On Friday morning, a politico texted me to ask what I thought about The New York Times' chief pollster, Nate Cohn, tweet series about their polling in Pennsylvania.
It’s obviously in Nate’s best interest to defend his polling, but he acknowledges that it’s leaving many people scratching their heads. How is Trump doing so well nationally, tied in the latest New York Times/Siena poll and losing in Pennsylvania, which votes to the right of the country?
Cohn’s margins with whites without a college degree (Trump +27) are slightly lower than those shown in exit polls in 2020 and 2016 (Trump +32 both years, according to exit polls). Even if Trump is doing somewhat better among non-college-educated whites (as The New York Times/Siena’s national polls suggest), it’s well within the poll’s margin of error.
The issue isn’t necessarily with the non-college-educated white vote; it’s with seniors.
Now, I know that dedicated readers have heard me scream about this repeatedly, but it’s worth saying again.
Trump is tied in the national polling but losing PA because of the disparity among seniors between the two sample sizes.
Let me explain why I think they’re wrong and it matters.
Past Senior Performance
It should come as no surprise that seniors have been one of the strongest voting demographics for Republicans. This has been true for decades. It’s also true that when you cross-examine seniors’ voting habits in Pennsylvania versus the nation as a whole, seniors in PA vote more Republican than the nation at large.
Here’s a comparison using state exit polling matched with national exit polls and Pew Research’s autopsy of the 2016 and 2020 elections.
Romney did exceptionally well with seniors nationally, higher than George W. Bush or John McCain, and still had higher results in Pennsylvania.
So, has Kamala Harris, the progressive candidate who’s the vice president to an unpopular incumbent, managed to break the code and win support from this group? It’s doubtful. Seniors are a relatively fixed group; their demographics don’t change radically because of mass immigration. It certainly wouldn’t be the case that seniors in Pennsylvania have moved left, while nationally, they’re leaning slightly more fitting.
I trust The New York Times’ crosstabs showing strong support from seniors because they match Pew Research’s poll, which had the largest sample size in the country. Pew surveyed 10,000 people twice and found that seniors supported Trump over Harris by six points in August and five points in September.
This means Trump is doing slightly better than he did in 2020 but not as well as he did in 2016. This would make sense given that COVID-19 significantly influenced seniors’ voting habits, and Joe Biden was a historically old candidate who represented their generation. Let’s not forget that he credited his upbringing in Pennsylvania for making him a working-class guy.
Harris has none of those things working in her favor, and like Clinton, the idea of a female president receives the most significant opposition from seniors.
So, I'm extremely skeptical that she’s broken a two-decade-old habit of voters in Pennsylvania by making them vote to the left of seniors nationally.
Polling Response Bias
There’s a genuine phenomenon that I’ve written about in great detail: polling response bias. Older liberals answer the phones more frequently than seniors, creating the illusion that seniors are supporting Democrats against Trump at a greater rate than they are.
I wrote about this in great detail for The American Conservative Magazine, but here’s a small part:
Quinnipiac’s last poll before Biden dropped out had him leading Trump among senior citizens by eight points; the New York Times/Siena and NPR/Marist polls gave Biden a four-point lead with Baby Boomers and the Silent Generation, Fox News had the 46th president tied with his predecessor. In contrast, the Washington Post/ABC poll gave Trump just a point lead among this crucial demographic.Â
According to Pew Research’s analysis of the last two presidential elections, Trump defeated Clinton by nine and Biden by four percent among voters over 65. Still, in the run-up to the election, pollsters predicted a blue wave led by an army of gray-haired voters.
In the last few weeks before the election, Biden was crushing Trump among seniors in most national polls: Quinnipiac said Biden would win voters 65 and up by 15 points, Emerson by 12, CNN by 11, The New York Times/Siena and Fox News by 10, NPR/Marist by nine, YouGov/Economist by four points USA Today by two, ABC/Washington Post had him up a point…
This wasn’t the first time this happened. In 2016, key swing states had Clinton outperforming Trump among senior citizens.Â
According to exit polls, Trump beat Clinton by 23 points among senior citizens in North Carolina, but polls leading up to the election had his victory far narrower. Monmouth and the New York Times had his lead at 11 points, CNN had him up five, and Elon University tied them.
Likewise, in Pennsylvania, Trump beat Clinton by 10 points among seniors in exit polls. Still, serious pollsters undervalued his support among this key demographic group in the last few weeks of the election. CNN had Clinton leading Trump by four, as did Monmouth.
The Pennsylvania Conundrum
Assuming that these national polls with large voter sample sizes are accurate and Trump is performing slightly better with seniors than in 2020, with the number being R+6 instead of R+4, it would explain the entire error in Cohn’s Pennsylvania polling.
Many detractors insist you shouldn’t get lost in the crosstabs, which I agree with for the most part. However, seniors are a massive voting bloc, and the numbers they’re reporting are far outside the margin of error.
Seniors’ overwhelming support for Harris also explains why Trump’s numbers among white voters are down significantly. He won white voters in Pennsylvania by 15 points in 2020 and by 20 points in 2016, and Romney also won white voters by 15 points in 2012.
Most of these polls have Trump doing as well among white voters as John McCain in 2008 when he lost the state by more than 10 points.
Conclusion:
Cohn is looking for the error in the wrong sample size; it’s not that his numbers with whites without a degree are off, and it’s throwing the results; it’s that seniors are off. The response bias among old white liberals in the Rust Belt through otherwise good polling firms numbers in 2016 and 2020, and it’s happening again.
That’s why your numbers nationally, which fit into the numbers provided by Pew, don’t match those in Pennsylvania.
This is a +2 R sample with +4 for Trump among seniors and yet is +4 Kamala in total https://x.com/thepoliticalhq/status/1837231517635948816?s=46&t=yoYoPgDGga-oDhZhAMG3WQ There was an A+ poll from yesterday that was +4 R sample yet +5 Kamala. I’d be concerned