Whatever else one can say about presidential elections involving Donald Trump, they certainly haven’t lacked drama. In 2016, Trump pulled off an upset win over Hillary Clinton, clearing the supposedly unscalable “Blue Wall.” In 2020, as an even bigger underdog, he narrowly lost to Joe Biden, as the wall held—and as Republican-leaning Georgia and Arizona fell. In each case, fewer than 105,000 well-placed votes—a smaller group than have packed a college football stadium—would have swung the election’s outcome. That’s fewer than one in 1,600 of all votes cast, or less than 0.07 percent. It’s a testament to the genius of our system that the Electoral College produced margins of over 70 electoral votes each time, while also avoiding the grisly specter of a nationwide recount. Still, both races went down to the wire.
Heading into the third consecutive presidential election with Trump as the Republican nominee, he’s neither the clear underdog nor the clear favorite, as the betting markets are split between the two candidates. One puzzling thing about the 2024 contest—an issues-versus-intangibles election—is that national polling isn’t lining up with statewide polling, at least not based on the two prior Trump elections’ results. Whether winning in 2016 or losing in 2020, Trump fared 3.7 (2016) and 3.6 (2020) percentage points better in the following seven swing states than he did nationally: Pennsylvania, Georgia, North Carolina, Michigan, Arizona, Wisconsin, and Nevada. Yet in the RealClearPolitics (RCP) average of recent polling, as of 11:00 EST on Monday, November 4, Trump is now faring just 0.7 points better in those states, on average, than he is nationally.
This means that one of three things must be true: national polling is underestimating Kamala Harris’s support; state polling is underestimating Trump’s support; or the gap between the national margin and the average margin in the seven key swing states will be notably smaller this time around. The race could hinge on which of these three things—or some blend of them—is true.
It’s hard to say with any certainty which of these possibilities is most realistic. Let’s start with national polling, which has underestimated both Republican and Democratic candidates in the past. President Barack Obama outperformed national popular vote projections by 3.2 points in 2012 (winning by 3.9 points versus a predicted 0.7 points in the RCP average), while Trump did the same in 2016 and 2020 races (besting the RCP average by 1.1 and 2.7 points, respectively). It is unlikely that national polls will get things exactly right this time around, either, and the direction of their error remains to be seen.
We’ve seen the second possibility—that the polls are underestimating Trump’s support in the seven key swing states—materialize in both 2016 and 2020. But important caveats are in order. In 2016, the polls’ underselling of Trump’s swing-state support was essentially limited to Michigan, Wisconsin, Pennsylvania, and North Carolina. In those states, Trump outperformed RCP and Nate Silver’s estimates by about 4 or 5 points, on average. In the other three swing states (Georgia, Arizona, and Nevada), both RCP and Silver missed by less than 1.5 points, on average, and were collectively as apt to overstate Trump’s support in these three states as to understate it.
In 2020, Trump again outperformed Silver’s estimates in the three Great Lakes states and North Carolina, and also in Nevada. RCP, however, was within 1.4 points of Trump’s performance in six of the seven states (all but Wisconsin), and essentially nailed his performance in Pennsylvania and Nevada. In other words, anyone who expects Trump to outperform RCP’s current projections in the non-Wisconsin swing states to any noteworthy degree is banking on the aggregator being less accurate this time around.
Let’s turn to the third possibility: that the gap between the polling nationally and in the seven swing states will be smaller than it was in 2016 and 2020. To assess that prospect, and to get a state-by-state sense of the election, we’ll examine the race in each of those states, assessing historical results and polling. As RCP’s Sean Trende has noted, nuggets of information gleaned from early voting are generally fool’s gold, so this discussion will steer clear of those.
Nevada: anybody’s guess, likely matters only if Harris upsets Trump in the South. The smallest of the seven key swing states, Nevada is also perhaps the toughest to gauge. The state, which has six electoral votes, hasn’t voted for a Republican presidential nominee in 20 years. Relative to the national average, however, the Silver State has shifted about 2.5 points toward the GOP in each of the past three presidential elections (moving from 5.2 points to the left of the nation to 2.1 points to the right of the nation in the process). If that trend continues, Trump will almost certainly win here.
