Harris leads in polling average in Pennsylvania and ties nationally

In the national poll, Kamala Harris and Donald Trump are tied among likely voters, both at 47 percent, a slight advance for Harris since our most recent national poll, conducted immediately before the debate.

At the same time, Harris held a four-point lead in a New York Times/Philadelphia Inquirer/Siena College poll of Pennsylvania, 50 percent to 46 percent.

Before we get into the details, let’s start with the big picture:

Not much has changed since the debate. Despite her strong showing, Vice President Harris didn’t gain much ground compared to our most recent polls nationally and in Pennsylvania. The poll is full of signs that our respondents thought Harris had a good debate—and Trump a bad one—but it hasn’t made a huge difference, at least for now and at least in our poll.

Pennsylvania, Pennsylvania, Pennsylvania. Harris may not have gained much, but her campaign will surely be happy with the numbers in Pennsylvania. The national result, on the other hand, is quite favorable to Trump (that’s the part that puzzles us and that we’re about to analyze). But our elections are decided in the Electoral College, and no state has a more relevant place in electoral math than Pennsylvania.

Now on to our conundrum: a clear lead for Harris in Pennsylvania, but a tie nationally? This is unexpected. Four years ago, President Joe Biden won the national vote by 4.5 percentage points, but won Pennsylvania by just 1.2 points. Similarly, our polling averages have shown Harris doing better nationally than in Pennsylvania. This poll is almost the opposite.

I would usually say this is statistical noise, the inevitable variation in polling results that is inherent to random sampling. And it may be, as we’ll see. But I think it’s hard to assume it’s simply noise, for two reasons:

It’s what we’ve noted before. It’s easy to dismiss any poll result as a statistical fluke. But we’ve found similar results in our two most recent polls nationally and in Pennsylvania.

This is becoming a trend for high-quality pollsters. Yes, our polling average finds Harris doing better nationally than in Pennsylvania, but the story is a bit different if we focus only on the highest-quality polls (which we call “select pollsters” in our table). Over the past month, many of these polls show Harris doing relatively poorly nationally, but doing well in the battleground states of the northern United States.

A note on “select” pollsters: To be considered select in our polling average, pollsters must meet two of three criteria: a track record of outperforming other pollsters, transparent methodology, and using a method that has a chance of reaching most or all potential voters. This isn’t a perfect approach (it leaves out some very good polls and includes some that aren’t so good), but it includes most of the heavyweights in the field and weeds out much of what doesn’t work.

If we focus on these higher-quality polls, we get a surprising picture: There are plenty of positive polls for Trump nationally, and plenty of polls showing Harris performing relatively well in northern battleground states like Pennsylvania.

A simple average of all “select” polls published since August 1 finds Harris leading by three points in Wisconsin and Michigan, and by two points in Pennsylvania and nationally. The Times/Siena poll doesn’t seem all that surprising in this context.

That pattern has been present for a while, but I assumed it was partly or mostly due to timing. Many of the most important national polls were conducted shortly after Harris announced her candidacy or just before the debate, while many high-quality state polls were conducted between the two times. This suggests that state polls might have recorded Harris at the height of her post-announcement rise, while national polls recorded her just before and after a political sugar rush. Today’s poll complicates this interpretation. And looking back, the explanation was never so simple.

Generally, lower-quality polls have shown the opposite relationship: better for Harris nationally, but better for Trump in key northern states. You might wonder why we include lower-quality polls in our averages, but many of them have value, and in some cases have even been more accurate than polls we typically consider higher quality. They carry less weight in our average, but they retain some importance, enough to sometimes override “select pollsters,” given their greater numbers.

What is clear is that recent results from the highest-quality polls are very different from those from the last presidential election. If true, it would suggest that Trump’s lead in the Electoral College, relative to the popular vote, has declined significantly since 2026.

This is not a surprise: almost exactly a year ago, I wrote that there were signs that Trump’s lead in the Electoral College was fading, including in the 2026 midterms. Indeed, today’s poll result is reminiscent of our poll before the midterms, which showed Republicans leading nationally, but Democrats doing well in Pennsylvania and other battleground states. It was hard to believe given recent history—I didn’t believe it, and neither did other pollsters I spoke to—but it turned out to be true.

We’ll revisit the case for Trump’s fading Electoral College lead shortly, including a dive into the geographic distribution of his strength in the Times/Siena national polls over the past year.

A mirage after the debate?
The period after a debate is always a tricky one for pollsters. One reason is that supporters of the candidate who won the consensus can be especially fired up and more willing to respond to a poll.

While the poll showed no major shifts, there were signs that Democrats were more willing to respond than usual, which may be cause for caution. Overall, white Democrats were 20 percent more likely to respond than white Republicans. (We limited the comparison to white respondents to isolate the effect of partisanship, not race.) And indeed, Harris’s strength in the poll was concentrated among white college graduates, the group you might expect to be most fired up after a debate.

There was also some indication that the most engaged Pennsylvania Democratic voters were especially likely to respond on the first or second night of polling, immediately after the debate. The share of voters who had previously contributed to Democratic campaigns, according to Federal Election Commission records in the L2 voter file, was unusually high during the first two nights, something not seen in the national survey. The debate was held in Pennsylvania, its media markets had the highest ratings in the country, and Harris campaigned in the state after the debate.

It’s hard to know whether all of this influenced the outcome of the poll, but it suggests that the debate influenced, at least temporarily, who wanted to participate in any polling. It’s also a reminder that the campaign plays out very differently in the most important battleground states.

Russian Doll Methodology
With this poll, we did something unusual: we ran a poll within a poll within a poll.

In collaboration with The Philadelphia Inquirer, this national poll was conducted and weighted as three separate polls: a poll of the city of Philadelphia, a poll of the rest of Pennsylvania, and a poll of the other 49 states and the District of Columbia. The three polls were combined (and balanced in proportion to population share).

If you’re wondering, the Philadelphia poll gave Harris a 79-16 lead (vs. 81-17 for Biden in 2026).

We’ve got some other fun national poll designs coming up next month.