Support independent journalism that matters — become a Vox Member today.
The 2024 presidential race is the closest on record. Donald Trump and Kamala Harris are separated by 1 percentage point or less both nationally and in the decisive battleground states. In the Electoral College, the two candidates are essentially tied, with the election turning on a deadlocked Pennsylvania, according to the New York Times’s polling average.
If one looks beyond the polling data, however, it is not difficult to tell a story about why each candidate is actually the favorite. Here are four reasons you should expect Trump to win the presidency Tuesday, and five to expect a Harris victory instead.
In 2016 and 2020, polls badly underestimated Trump’s support, particularly in battleground states. There are reasons to fear that pollsters will miss in the same direction yet again.
One leading theory of pollsters’ recent woes goes like this: Americans who participate in surveys are systematically different — and more Democratic — than those who do not.
If you are highly politically engaged, very trustful of strangers, or both, then you’re going to be more inclined to respond to a phone call or text message from a pollster. After all, people who are eager to discuss politics are more liable to share their voting intentions when prompted. And those who aren’t wary of other people are more likely to engage with an unsolicited request for conversation.
This has always been true to an extent. But in the past, Republican voters weren’t dramatically less likely than Democratic ones to be highly engaged or trustful. Trump’s conquest of the GOP changed this. As Republicans unified behind a populist, anti-system conspiracy theorist, they made inroads with less politically active and trusting voters, while Democrats racked up large gains with college-educated voters, who have relatively high turnout rates and levels of social trust.
Throughout the past year, there have been signs that this realignment has continued apace, with Trump polling much better among low-propensity voters than reliable ones. And some pollsters are finding it much easier to reach Democrats than Republicans as a result. Nate Cohn, chief pollster for the New York Times, writes that white Democrats were 16 percent more likely than white Republicans to participate in his final surveys.
There are many ways for pollsters to counteract this bias, but the fundamental challenge it poses may prove insurmountable. You can try to make sure that low-propensity voters (i.e., those who sat out recent elections) are well-represented in your survey. But as Cohn notes, in his polls of previous elections, low-propensity voters that agreed to participate ultimately voted at much higher rates than such voters in general. Put differently: The politically disengaged voters who respond to polls are not representative of such voters, as they are more politically active (and thus, theoretically, more Democratic). Which isn’t entirely surprising since the very act of taking a poll is a form of political engagement.
Pollsters don’t want to underestimate Trump again. But the inescapable limitations of their methods may lead them to make the same mistake thrice.
Since the onset of post-pandemic inflation in 2022, incumbents the world over have struggled to retain power. Ruling parties have either lost seats or control of government altogether in Japan, Austria, Britain, Italy, and Germany, among other nations. And if polls hold steady, Justin Trudeau’s governing Liberal Party is poised to suffer a landslide defeat in next year’s Canadian election.
Americans appear to share this transcontinental desire for change. Both NBC News and YouGov find roughly 65 percent of Americans saying America is on “the wrong track,” while 26 percent say they’re “satisfied with the way things are going in the United States” in Gallup’s polling.
Harris has tried to sell herself as a change candidate, centering her campaign on the slogan, “We’re not going back.” But she is still the sitting vice president, and Trump’s team has worked relentlessly to link her to the exceptionally unpopular commander-in-chief.
Joe Biden’s approval rating is currently 18 points underwater. And as CNN’s Harry Enten notes, since World War II, whenever a retiring president has had a net-negative approval rating, their desired successor has lost: Adlai Stevenson failed to overcome Harry Truman’s unpopularity in 1952, Hubert Humphrey couldn’t withstand Lyndon Johnson’s in 1968, and John McCain was undone by George W. Bush’s in 2008.
Thus, if undecided voters — like their peers abroad — are in a mood for change on Election Day, one might expect them to break for Trump.
For virtually all of the past 30 years, more voters have identified as Democratic than Republican in Gallup’s surveys. Now, the GOP boasts a 3-point lead on national party identification.
There has been a strong correlation historically between this figure and election outcomes: In 2004 and 2016, when Democrats’ lead on partisan identity was slim, Republicans won the presidency. By contrast, when Democrats’ advantage on this measure was larger than normal — as in 1992, 1996, 2008, 2012, and 2020 — they won the White House.
Gallup isn’t alone in detecting a historically unusual advantage for Republicans in partisan self-identification. NBC News and Pew Research found the same result. And this shift toward the GOP has also surfaced in some states’ voter registration data: In Pennsylvania, there were 630,000 more registered Democrats than registered Republicans in March 2021. By October 2024, that advantage had fallen by more than half to around 300,000, according to NBC News.
