Pennsylvania has been one of the keys to the Electoral College for decades, but this year it has reached Ultimate Swing State status. Nate Silver lists Pennsylvania as the most likely tipping-point state for either Kamala Harris or Donald Trump to achieve victory on Election Night, surpassing other perennials like Michigan and Wisconsin.
Both campaigns are treating the state as essential, flooding it with staff, money, and endless ads to prod exhausted voters to the polls and take home its 20 electoral votes. So which candidate is in a better place to do so? To get a sense of how things look in the Keystone State, I spoke with Berwood Yost, the director of the Center for Opinion Research at Franklin & Marshall College, who also runs the college’s closely watched poll.
The FiveThirtyEight averages have Kamala Harris with a 0.7 percent lead in the state right now. I know the latest F&M poll in August showed her up three in August, but do you see the race as the 50-50 proposition it’s being portrayed as?
The last two cycles have been exceedingly narrow races — a 1.2 percent margin and 82,000 votes last time around, and fewer than that in 2016. So history suggests that this will be close. There’s a good many Trump supporters here — we have a lot of white working-class voters, which is the base of his appeal. But I’ve also written about the fact that Trump is a known commodity. Is he at his ceiling or his floor in the polling? I think that remains unanswered.
Relatedly, I read something you wrote about how to define undecided voters. Only about 3 percent of people are truly undecided, you posit, but there’s a much larger bloc of moderates who may be leaning one way or another but haven’t fully made up their minds. And those voters don’t like Donald Trump at all, which tracks with your point that maybe there isn’t that much more room for him to gain.
I think that’s true. If you look at his favorability ratings in the state, they’re basically where they were in October of 2020. Everybody knows who he is and they know what they’ll get.
Polling shows that Harris actually holds her own pretty well with white voters compared to Biden’s 2020 margins, but that she’s still behind him among Black and Hispanic voters. This kind of surprised me.
That’s probably a working-class thing. The divide in this state to me, when I look and think about what’s going on here — one is an educational divide and one is a geographic divide. And the two work together.
If you look at what’s happened in Pennsylvania over the last 20-plus years, the state has flipped itself on its head. The Democratic strongholds of southwestern Pennsylvania that Al Gore won pretty comfortably completely flipped to Biden. And in the Southeast, you’ve seen the opposite. I’ve often wondered how fracking has played into this. Because if you think about the fracking belt in Pennsylvania, it runs from the southwest and comes across the northeast. Luzerne County was a county that Gore won with I think 54, 55 percent of the vote in 2000. And by 2020, Biden had gotten about 42 percent. For Harris, the question is: Can she cut the margins in those communities that support Republicans, and can she build or maintain the margins in those urban centers of Philly and Pittsburgh and the southeastern counties up through the Lehigh Valley? It’s very much the same strategy Biden used.
Pennsylvania is quite similar, demographically, to Wisconsin and Michigan. But Harris holds slightly larger leads in both of those states. These are small differences, but obviously small differences can matter a lot. Do you have any theory as to why Pennsylvania is stubbornly closer?
Your point is an interesting one, because if you think about the white non-Hispanic population overall in Wisconsin, it’s a larger share than in Pennsylvania. And there are fewer white working-class voters here, probably. Pennsylvania is an interesting mix: It has more foreign-born people than either of those other states. I believe we have higher rates of educational attainment. We also have more people who speak a language other than English at home compared to those two states — it’s about 12 percent here compared to 8 percent or 9 percent in those two states.
The other thing that’s different here is our median household income is higher. So it is sort of perplexing when you think about the college attainment and the higher incomes and the foreign-born population, which all seem to say Pennsylvania should be a better target for Harris than either Wisconsin or Michigan. But at the same time, Pennsylvania is famous for its localism. I think there’s a political culture here and you have to understand the political geography, and that probably plays into this as well.
Polling has been way off in all these states the last two presidential cycles. The F&M poll had Hillary Clinton up 11 points a few weeks before the 2016 election and had polls with Biden up several points shortly before the 2020 election. Both times, the state ended up being decided by a point and a half or less. And both results really surprised people because almost no polling portrayed the races as being so close. I know accuracy has been better in midterms, but do you think your polls and others have corrected the old errors? Or is that impossible?