In 2020, RCP nailed this state exactly, projecting a 2.4-point Biden win in a race he won by 2.4 points, while Nate Silver underestimated Trump’s performance by 3.7 points. Today, RCP has Trump up by 1 point, and Silver has him ahead by 0.4 points. Of the two polls that most accurately pegged this state in the two prior Trump elections, Gravis and Emerson, the former hasn’t polled the state this time around, while the latter has the race tied. Nevada’s outcome is most likely to affect the overall result if Harris upsets Trump in Georgia or North Carolina—in which case, winning Nevada would help open an alternate path to victory for the former president, one relying more on the Great Lakes states and the West.
Georgia: a close battle in a key state for Trump. Before Joe Biden won Georgia by 0.2 points in 2020, Republican nominees had emerged victorious in the Peach State six straight times, with their five most recent wins all being by at least 5 points. Polling here has generally been quite accurate: in both 2016 and 2020, RCP and Silver were each less than 1.5 points off the actual margin. RCP now has Trump up by 1.9 points, and Silver has him up by 1.3. On the other hand, the New York Times/Siena poll, which was extremely accurate in Georgia in 2020 (predicting a tie), has Harris up by 1 point. Another accurate poll in 2020, Emerson (which overestimated Trump by 1.2 points in that race), now has the former president up 1 point. A Harris win here would be a major blow to Trump’s chances, forcing the former president to win two out of three from among Pennsylvania, Wisconsin, and Michigan, or else win both Pennsylvania and Nevada.
Michigan: Harris attempts to hold on in a near-must-win state. Of the key swing states, Biden in 2020 enjoyed his largest margin of victory in Michigan, which has voted for the Democratic nominee in all but one of the presidential elections across the past three decades—when it flipped for Trump by 0.2 points in 2016. In other words, this should be the easiest of these seven states for Harris to win. But the polling doesn’t show a comfortable advantage for the vice president, who may be suffering the political consequences of the Biden administration’s electric-vehicle mandates.
Both RCP and Silver missed Michigan badly in 2016, as each suggested a relatively easy Hillary Clinton win; Silver was even further off in 2020, forecasting an eight-point Biden win (he won by 2.8). RCP, however, was closer in 2020, being within 1.4 points of the actual result. In all four instances (twice for RCP, twice for Silver), Trump has outperformed expectations. Right now, both show Harris winning (RCP by 1.2 points, Silver by 1.1), and the fact that the two polling aggregators are so close, and that RCP did well in Michigan last time, suggests that the polls on which Silver most heavily relies have probably—finally—figured out how to capture voter behavior in this state reasonably well. Insider Advantage, the most accurate Michigan poll in 2020 (showing Biden up 2 points when he won by 2.8), now has the race tied.
Harris is probably the slight favorite here, but the Wolverine State is hardly the gimme she’d like it to be.
Arizona: advantage Trump. RCP and Silver currently have Trump ahead by 2.5 and 2.6 points, respectively, here. This border state, which has been particularly affected by the immigration crisis, was 4.2 points to the nation’s right in 2020 (Biden won by just 0.3 points here versus 4.5 points nationally). While that was down from 5.6 points to the right of the nation in 2016, and more than 10 points in each of the two Obama elections, this still appears to be one of the more right-leaning of the seven swing states. In terms of individual polls, both Marist’s and CBS/YouGov’s polling have fared well in Arizona during the past two presidential elections; Marist now has Trump up by just 1 point, while CBS/YouGov last polled the state in mid-October and had Trump up by 3. This appears to be the swing state in which one candidate has the clearest advantage, albeit in a still-competitive race.