Polls have routinely found voters rating the economy and immigration as two of their top issues, and favoring the GOP on both. Further, when Gallup recently asked voters which party is better able to handle the issue that matters to them most — whatever that issue may be — they trusted the Republicans over the Democrats by a 46 percent to 41 percent margin.
In the past 75 years, no political party has won the presidency while trailing on this question in Gallup’s poll.
Harris is significantly more popular than Trump: Voters disapprove of the Democratic nominee by roughly 2 points while they disapprove of the Republican one by roughly 9 points, according to FiveThirtyEight’s polling average.
As CNN’s Enten notes, since 1956, the more popular candidate has won the White House 16 out of 17 times (the one exception was Trump in 2016).
As noted above, it’s possible that pollsters are doomed to perpetually underestimate Trump’s support due to his disproportionate strength with the politically disengaged.
But it’s also quite possible that pollsters are overcorrecting for that problem. Certainly, every polling firm has a strong incentive to avoid understating Trump’s support for a third election in a row. The reputational damage of missing in the other direction — by overestimating him — is liable to be less severe; it’s more understandable to make new mistakes than to repeat old ones.
There are many ways that pollsters could yield to this incentive. Survey firms don’t publish their raw results. Rather, they each apply their own idiosyncratic series of weights to those results in an effort to make them more representative of the expected electorate. If your sample is more educated than American voters as a whole, for example, then you will increase the weight of responses from non-college-educated voters in your final tally.
This is undoubtedly necessary for generating accurate results. But it also gives pollsters a lot of freedom to bend their findings in a safer direction. If your initial result shows Harris winning Wisconsin by 7 points, you can adjust your weights until you have a finding that’s more in keeping with polling averages (and less likely to get your shop laughed at post-election). As Vanderbilt University political scientist Josh Clinton recently demonstrated, a pollster’s decisions about how to model the 2024 electorate can shift the results of a given survey by 8 points.
And there are signs that pollsters might be abusing their discretion. As NBC News reported last week, recent polls of the top seven swing states have shown an improbably close race: 124 of 321 surveys showed margins of 1 point or less.
The probability of so many polls showing such a small gap is extremely low, absent pollsters putting their thumbs on the scale. Even if Harris and Trump really are virtually tied in all major swing states, the inherent randomness of polling should produce a more varied series of results than we’re seeing. That suggests many pollsters are playing it safe and engineering results that conform to the consensus. If so, they could be erring on the side of underestimating Harris since that is reputationally safer than underestimating Trump.
Polls have long shown that voters trust Republicans over Democrats on inflation and immigration. And the surge in prices and border crossings under Biden coincided with the collapse of his approval rating.
Over the course of 2024, however, these issues have become less pressing. Inflation has fallen sharply over the past two years, and last week, the Federal Reserve’s favorite inflation indicator — the Personal Consumption Expenditures index — increased by just 2.1 percent, roughly in line with the central bank’s target rate.
At the same time, economic growth continues humming along at a roughly 3 percent rate, stocks have been hovering near record highs, and unemployment is low. And there are some signs that this is beginning to make an impression on voters: Some recent polls have shown Harris eroding Trump’s advantage on the economy.
Meanwhile, migrant encounters at the US-Mexico border dropped by 77 percent between December 2023 and August 2024, according to a recent analysis from the Pew Research Center. There’s no guarantee that improving objective conditions will change the electorate’s subjective impressions. But there is some evidence that immigration has receded a bit from the public’s consciousness in recent months: Between April and October, the share of voters naming immigration as America’s “most important problem” fell from 27 percent to 21 percent.
As noted above, Democrats now do disproportionately well with America’s most politically engaged voters. That might lead polls to overrepresent Democratic voters. But all else equal, every political party would prefer to be popular with Americans who reliably vote in elections than with those who do not.
Democrats’ strength with the civic-minded has helped them dominate off-year special elections and hold their own in the 2022 midterms, despite public discontent with inflation and Biden.
There’s reason to think that Harris’s coalition might be even more reliable than Biden’s was in 2020, as some polls have shown her gaining ground with women and college-educated voters, both of whom turn out at elevated rates.
Even if polls are perfectly accurate, they are likely to underestimate whichever candidate does best with voters who make their final decision on Election Day or just before it. And in the New York Times’s final battleground state polls, Harris led Trump by 9 points among voters who’d only recently picked their candidate.
This finding isn’t entirely surprising. The headlines in the campaign’s final weeks haven’t been ideal for Trump, with his former chief of staff calling him a fascist and his rally in Madison Square Garden convincing many Hispanic voters that he is a racist.
If late deciders are indeed breaking toward Harris, then she is likely to win the presidency.
