Introducing Our 2024 Poll Tracker

What do the polls say about the race for the White House?

With less than five months to go, new poll averages from The New York Times, which debut today, show a very close race nationally and in the critical battleground states.

On average, the recent national polls show President Biden and Donald J. Trump each winning about 46 percent of the vote in a head-to-head race. They are essentially tied as they approach their first debate Thursday night.

A close race is a slight improvement for Mr. Biden. He’s trailed in other polling averages — even if only narrowly — on almost every single day of the last nine months. But he narrowed his deficit in the wake of the State of the Union and with the clinching of the nominations in March. He’s closed the gap entirely since Mr. Trump’s felony conviction in New York.

The race is also close in key battleground states, but here Mr. Trump is slightly ahead in the most recent polls. While he often leads by only a point or two, he does nonetheless hold the edge in Pennsylvania, Wisconsin, Michigan, Nevada, Arizona and Georgia — states that would be enough for Mr. Trump to win the Electoral College and therefore the presidency.

Of course, the election will not be held today and the polls will not be exactly right. Polls are inexact, fuzzy measurements subject to many sources of error. Mr. Trump’s lead in the battleground states is more tenuous than most candidates’ polling advantages in past elections. In fact, today’s polling average shows a closer race than the final November polls did in any election in the last 20 years.

Mr. Trump leads by no more than two points in Pennsylvania, Michigan and Wisconsin — three states that would be sufficient for Mr. Biden to win the presidency, provided he held the states and districts where he currently leads. In a typical election, the polls err one way or another by about two points. And this cycle, there are many additional sources of uncertainty: the independent candidate Robert F. Kennedy Jr.; a large number of “double-hater” voters who dislike both major candidates; and Mr. Trump’s dependence on support from traditionally Democratic low-turnout voters. If the election were held today, it would be no surprise if Mr. Biden prevailed.

It’s important to emphasize that a polling average is not a forecast. It’s just our best measure of the state of the race today. It doesn’t account for, say, the partisanship or demographics of a state. It doesn’t allocate undecided voters. It can’t simulate what will happen in the debates and at the conventions.

Why look at a polling average? Individual polls are small, noisy and subject to a sizable margin of error. In theory, the average of polls benefits from the combined sample size of many polls. It also benefits from the collective wisdom of the pollsters, who make necessary but debatable decisions about how to conduct their polls. Over the last two decades, poll averages have outperformed individual high-quality polls.

It’s worth cautioning that the case for polling averages has worsened in recent years. There are fewer high-quality polls than there used to be, as rigorous survey research has become significantly more expensive. At the same time, it’s never been easier to field a cheap online poll. Even high school students have fielded national and state polls in recent cycles. And while the last decade hasn’t been kind to the traditional pollsters, the cheap online and partisan polls still systematically underperform, creating serious risks to any poll average.

The 2022 midterm election was arguably the first cycle where it was unequivocally better to look at a single high-quality nonpartisan poll, like the Times/Siena College poll, than to look at an average. There just weren’t enough high-quality polls to include in an average, and the lower-quality polls were off target.

Of course, the lesson of the last election is not necessarily a lesson for the next. A hypothetical poll average that would have excelled in 2022 might have fared terribly in 2020, and vice versa. Our average has been optimized on elections going back to 1980, and it attempts to account for the differences between different kinds of polls and pollsters, including giving more weight to the highest-quality polls. But in any given election, any given poll average might ultimately fare better or worse, regardless of how it would have performed historically.

Still, the Times polling average, which will be updated every day through Election Day, is our best effort to help you understand what the polls are saying about the race. And right now, the average shows a very close race.

Want to learn more about how our poll average works? You can read our methodology here.