In this week’s politics chat, we examine what we know about how presidential elections work — or, what we think we know — and whether Donald Trump upends the apple cart. The transcript below has been lightly edited.
micah (Micah Cohen, politics editor): Let’s talk about Trump and what he means for the political science interpretation of elections. There are a lot of research-based maxims, but will they apply here?
To start us off: What are the main poli-sci maxims?
harry (Harry Enten, senior political writer): “It’s the economy, stupid.”
micah: How about: Partisan identification is the overwhelming driver of vote choice. And relatedly: The candidates don’t matter that much.
julia (Julia Azari, associate political science professor at Marquette University and FiveThirtyEight contributor): So I’d divide the poli-sci approaches into three rough camps: Group No. 1 focuses on the economy, as Harry said. The economy is what we like to call a valence issue; everyone wants a better economy. On the other hand, there’s camp No. 2, which focuses on party identification and polarization. How often are Democrats going to vote for a Republican because of the economy (or vice versa), in other words? The “campaigns matter” crowd falls into this group because campaigns help inform voters about which candidates share their positions. And finally, there’s group No. 3, which is much smaller, who see campaigns, events and other what we like to call “stochastic factors” as the main drivers of election outcomes.
And now everyone knows why political scientists are so much fun at parties — we use words like “valence” and “stochastic” to describe simple ideas.
micah: What about the idea that it’s really hard for a party to keep the White House for three consecutive terms?
natesilver (Nate Silver, editor in chief): I’m going to get in a lot of arguments with you guys about the two-term curse, which is based on a sample size of post-World War II elections and doesn’t really hold up at all if you look before that.
julia: Bring it! (You are certainly right about pre-WWII.)
natesilver: But I think it’s safe to say that political scientists agree there’s a significant advantage in being an elected, first-term incumbent (as President Obama was in 2012) and then there’s less agreement about what happens after that.
julia: Yes. One prominent forecasting model, Alan Abramowitz’s “time for change,” takes into account exactly that.
harry: You mean the “time for change” model that decided to include a polarization ad hoc variable, which proved to provide a less accurate forecast in 2012?
julia: Ha, yeah, that one.
micah: Shots fired!
natesilver: Guys, let’s not debate individual models. There’ll be plenty of time for that later.
micah: OK, so let’s take the idea that party ID is so determinative. There’s been a lot of talk about white, working-class Democrats voting for Trump and better-educated, suburban Republicans, particularly women, voting for Hillary Clinton. Is there evidence partisanship will be less of a driving factor this year?
harry: There’s not much evidence of that at this point, Micah. Trump is coalescing Republican support like prior Republicans. The question to me is more along the lines of whether people who had identified as Republicans no longer identify as Republicans or independent-leaning Republicans.
julia: I have two thoughts that sorta dodge this question in two ways. First, I think the sort of evidence that would address that directly is not fully available yet, though polling gets more predictive as we get closer to the election (the sky is also blue, the pope also Catholic).
Less tautologically, and with the caveat that Harry’s right about Republican support, one wonders if the campaign will reveal Trump to be some weird hybrid of a Republican and a third-party candidate because his views are so out there.
It’s hard to identify flat-out lack of preparation on policy issues on the left-right spectrum, but I’ll submit that this lack of preparation makes it more likely that Trump deviates (more than he has already) from standard party positions.
And there’s the whole “not actually a Republican” thing. I think it may matter that this is a pretty unique party candidacy.
micah: It’s hard for me to imagine Trump’s support among Republican women who live in the suburbs of Philadelphia being completely normal.
natesilver: I’d also point out that the advantage in partisan turnout and enthusiasm can vary quite a bit from election to election, and we don’t know a lot about how it will look in 2016 yet. There might be a lot of grudging Trump or Clinton voters who are showing up in registered voter polls, but who aren’t liable to show up to vote.
julia: Lots of talk about suburbs, but don’t most suburbs trend blue? (Except the ones a few miles from me here in Milwaukee!)
micah: There are lots of Republican suburbs in the Sun Belt too. The Southwest. The South.
harry: It’s still early, but Arizona looks vaguely competitive, which would be a sign that something is different this year. In 2012, Republican Sen. Jeff Flake of Arizona — he’s not a fan of Trump — nearly lost his bid for the U.S. Senate. And polls show both John McCain and Trump potentially in trouble out there.
julia: True. I would think the Atlanta suburbs or Phoenix ones might be more of an issue than Philadelphia.
micah: Yeah, that’s fair.
