Member Newsletter
I wanted to take a look at the polling news from this weekend and try to help you make some sense of where the race is. Obviously I can’t tell you what’s going to happen in November or necessarily which polls to believe. But I think I can provide some overview of and context for why different polls might seem to show different things, and how to think about that difference. Yesterday, NBC News released a poll showing Harris beating Trump by 5 points nationwide and 6 points if third-party candidates were added. Another national poll from CBS showed Harris 4 points up over Trump nationwide. But it was the NBC poll which got the most attention because poll watchers still give some extra credit to the big, largely phone-based polls from the major national media organizations.
Obviously, no single poll should bulk too large in anyone’s thinking. But what gave the NBC News poll a lot of attention wasn’t so much the result, which was obviously good for Harris, as the fact that it tended to match and confirm and perhaps amplify the trends we’ve seen from a lot of other polls since the debate. Those polls show Harris solidifying a small national lead, consolidating small leads in the Blue Wall states while running about even through the southern tier swing states. There’s been a large volume of polls showing that. But people wanted to see one of those big, high-priced, phone-based polls say the same thing. In part, that was because you have the Times-Siena poll, which as I’ve explained in the past is very respected but also has a totally disproportionate impact on the media narrative about the race, saying something different. That poll has continued to show a much closer race than the great majority of other polls. A nationwide poll from last week from Times-Siena showed a tied race at 47 percent after one a few weeks earlier that showed Trump ahead by 1 percentage point. So that NBC poll wasn’t just another solid poll for Harris. It made it seem a bit more like Times-Siena is an outlier. Not wrong necessarily but an outlier from the majority of campaign.
There are some additional data points out of the NBC News poll that are notable. It shows Harris winning on the issues of abortion, temperament and change. Trump remains ahead on inflation, the economy and the border. The NBC write-up also puts a lot of emphasis on the steep increase in Harris’ favorability numbers — a point, I’m happy to say, we’ve been discussing with you for weeks. This poll shows her at net +3 favorability. And they note that this is the biggest popularity bump of any presidential candidate in the 35 years they’ve been doing this poll. (The only bigger one is the jump George W. Bush got after 9/11, which is obviously not really comparable.) I’m increasingly thinking Harris’ net favorability numbers and this “change” marker are the signal in the noise of the last two months, the underlying drivers of the horse race.
So far so good.
Then this morning Times-Siena released a new poll of the southern tier swing states of Arizona, Georgia, and North Carolina and they’re some of best numbers Trump has seen in weeks. Trump’s up in that poll 5 points in Arizona, 4 points in Georgia and 2 points in North Carolina.
Needless to say, this new round of polls spawned talk of a Trump resurgence, tears for anxious Democrats and new consternation about what the polls are telling us. So how do we make sense of this? The simplest and best answer is that Times-Siena is simply looking at a different electorate. You can see this really across the board in all its polls this cycle. It’s consistent from Biden through Harris. They show Harris doing much better than Biden, like everyone else. But they still tend to show a slightly Trumpier electorate than most other pollsters, even across the change from Biden to Harris.
Now what does that mean when we say that they’re looking at a different electorate, since obviously all pollsters are in fact polling the same electorate, either registered or likely voters in the United States in the fall of 2024? Put simply, each pollster has to make a series of educated guesses about the make-up of the electorate. How many of different kinds of people are there? What are their different propensities to vote? Calling these “guesses” probably gives too little credit. It might be better to call them theories of the electorate. From a good pollster, those should include an analysis of recent elections, comparisons of earlier polls with actual election results, other demographic information. In other words, from a good pollster these aren’t based on vibes. They’re based on trying to come up with an accurate understanding of the different groupings which make up the electorate and their characteristics.
I think every serious polling observer would agree that Times-Siena is simply modeling an electorate that is, not dramatically but significantly, different from that which most other pollsters are using. Once that’s clear, it’s not a huge surprise that you have a result like this and the NBC poll coming out at about the same time. Given what we’ve seen this cycle from Times-Siena, it would be surprising if they were more similar. What we can’t really know until the election is which of these models is “right.” All we can say at present is that Times-Siena seems to be an outlier. If they were some random pollster, they’d be easier to dismiss. But they’re not. They’re well respected. But they’re largely on their own with this theory or model of the electorate. To the extent there’s a problem it’s that this one poll has a wildly disproportionate impact on media narratives. You can’t just dismiss it as some crazy, junk poll. Because it’s not. But the attention to this one poll would suggest that it’s been uniquely predictive in the past or has some dramatically different level of quality than other polls. Neither is the case. It’s one out of one or two hand fulls of high quality public pollsters in the U.S.
One additional point about these Times-Siena sunbelt polls. They show Trump with substantial leads in those states when the averages show them tied or close to tied. But Trump is definitely doing better than he is in the Blue Wall states. So they’re not wildly off what other pollsters are showing in those states. Times Siena paired up with the Philadelphia Inquirer on a Pennsylvania poll last week and it showed Harris up by 4. That’s a touch higher than the current consensus of PA polls. How one reconciles all these things together I have no idea. It’s always good to remember that there’s a decent amount of variability in any single poll result.
This is a case where averages are our best guide. The consensus of polls is usually close to right. And I suspect it is in this case too. Polling errors happen with relative frequency. But not always in the same direction. My own personal hunch is that we’re in a period in which there’s some mix of a Dobbs effect and pollsters having slightly overcorrected in response to the 2020 polling error. But everything in this final paragraph is my hunch. For a reliable and I think fairly uncontroversial way to look at current polling in general, look at everything prior to this paragraph.