North Carolina: a close race, but polls might be underestimating Trump. In the past two Trump elections, the Tar Heel State was more than 5 points to the right of the nation (in terms of margin of victory) each time, and the last time that it wasn’t more than 5 points to the right of the nation was in 1980 (a landslide year for Ronald Reagan). What’s more, in the Trump era, polling has proved problematic in this state: RCP and Silver both underestimated Trump each time, by an average of 2 points in RCP’s case and 3.7 points in Silver’s case. Pollster Patrick Ruffini writes that North Carolina, like Wisconsin, has a lot of “rural working class” voters who haven’t “been politically polarized until recently.” These are voters who not long ago went for Obama and now support Trump, and they seem to be inordinately hard to poll.
Trump is currently ahead here by 1.5 points per RCP and 1.1 points per Silver—only about 1.5 points to the right of the national polling in each case. Given both the state’s voting history and the pollsters’ track record in evaluating it, there’s good reason to believe that these margins may be a bit low. The individual polls that have most accurately gauged North Carolina in the prior Trump elections currently have him up by an average of between 1.5 and 2 points. This is probably the second-most promising of the seven states for Trump, and given that he won it in 2020 (while losing the other six states), a Harris upset here would bode quite well for her chances of victory.
Wisconsin: a near toss-up that could swing the election. If Trump holds on to win in Georgia, North Carolina, and Arizona, Wisconsin could put him over the top (giving him 272 electoral votes). This swing state has utterly confounded pollsters in the past two Trump elections. In 2020, the RCP average was off by 6 points in Wisconsin; Silver’s projection was even worse, being off by 7.6 points. This time around, polling looks a lot more realistic here, but the best guess is that the polls are still somewhat underestimating Trump’s support. The polls in the RCP average that were closest to pegging this state in either prior Trump election were Trafalgar and Susquehanna in 2020, when Trafalgar suggested Biden would win by 1 point and Susquehanna projected him winning by 3 points; he actually won by 0.7. Trafalgar now has Harris up 1 point, while Susquehanna has yet to release a Wisconsin poll this time around. Overall, RCP has Harris up by 0.4 points, while Silver has her up by 0.8 points. Putting all this together, the Badger State is a near-toss-up—and, aside from Pennsylvania, it may be the state most likely to decide the race.
Pennsylvania: the biggest prize, with no clear favorite. The largest of the seven swing states is the most important up-for-grabs prize on the board. The state is extremely competitive, with RCP showing Trump up 0.3 points and Silver showing him up 0.4 points. It’s noteworthy that the two polling aggregators have converged, after RCP was 3.5 points more accurate than Silver in 2020 (RCP projected Biden’s 1.2-point win exactly; Silver estimated he’d win by 4.7 points). It suggests that the polls that Silver prefers, like the New York Times/Siena poll, may finally more or less have figured this state out. (That poll now shows the state tied.) Indeed, there is less polling variation in Pennsylvania than in any of the other six states, with every poll in the RCP average showing a margin of 2 points or less, either way (versus 6 points or less, either way, in Nevada). Susquehanna (which overestimated Trump by 2.2 points in 2020) and Rasmussen (which underestimated him by 1.8 points), were the most accurate of the RCP-utilized polls in 2020; Susquehanna now has the race tied, and Rasmussen has Trump up 2. Trafalgar, which essentially nailed the race in 2016 (within 0.3 points) but then overestimated Trump’s support by 3.2 points in 2020, now has Trump up 1. Pennsylvania was about 3 points to the right of the nation in each of the past two Trump elections.
The winner of the aptly named Keystone State will certainly have the inside lane to the White House. If Trump wins Pennsylvania, Harris will have to win either Georgia or North Carolina. If Trump wins Pennsylvania along with North Carolina and Arizona, Harris would have to sweep the other four states. On the flip side, if Harris wins Pennsylvania, Trump would have to win Wisconsin or Michigan along with a majority of the remaining key swing states.
In all, don’t be surprised if the polls are at least slightly underestimating Trump’s support in Wisconsin and North Carolina—and assume that for Harris to win, she’ll likely have to win the national popular vote by a greater margin than the narrow one she is currently projected to hold. With that said, it’s still anybody’s race. We’ll find out the result on Tuesday night—or perhaps in a few months.
Photos: KAMIL KRZACZYNSKI/AFP via Getty Images (left) / Brandon Bell/Getty Images (right)