All right, so I’m going to pronounce the partisanship-drives-everything rule “shaky” in the age of Trump.
natesilver: Well, that might be going too far, Micah. There’s very strong evidence that polarization and partisanship have been on a historic uptrend. It’s one of the most important facets of American politics in the past quarter-century. There’s a question as to when and whether that trend might reverse itself, and whether 2016 could be a manifestation of that in some ways. And there are some technical questions as to how you measure partisanship and how it might affect how you build an election model. But I’d say the partisanship-drives-everything rule is holding up pretty well right now, including the House speaker endorsing the GOP candidate despite denouncing him for having made racist statements.
micah: Next up: It’s the economy, stupid.
julia: I learned a lot listening to Ben Casselman talk about this on the elections podcast. One thing that strikes me is that economic indicators are really ambivalent. A recent book by Marc Hetherington and Tom Rudolph suggests that polarization has brought with it some serious motivated reasoning when it comes to assessing the state of the economy. The president’s partisan supporters look at the indicators of strength; opponents find indicators of weakness. The “soft recovery” from the 2008 recession is especially susceptible to that.
micah: Interesting … so now the economy gets filtered through a partisan lens?
harry: Yes, I mean, I guess I’m not surprised by that. Everything goes through a partisan lens these days.
julia: Yes. Hetherington and Rudolph find that while partisanship is not a complete substitute for reality, Republicans and Democrats have pretty different perceptions of whether things have improved or not (they also look at national security), and this, in turn, affects how much people trust the government.
While I’m aggressively political science-ing here, I’d like to mention a qualitative study I’m reading about politics in Wisconsin, “The Politics of Resentment” by Katherine Cramer. She spent time talking to rural voters in Wisconsin, and there are lots of interesting points raised, but one is that people’s economic perceptions are not clearly distinct from their cultural identities and other non-economic attitudes. These things are often tied together.
I don’t know if Trump’s support taps into that exactly, but his main two talking points are trade and racial resentment/xenophobia.
natesilver: We can get into some relatively technical territory — or not — in terms of how you translate the economy into a forecast. At a very basic level, a strong economy helps the incumbent party and a weak economy hurts it. But the relationship is rougher than some of the models imply, in part because there are so many different objective and subjective ways to measure the economy.
julia: Yeah, this is one of the oldest debates in the forecasting literature and the economic voting literature more generally — is it GDP, employment, income/wealth, etc.? Perceptions of the economy in general or personal financial situation?
harry: Indeed, I’d argue that part of what a campaign is about is convincing voters that a candidate’s views on the economy are correct. That is, when the different measures conflict.
julia: “Are you better off than you were four years ago?”
natesilver: With a small sample — only like a dozen elections — I’d argue there’s really no way to know exactly which economic variables move the needle. Using some sort of economic composite is best, in my opinion (I have strong feelings about this), and that’s what FiveThirtyEight’s model does. Even then, the confidence intervals are fairly wide if you’re calculating them in the right way. Ben Lauderdale and Drew Linzer are good on this point.
harry: I think this year, as Ben Casselman pointed out on the podcast, there isn’t any real clear sign of where the economy is pointing. If it was really negative, then Trump losing would show that he upended the fundamentals. Conversely, if he won with the economy cruising along, it would suggest that perhaps he is a really good candidate. Instead, I think this is the perfect year where a good or bad candidate can break a party’s shot of winning.
And so far the evidence is that Trump is more bad than good.
julia: That’s a great point there, Harry.
One thing that I keep coming back to is that I think in some ways a Trump defeat is overdetermined. This is good news if you are someone who is terrified of a Trump presidency, but bad news if you’re someone interested in the substance of elections. I’m not just interested in which economic indicators make the best forecasts, but also in the question of why someone would cast a particular vote and what it means for parties and governance going forward.
micah: So that brings us to another maxim: Candidates/campaigns don’t matter that much. (Although maybe that’s a bit of a straw man?)
natesilver: I dunno if it’s that much of a straw man. I’ve gotten pretty deep into a couple of #nerdwars about it before.
julia: Agreed about the straw man issue. The study of presidential politics is like the worst experiment ever. Not enough repeat candidates to really get a sense of whether candidates matter. Like, we can probably conclude that Henry Clay, Adlai Stevenson and William Jennings Bryan were bad candidates.
harry: Heck, I’d even quibble about Stevenson.
julia: RIGHT! You try running a successful campaign against an affable, shape-shifting general who won World War II.
natesilver: In Senate and gubernatorial elections, where you have a much larger sample size, there’s evidence that candidates matter quite a bit. Or at the very least, that there are large residual effects that aren’t well-explained by fundamentals.
