Paul Krugman is many things: a Nobel Prize-winning economist, a beloved professor, and a former New York Times columnist, among others. But now he’s also a fellow Substacker, publishing his eponymous must-read newsletter every weekday—and a special video feature on the weekends. So it was with the utmost pleasure that I joined his weekly video chat, along with G. Elliott Morris of Strength in Numbers, to discuss one of my favorite topics: special elections. At least, that’s where we started! With a fellow like Paul, I expected a wide-ranging conversation, and that’s exactly what unfolded. We not only talked about why Democrats are racking up massive overperformances in special elections, but we also explored: * Why polls don’t look quite as strong for the Democrats—or do they? * What might happen if Trump floods the streets with troops, L.A.-style, just before the midterms * How legacy media is failing to meet the moment—and why independent outlets are more important than ever And so much more. My only regret is that we stopped the recording before we chatted about our mutual love for Isaac Asimov’s “Foundation” novels (and why the TV adaptation is no good at all!). You can watch our entire discussion above, and you can also read a complete transcript below. I hope you enjoy watching this conversation as much as I enjoyed participating in it! PAUL KRUGMAN: So hi, everyone. As regular viewers know, I try on Saturdays to have a video conversation with interesting people. And this time is going to be a little different because I'm talking with two people, David Nir and G. Elliott Morris, with at least the starting topic suggested by David, being: what can we learn from the special elections that have been taking place? But I think this will probably devolve into a general discussion of the current political scene. And it also, maybe not so unusually, is a subject of which I know nothing. So I'm going to rely on two guys who do know something responding to questions from an ignorant interlocutor. So let's just see how this all plays out. And I'm going to actually start with David, because it was your idea that we have this conversation and talk about special elections, but I think that we want to get into some broader things. But why is now a good time to talk about special elections? DAVID NIR: So Paul, interestingly, you were the inspiration for this conversation because recently you published a piece in which you linked to something that Elliott wrote at his Substack, Strength in Numbers, and Elliott's piece relied on work that me and my team have done at our Substack, called The Downballot. We’ve spent many years looking closely at special elections because they can give us insight into the broader political environment because these are real elections. People are showing up to express their political preferences. And if you look at them the right way, they can actually tell you what people think, what is going on in people's minds, what does the broader political environment look like? And at this point, in this election cycle, in Trump's second term, we've seen about 30 special elections across the country. And we know that compared to expectations, Democrats are crushing it. They are beating expectations by huge margins. On average, when we look at them collectively, they are running 16 points ahead of where we would expect a normal Democrat to run in these elections. And if these patterns hold true, that would be more than enough for Democrats to take back the House next year. KRUGMAN: Okay. The thing about special elections, on one hand, they're real votes, and we want to talk a little bit about polling, but the good news is they’re real votes. This is not asking people what they think, which may be what they think you want them to think, or whatever. This is actual votes. But on the other hand, special elections are not a representative sample of the population. So what should we be making of that? NIR: I think you nailed that question there. It's a really, really important one because we know that the kind of people who show up for a special election, they tend to look different than the folks who show up in a midterm election and certainly from the folks who show up in a presidential election every four years. These folks tend to be much more politically engaged, typically higher educated, typically more affluent. Despite those differences, Paul, the fascinating thing is that if we look back several decades, we see that these special election performances, again, taken in aggregate, have typically correlated very closely with House general elections every even-numbered year. It's very interesting that there is this correlation because you might expect there not to be, but it turns out they do have predictive value, even though the electorate, like I said, tends to be better educated, more engaged than the usual midterm electorate. That said, in 2024, this normal correlation that we've seen historically, it broke apart. Democrats actually were doing pretty well in special elections, and they did poorly in the general election for the House. And of course, Republicans retain control. So the question is, will this correlation reassert itself in 2026, or is this correlation over because these electorates are too different? And I think that there are reasons to think it actually might reassert itself. And Elliott has done some work and has some thoughts, I know, on that particular question. KRUGMAN: So maybe this is for you to come in, Elliott. And you've also, I mean, obviously, you look at polls as well, the polls are telling a related but not quite as stark a story, as I understand it. But do you want to talk about where we are on all of that? G. ELLIOTT MORRIS: Yeah, and I'll just start that by asking a contextual question for David. So you say that 2024, 2025 is a 16-point swing in Democrats' favor, right? Is that larger than in 2018? Is that, like, notably significant? NIR: Yeah, it's pretty fascinating in 2017 and 2018. Over that entire cycle, we had more than 100 special elections. And in those elections, Democrats, compared to the presidential election in 2016, ran about 10 to 11 points ahead. So at this point, even though it's only June in 2025, Democrats are now running 16 points ahead. So this aggregate overperformance, that's the nerdy term we like to use, is bigger than it was in 2017-2018. And of course, we know that 2018 was a hugely successful midterm for Democrats, where they flipped 40 seats and retook control of the House. MORRIS: Yeah. So bigger than previously. This is surprising because that's not what we see in the polls. And if you take Donald Trump's approval rating as your benchmark, for example, he's at minus 10 right now. In the adult population, he's at closer to minus seven among voters. That's only an eight-to-nine-point swing since 2024. So, it does seem like there's a difference between these populations. Maybe the people who are turning out in the specials are, like, too Democratic than the overall population. That's my theory at least, and that's backed up by some of the voter file analysis. There's just Republicans dropping off. And if that's the case, then it slightly exaggerates the backlash. But there's still a really large backlash in the polling right now, a minus-seven approval rating. Among voters, minus 10 among adults for Trump is like a historically large deficit for a president at this point in their term. The only one that comes close is actually Bill Clinton. Now we're in Trump's 2nd term, and in Trump's first term, of course, he was even worse. KRUGMAN: Yeah. So Trump is slightly less unpopular if we believe the polls than he was at this point in his first term. But the special elections are basically saying that people really, really basically hate his party. None of these things are saying that Republicans are in good shape, but just how bad it is depends on what your data source is. Am I right? MORRIS: Yeah, that's right. And it looks like Democrats are fired up and ready to go to vote out Donald Trump, to use the old adage. NIR: As with everything else, polls, of course, come with errors. There's always a possibility that they can be wrong in one direction or another. And even a metric like special elections, when you're using it for this kind of historical analysis, also like any other thing that you measure, comes with error bars. And so they could actually be a little bit closer than you think, because if polls are slightly understating how unpopular Trump is and special elections are overstating how unpopular Trump is, maybe they actually meet in the middle. The thing I would say is answering a poll is obviously a lot easier than casting a ballot. Someone who is reasonably warm on Donald Trump, it's easy enough for them to answer a poll, especially these days. It's often by text message or on your computer, whereas actually showing up and going to the ballot box or turning in an absentee ballot, that takes a little bit more effort. So that suggests that if the enthusiasm is with the Democratic side, that could also help explain that gap. KRUGMAN: And also people…maybe “lie” is too strong a word. But I guess I'm allowed to use it now that I no longer work for the New York Times. One thing that struck me about one of Elliott's posts was on the question of asking people who voted for Trump, would you still vote for him now? And overwhelmingly, they say yes, except that other indicators suggest that quite a few of them would not. It's basically because people won't admit to making mistakes in the past, and that may be particularly an issue in polling right now. MORRIS: In our polling, 86% of people who voted for Trump say they would vote for him again, which is high, but that 14% defection rate is high enough to significantly change the outcome of the election. I would think changing vote preferences is really hard. Peo