How long do we have?
I’ll give you a few minutes on this one.
One of the things you’ll notice in our polling is we usually stop about three weeks before the election. We don’t pretend that we’re trying to predict what will happen. We say that this is a description of where things are and where they might go. And there is no poll of 500 or 800 people precise enough to find what is basically a 50-50 race. Polls are blunt instruments, and I think it’s important to remind people of that. I think one of the problems we see that comes from all this poll aggregation is it makes them seem like they should be perfect, and that’s just not what polls are. If you look at the internals of the polls, they can tell you a lot. They give you a sense of the race. But I encourage people not to overrely on what that horse race shows.
In terms of the Hillary Clinton poll, we had the misfortune of finishing our polling the day that Jim Comey announced that he was reopening the investigation into her email server. I mean, what’s the poll supposed to do about that? I think anyone who thinks that these polls are going to be precise — I don’t think that’s the best use of them. I think the way that you think about polling is, what are the issues that are driving people? Who’s undecided? What makes them undecided? Where could they go? The polls were pretty good in 2020 and 2018 because there were no changes in the last few weeks of the election. Things went the way they were going.
Sure, but there was also a consensus among people that in 2016 and 2020 there had been a systematic underweighting of the white, non-college-educated educated people who flock to Trump. In 2020, it was a little more confounding because people corrected for all those errors and still were off. I’ve talked to Nate Cohn and others about all the various theories of why that was. Maybe Democrats were responding more to polls during COVID, for instance. People don’t really know. But did you guys adjust for that this time around or use a different methodology?
Oh, yeah. Look, if you’re a pollster, your life is constant change. You’re always adapting your methods. In 2020, we were one of the few organizations that were weighting based on education. We had spotted that as an issue and were working on it. The timing of our poll was just unfortunate — you could see things bend after that in the state. And in fact, the polls nationally, those numbers were pretty good.
2020 was a weird year. It was the first year we ever had mail-in voting. And it happened in the midst of a pandemic where Democrats, from my understanding, were not out knocking on doors the way Republicans were. That probably explains why we don’t really know what happened, because it was sui generis. I’m not making excuses; I’m just saying. And so after 2020, we began additional efforts to reach voters, particularly younger voters. We’ve adapted text messaging now as part of our follow-up strategy, which we didn’t used to do. Pollsters are constantly trying to make sure we’re capturing an accurate representation of the electorate.
One theory is that there’s something about Trump being on the ballot that skews things, that there’s a bloc of people who vote for him and don’t really care about other Republicans, and that they’re very hard to poll.
If you look at what Siena College and the New York Times are doing — and we’re doing some of the same thing — it’s trying to make sure we capture people who are less likely to respond.
Lower-propensity voters.
It’s not just lower-propensity voters. The Census Bureau provides estimates at the Census-block level of responsiveness to surveys. And you can use some of that data to help identify places where people would be less likely to respond. So we’re trying to get people into our samples who are harder to reach, less likely to cooperate. We’re all doing that, but there’s a danger to it. And the danger is you overrepresent people who won’t be there on Election Day. The other thing I’m thinking about is if you go to one of those AAPOR reports and look at polling variations — I don’t like to call them polling errors — around elections going back to the ’30s, it’s rare that you have more than two presidential cycles in a row where the error is in the same direction. So I think it’s also fair to think, Is it possible that maybe all these corrections —
That you’ve overcorrected.
And in fact, that’s why I wrote the piece about undecided voters, to say, “Hey guys, polling is an effort to capture how people are feeling at a polling time.”
But people are people, and they could change their minds. And the way polling is done, many pollsters are encouraging responses and discouraging the “don’t know” category, which is why I think it’s important to ask about the certainty of the poll. And that changes the picture.
I think there’s still people who are thinking about it, who maybe they’ve made a choice and maybe most of them will go with that choice. But I think the most telling part of the piece is if you look at the factions. And it does make sense that the traditional Republican faction still has their doubts. Maybe they’re going to vote for Trump at the end of the day. In these polarized times, that’s certainly likely. But there’s some doubt there. And frankly, the drum beat of Republican officials, starting with Dick Cheney …
I was about to bring up Cheney. Maybe he’ll swing this election, which would be unexpected, to say the least.
