Harris and Trump are tied in the polls with 2 weeks to go. Here's how to make sense of the numbers.

Photo illustration: Yahoo News; photos: Jacquelyn Martin/AP, Evan Vucci/AP
Photo illustration: Yahoo News; photos: Jacquelyn Martin/AP, Evan Vucci/AP

With only two weeks to go until Election Day, Vice President Kamala Harris and former President Donald Trump are busy making their final pitches to voters — and trying to pull ahead in the polls. But surveys have shown a stubborn tie, both nationally and in the key battleground states. Here’s a quick Yahoo News guide to what the numbers can — and can’t — tell you about who will win the 2024 presidential election.

The polls are saying this could be the closest presidential election in modern U.S. history.

Sure, you can find national surveys that show Harris ahead by 5 percentage points; you can also string together individual swing state polls to give Trump a potentially decisive lead.

But the most prudent way to partake of polling in the final days of a big election is not by cherry-picking results that suit your political preferences. It’s by absorbing all the numbers in aggregate.

There are a few transparent, trustworthy sites that average everything together — both the national surveys and their swing-state counterparts — while also statistically accounting for each pollster’s track record of accuracy and any partisan affiliations they might have.

Here’s where those sites put the race right now:

Silver Bulletin: Nationally, Harris (48.8%) is leading Trump (47.2%) by 1.6 percentage points. In Pennsylvania — which is poised to be the pivotal swing state — Harris and Trump are tied at 48%.

FiveThirtyEight: Nationally, Harris (48.2%) is leading Trump (46.3%) by 1.9 percentage points. In Pennsylvania, Trump (47.8%) is leading Harris (47.6%) by 0.2 percentage points.

New York Times: Nationally, Harris (49%) is leading Trump (48%) by 1 percentage point. In Pennsylvania, Harris and Trump are tied at 48%.

In other words: it’s a dead heat. Even the two tightest presidential contests of the last 65 years — John F. Kennedy vs. Richard Nixon in 1960 and George W. Bush vs. Al Gore in 2000 — weren’t this tight with two weeks to go.

Yes. To win a U.S. presidential election, you don’t necessarily need to win the “popular vote” (i.e., the combined tally across all 50 states). You need to stockpile “electoral votes” by finishing first in individual states.

With minor exceptions, the candidate who gets the most votes in a particular state on Election Day is awarded all of that state’s electoral votes, which are proportional to the state's population. Whoever finishes first in enough states to secure at least 270 electoral votes — a majority — wins the Oval Office.

National polls — the ones showing Harris ahead by 1 or 2 points — reflect only the national popular vote. To get a sense of what might actually happen in the Electoral College, swing state polls are more relevant.

The problem is that they’re also harder to conduct. Small sample sizes, big margins of error, questionable (often partisan) pollsters. So you have to take swing-state surveys with a grain of salt and (again) look at them in the aggregate.

And what do they reveal if you do? That the race is … still a tie. According to the current Silver Bulletin averages, Harris is narrowly ahead in Wisconsin (by 0.6 points), Michigan (by 0.6 points) and Nevada (by 0.5 points); Trump is ahead by slightly more in North Carolina (1 point), Georgia (1.3 points) and Arizona (2 points).

A tie in Pennsylvania — which is what Silver Bulletin shows — means that if the election were held today, and if today’s averages were exactly right, Harris would get 257 electoral votes and Trump would get 262. Pennsylvania’s 19 electoral votes would be the tiebreaker.

“Big” might be an overstatement — but if anyone has been making gains in recent days, it’s Trump.

In fact, all of the Silver Bulletin’s key swing-state averages show margins that have moved in Trump’s direction over the last week (by anywhere from 0.2 points to 0.6 points) and over the last month (anywhere from 0.4 points to 1.9 points).

Nationally, the same site says Trump has closed the gap with Harris by about one point over the same period.

To be sure, these are modest improvements. Yet the fact they’ve consistently surfaced across various polls and places suggests a certain narrative: that whatever slight bounce Harris enjoyed after dramatically replacing President Biden atop the Democratic ticket and “winning” her one and only debate with Trump has faded, and now the contest is reverting back to the nail-biter it was always destined to be.

