recent issue with polling recently has been undersampling Trump supporters, which was an issue in 2016 and 2020, but less so in 2022.

  2. Worth thinking back 4 years, when the polls were -- to put it bluntly -- varied.

    The New Yorker, in November 2019, said
    Yesterday the whole hep political world revolved around discussions of some new data from the New York Times/Siena College polling combine, with analysis from the Upshot’s Nate Cohn. To make a long story short, the data and analysis (based on polling of key 2016–2020 battleground states) suggested that (a) Donald Trump is in much better shape for reelection that is commonly assumed, and (b) that the fast-ascending Democratic candidacy of Elizabeth Warren is displaying some potential general-election weaknesses.

    Today, as though intended to provide balm to freaked-out Democrats, the ABC/Washington Post polling combine released a new national poll showing Elizabeth Warren, Joe Biden, and Bernie Sanders all trouncing Trump in a general-election trial heat by double-digit margins (Biden up by 17, Warren by 15, and Sanders by 14). And also today, Nate Silver introduced and released the latest edition of FiveThirtyEight’s unique pollster ratings. And guess what? Both NYT/Siena and ABC/WaPo got the coveted A+ rating (five out of 532 polling outfits got that gold-standard rating).

    Most people aren't thinking about the "real" choice that will be on the ballot in 2024. And they certainly are not thinking about whatever current events at most salient in the 3 months before the election (and how they relate to prior events).

  3. Polling is certainly one "snapshot" of the status of the race. This far out, results don't have a great record for predictive power. If I remember correctly, early November 2019 was when polling saw a surge of expectation that Biden was fading, and Warren and Harris were rising, and that all of them would romp over Trump (or were within MoE).

    In the world of actual voting, those tightly tied to abortion restrictions underperformed in 2022. Those unwilling to say Biden won in 2020 underperformed in 2022 and 2023.

    Here's hoping there will be discussions of campaign finance results as another sort of snapshot.

  4. I think you’ll find Ezra Klein’s recent podcast with Michael Podhorzer about these polls really interesting. https://www.nytimes.com/2023/11/14/opinion/ezra-klein-podcast-michael-podhorzer.html?smid=nytcore-ios-share&referringSource=articleShare


Leave a comment