There have been very few post-debate polls in any of the seven states. In fact, many of the changes in our state averages are attributable to polls fielded before the debate, but released after.

What changed: pre-debate

It’s worth divvying up what’s changed into two groups: the polls fielded before the debate, and the polls since.

First, the pre-debate polls released this past week: They might sound like old news, but it’s not so simple. In many states, there haven’t been many recent surveys, as the Labor Day holiday in the US put a pause on many polls.

A string of high-quality state polls found Harris faring well, at least in relative terms, in Wisconsin, Iowa, Virginia, North Carolina and even Alaska. Trump, on the other hand, earned solid results in Georgia and Michigan.

These results didn’t shift our state polling averages too much, but they were still good news for Harris — and arguably a bit of a surprise. After all, the polls we’d seen before the debate showed a close and tightening race, with several high-quality outlets — Times/Siena, YouGov/Economist, Marist, Pew Research and KFF — finding the race within 1 point or so. A close race nationwide might have been expected to yield a clear Trump lead in the battleground states.

But that’s not what happened. Instead, Harris inched ahead in our average in North Carolina, after Quinnipiac and SurveyUSA polls found her up 3 points, while Trump now has the edge in Georgia. And Harris firmed up her modest lead in Wisconsin, where the venerable Marquette Law School poll found her up 4 points.

What changed: post-debate

The relatively few post-debate polls show signs of a Harris bounce.

According to our average, she has already gained about 1 point nationwide, rising from a 1.7-point lead Wednesday morning to a 2.7-point lead as of Monday morning. That 1-point shift has been reasonably consistent across the six national polls that took surveys before and after the debate.

In the days ahead, Harris may gain even more ground. For one, most of the post-debate polls have come from online panels. They tend to shift less than other polls, as they’re often composed of highly engaged voters and weighted more heavily than the typically more volatile phone surveys, which will probably arrive this week.

There’s another reason: Many people don’t watch debates, but they do hear the post-debate coverage — like the continuing discussion of Trump’s false claim that Haitian immigrants in Springfield, Ohio, have been eating pets. Extended coverage of a debate can help the perceived winner just as much as the debate itself, and yield additional polling gains in the days or even weeks thereafter.

Will a polling bump last?

The time after a debate is fraught for pollsters.

Loading

On the one hand, debates can change the race in a lasting way — just ask President Joe Biden. Even setting aside that recent example, the polls routinely shift after the first debate without ever reverting to their previous levels. That’s what happened in 2000, 2004, 2008 and 2012.

On the other hand, debates can create shifts in the political environment that simply don’t last. Just ask Hillary Clinton, who built a big lead after the debates in 2016, only to see it fade in the final week. Something similar happened to Biden in 2020.

These post-debate swings were fleeting, but they could have reflected real changes in stated preference — that is, driven by people who would have given different answers to a pollster before the debate. You can imagine, for instance, some Nikki Haley voters who were coming around to Trump, but who might now say they’re undecided after watching the debate. Even if these voters eventually come around to Trump in the end, we’d see him slip in post-debate polls.

It’s also possible that these post-debate swings were partly a mirage: They might simply be attributable to changes in who responds to surveys, not any changes in the preferences of voters.

Either way, it’s an excellent reminder of the limitations of polling. If the election were held on Tuesday, would Harris’ bounce really materialise? Would those undecided Haley voters ultimately vote for Trump, stay home or vote for Harris? Did Harris secure new support from voters who hadn’t heard enough about her, but saw in the debate someone capable of handling the presidency? We just can’t say.

All of these same questions will exist on the morning of the election. We won’t know if undecided voters will break in one direction, if the polls have been skewed by varying response rates or if, maybe, the polls are dead-on.

What we do know is that the polls are so close that even an ordinary polling error could give either side a decisive victory.

If the polls erred as they did in 2020, Trump would sweep the battlegrounds. Conversely, Harris would win big if it turned out the polls were wrong as they were in 2022.

It’s easy to imagine either scenario. It’s hard not to see a result like “Harris +4 in Wisconsin” without feeling a sense of foreboding, as Wisconsin was ground zero for polling error the last two presidential elections. On the other hand, there has been a deluge of polls from Republican-leaning firms over the past few weeks, just as there was ahead of the 2022 election, when a promised “red wave” did not materialise.

It’s also possible that these two phenomena will mostly cancel out, and the polls will have their best year in a decade. After all, the polls today show a near repeat of the 2020 election, with almost every state polling within a point or two of the result four years ago.

If anything is easy to imagine in today’s polarised country, it’s a repeat of the last election.

This article originally appeared in The New York Times.

Get a note directly from our foreign correspondents on what’s making headlines around the world. Sign up for our weekly What in the World newsletter.