Sort of. Numerous forces and factors are in play right now — the state of the economy, the unusual semi-incumbency of both candidates, etc. But the one that’s best captured in the polls is probably demographics.

A caveat before proceeding. The more you slice and dice a survey into separate groups — by age, race, education and so on — the smaller (and less representative) your sample becomes. Your margin of error grows accordingly.

So it’s hard to quantify the precise gap between Harris and Trump among, say, Latino voters. One defining trend has emerged, however. During the Trump era, Democrats have performed better and better among college graduates, while Republicans have performed better and better among voters without a college degree. Education is now the major polarization point in U.S. politics — even across racial lines.

This has resulted in some significant movement among key demographics. Black voters, for instance, still overwhelmingly support Democrats. But Trump’s support grew from an estimated 7% in 2016 to an estimated 9% in 2020 — and could go as high as 15% in 2024, if the latest polls are to be believed. Black men in particular seem to be responsible for this shift.

Latinos are a similar story: Harris is doing better with them than Biden was before he dropped out, but not quite as well as Biden did in 2020.

On the flip side, Harris is benefiting from increased support among white, college-educated voters. In 2016, Clinton won this group — who previously leaned Republican — by 5 points. In 2020, Biden won them by 9 points. Now, a recent CNN poll shows Harris leading among white college graduates by 18.

So why is the 2024 race so close? Because surveys suggest that these changes could cancel each other out — depending, of course, on who actually shows up to vote.

Polls aren’t elections. No pollster has the time or resources to survey all 160 million Americans who are likely to vote this year.

But here’s the good news: it’s not necessary to poll every voter to get a representative sample of them. In fact, that’s the whole idea behind polling. Instead, all you need is a group of about 1,500 to 2,000 U.S. adults who match the overall population in terms of age, gender, race, education and voting behavior. Get that and you can get a pretty accurate idea of public opinion — within a margin of error of 2.5% to 3%.

Want proof that polling isn’t wildly inaccurate? Since 1972, the final national polling average in a presidential race has differed from the actual national vote margin by just 2.5 points, on average. In 2012, the polls were “off” by 3 points. In 2016, they were off by 2 points. In 2020, they were off by 4 points. And in the 2022 midterms — the most recent major election cycle — they were “more accurate … than in any cycle since at least 1998.”

In other words, polls rarely predict the exact election results; there’s almost always some degree of “error.” But it tends to be smallish.

Now, a smallish error can still seem big. In 2016, for instance, the final national polling average showed Hillary Clinton ahead of Trump by about 4 points; she eventually won the popular vote by 2 points. That’s actually a smaller miss than usual. But because Clinton narrowly lost the Electoral College — and because polls in the decisive states consistently overestimated her support — a lot of observers were shocked by the result.

Conversely, a biggish error can seem small. In 2020, FiveThirtyEight’s final Wisconsin polling average put Biden ahead of Trump by more than 8 percentage points. Biden wound up winning the state by less than 1 point — but because he won the Electoral College with votes to spare, fewer people were upset by the miss.

There will likely be another round of polling errors in 2024. The polls could systematically underestimate Trump again, like they did in 2016 and 2020. Or they could systematically underestimate the Democratic nominee again, like they did in 2012 and 2008. There’s no way to tell in advance. And when the polls are this close, the direction of even a typical error — 2.5 points for Trump, 2.5 points for Harris — could make all the difference.

Or the polls could be spot on.

Pollsters do everything in their power to adjust for the problems of previous cycles. In 2020, something called nonresponse bias — essentially, Biden supporters seemed more eager or available to respond than Trump supporters — might have made Biden’s lead look bigger than it was. In 2016, undersampling voters without college degrees likely had a similar effect on Clinton’s numbers, especially in key swing states.

To account for these issues, pollsters have tweaked the statistical methods they use to further dial in the demographic balance of respondents — a process called “weighting.” But no one will know until after the election whether they got it right.