Has there been a shift toward Trump in the final days of the race? Here’s what the polls show.

Uncredited/Associated Press

If you’re feeling confused by political polling this year, you’re not alone. This year’s presidential race between Vice President Kamala Harris and former president Donald Trump is turning out to be among the closest in recent memory.

Another reason for uncertainty? There are more polls than ever. The number of active public pollsters has more than doubled since 2000, according to Pew Research Center. Pair that with the emergence of rival polling averages, and it can be hard to even know what the polls are saying.

Lastly, “pollsters offer a reminder there’s only so much one can do to predict the future,” Globe D.C. reporter Tal Kopan wrote earlier this month. “Ultimately, some voters will make up their minds at the last minute.”

Here’s what we know about the state of the presidential race.

Has there really been a late shift toward Trump in the polls?

National and battleground state polls alike are consistently showing an incredibly close race and are not designed to drill down to the level of precision required to detect swings of a point or two, experts say. Keep in mind it is common to see a margin of error of several percentage points, so when a poll shows a one-point lead for Harris, as a recent Suffolk University/USA Today poll did, a 3 percentage-point margin of error means Harris could either be in the lead or trailing.

Polling averages have thus emerged in an attempt to detect trends over time. Averages are typically made up of a selection of high-quality publicly available polls and take into account the sample size, recency, and other factors. If we look at the national averages, the trend in recent weeks shows a slight erosion of Harris’s very narrow advantage.

Averages calculated by FiveThirtyEight, The New York Times, and the Cook Political Report each show that Harris’s national polling advantage has shrunk by a point or two since mid-September.

Looking at battleground state polling averages created by FiveThirtyEight and the Times, the widest lead among the seven battleground states is just two percentage points, leaving the electoral college a true tossup. FiveThirtyEight shows Trump with a two point lead in Arizona and Georgia. All other states show a lead of one point or less between Trump and Harris.

How important are polls at this stage in the election? How much should I pay attention to them?

Courtney Kennedy, vice president of methods and innovation at Pew Research Center, urges those anxious about the outcome to be aware of what polls can and cannot do.

“I get asked a lot, ‘Can I trust the polls?’ And I push back and I say, ‘Trust the polls to do what?’ If you’re asking me, ‘Can I trust the polls to reliably tell me who’s going to win a close race?’ The answer is no,” she told the Globe Friday. “They are not designed to be able to get it right if the race is within a few percentage points. If the question is, ‘Can I trust polls to tell me how Americans responded to the Dobbs decision, or how do Americans feel about the economy?’ Polling is absolutely up to that task.”

Polling averages allow us to look beyond some of the individual error that might appear in a given poll to see the bigger picture, Kennedy said. But they, too, have their limits.

“If the polls are off this year, chances are they’re going to be off in a similar direction, at least that’s what we saw in 2016 and 2020. Polls really had the same weakness. So we shouldn’t have a false hope that averaging is going to miraculously fix it,” she said.

There is also evidence that the polls are less accurate when Trump is on the ballot.

“They have absolutely struggled to fully measure his base of support in years when when he’s running,” Kennedy said, adding that a Pew analysis found the majority of pollsters have changed their methodology since 2016 to address the polling problems of previous cycles.

“But it’s not the case that like, ‘Oh, we’ve figured out the silver bullet. Everybody’s fixed it, and we’re good to go.’ Instead, you have pollsters trying an array of different fixes, and we’re hopeful that in the aggregate, that will result in greater accuracy.”

Finally, we’re about to get a lot less information about the state of the race. Major polling operations historically stop releasing new polling data in the days immediately before the election.