Oh my God, and Alberto Gonzales Thursday morning. There’s enough permission from these Republicans that it could be determinative among a small group of people. I also do think those true independents, which are probably 8 to 10 percent of the voters, they’re the ones ultimately that will make the choice. And if they break in one way or the other, that’s who wins.
My only data point is that my gun-loving libertarian friend who lives outside of Scranton told me that all the guys at his country club who had voted for a mix of Democrats and Republicans in the past are all going for Trump this time.
That’s Luzerne County, though, right?
Yeah.
Luzerne and Lackawanna will be part of it. But I’m looking at Erie, I’m looking at Northampton, I’m looking at Monroe County. Those are the three swingiest in the state. Biden won Erie, but he squeaked by, didn’t even get 50 percent. And that’s the other thing that we’re not talking about here, is there will be at least four candidates on the ballot. Maybe five.
Remind me if RFK Jr. made it?
No, RFK will not be on the ballot, but Jill Stein will be. There’s a libertarian, Chase Oliver, on the ballot. And I believe there’s one other. So if you go back to 2016, actually even to 2020, and look at our final polls, even though they’re three weeks out, they pretty much nailed the Democratic vote share. But the problem for Hillary was that the third-party candidates pushed the race to Trump. If she’d got Jill Stein’s voters, she would have won. I don’t know how many pollsters include all the potential candidates, but we certainly do that on our part.
The infrastructure in the state of the two campaigns seems pretty lopsided. Harris’s campaign has opened many more offices than Trump’s team has. They have more money, and they’re spending it freely, though both sides are blanketing the airwaves. From what I read, they also just seem better organized. Does that stuff actually end up mattering in the end, do you think?
Yes, I do. Absolutely. And that’s another thing that polls have a tough time capturing. That what happens on Election Day to get people to the polls — that can make a difference. When you’re talking about 40,000 or 80,000 votes, the last election, every little thing potentially could be the difference. And I think having a good get-out-the-vote effort is fundamentally important. The Trump campaign is testing a theory about volunteers and targeting low-propensity voters. The theory of the case that they’ve put forward is that there’s a significant number of voters who don’t vote regularly, who feel like Trump represents them, and their efforts are to get them to the polls. So it’s less outreach to the middle and more outreach to those voters. I think the Harris campaign seems to have a lot of enthusiasm, and donors and volunteers seem to be far ahead of where Trump is. But we’ll have to see.
Another thing I’m keeping an eye on is voter registration in the state. Pennsylvania is one of those states where you register with a party, and as the rules stand now, you have to be registered as a Republican or a Democrat to vote in those primaries. In 2008 and 2009, the Democrats in this state had about a 1.2 million voter advantage. That’s shrunk to about 350,000, 360,000 voters. So the patterns in the state have just sort of been more and more competitive.
At the time of the 2016 election, the gap between the registered voters in the state was much bigger than it is now. And so we will see if that has an effect. I suspect what that is mostly is people’s registration catching up to their voting behaviors. But it is something I keep an eye on, and we’ll see if it’s indicative of enthusiasm. There was clearly a lack of enthusiasm in the state until Biden left.
What about Harris’s vice-presidential pick — her decision to choose Minnesota governor Tim Walz over Pennsylvania’s popular governor Josh Shapiro?
I wrote something before she picked Walz about why Shapiro would make such a good running mate for her. And I’ve talked to some Republican county chairs about her selection. One of them just said that was a fatal mistake. Another one said, “I’m going to send her a thank-you note.”
Shapiro strikes fear into the heart of Republican leaders in this state because he’s a pretty savvy politician who seems, to many of the state’s voters, to be more moderate than progressive. And you look at his stance on fracking, for instance. Or the fact that the first executive order on the first day he was in office was to open state jobs to people without a college degree.
There’s that localism to Pennsylvania politics. I think if she had picked him, she’d have won the state. I think she still can win the state. But then again, we have no history of electing female executives at the state level. We’ve never had a governor, a female governor. I don’t believe we’ve ever had a female senator.
This interview has been edited for length and clarity.