The argument is that presidential elections are different, though, because the nomination is a hotly contested, difficult process, which tends to produce good (or at least decent) candidates.
harry: And that’s what makes Trump so interesting. Most think Trump isn’t a good candidate.
micah: What are good candidate qualities and what are bad ones?
julia: Yes. And we are brought to the question of whether presidential campaigns are unique. This is especially interesting this year since the gender variable has now been introduced in the same study as the “random non-party xenophobe.” I would seriously fail someone if they turned this in in a research design class.
natesilver: If Trump loses by like 9 percentage points, though, isn’t that pretty strong evidence for “candidates do matter”? And if he wins or at least comes super close, isn’t that strong evidence that they don’t? I know it’s just one data point, but it’s a pretty darn interesting data point given that the “fundamentals” of the election are otherwise close.
julia: I’ve been giving this a lot of thought, and there was some useful psychology research that I referenced in my piece on electability a while back.
I think there are some straightforward ones: experience, perceptions of intelligence, authenticity. I also think perceptions of these characteristics are gendered and raced as hell.
harry: I would add that time in elected office is a big one in a lot of models that try to measure this. Trump, of course, is a zero on this score.
julia: Yeah. A lot of political science studies use previously held office as a measure of candidate quality. I’m working on something related to this right now … stay tuned/get excited!
natesilver: There may be some basic blocking-and-tackling stuff that Trump gets wrong, owing to a lack of experience as a candidate. And although you can cite counterexamples like Jesse Ventura or Arnold Schwarzenegger, in general those cases are the exceptions, and candidates without elected experience tend to do poorly relative to the fundamentals. Then again, Trump won the primaries, which is a significant feat.
So a big question is: Are the factors that allowed Trump to win the primaries pertinent to the general election, or were they unique to the Republican Party circa 2016?
julia: Yeah, I think what brings those three candidates together and distinguishes them from many other nonprofessionals is having some media experience.
natesilver: That’s interesting, yeah.
julia: Maintaining my dodging theme: I think the factors that allowed Trump to win the primaries are relevant, but not determinative.
natesilver: All three arguably took advantage of flukish circumstances, too. Schwarzenegger winning in a recall election with dozens of candidates on the ballot. Ventura being against two very unpopular opponents. Although, this reasoning gets circular. “Trump won because all the other candidates stunk” is sort of tautological.
julia: Plausible counterfactual outcomes to the 2016 race is an analysis that deserves more attention, I think.
micah: All right, are we missing any other maxims?
julia: I think the person who gets the most votes will probably win. (Though not necessarily!)
harry: Ronald Reagan had never won elected office before becoming governor of California. And yes, he was involved in politics, but Ventura (a small-town mayor) and Schwarzenegger were also involved in Republican politics for years, if not elected. So sometimes these guys can be stronger than we let on. Trump is arguably the least involved in actual politics of any of these guys, however.
natesilver: So let me ask an annoying question. Personal preferences aside, is there an outcome that would be better or worse, from the standpoint of the political science consensus?
harry: I guess the question ultimately comes down to: What is the political science consensus? I know that PollyVote, which is run by political scientists and averages all sorts of different types of forecasts, has Clinton as a favorite right now.
natesilver: Yeah, but that’s cheating by looking at polls and betting markets. The “fundamentals” models mostly show a tossup, or maybe even generic Republicans as a slight favorite.
micah: Wouldn’t Trump losing by 6 to 8 percentage points be a sign that some of the normal rules apply? Candidate quality matters, for instance.
If the economy stays so-so, we won’t get a good verdict on that.
julia: I hesitate to refer to anything as a political science “consensus,” but the fundamentals approach is pretty popular. However, the ambivalence of the fundamentals obscures the risks for political science. If Trump won despite the fundamentals favoring Clinton, then we’d probably implode.
Listen to the latest episode of the FiveThirtyEight politics podcast.
harry: I think we should be clear that there are plenty of political science models that look at the polls, so I don’t want to call it “cheating.” There’s more than just the economic fundamentals as far as political science is concerned.
julia: Oh yeah, even the most fundamental models take some polling into account — presidential approval if not election polls.
natesilver: Which is also cheating!
harry: I’d argue, for instance, that the president’s approval rating is a pretty big “fundamental,” and Obama’s approval rating is up and based on a regression (admittedly with a small sample size) would have Clinton favored.
micah: OK … closing thoughts.
natesilver: In my view, Trump’s a pretty good test of how much candidates matter. If he loses big in what should otherwise be a close election, that’s evidence that they matter quite a bit. And I’d argue that would go more against the political science consensus than toward it, although Julia’s absolutely right that it’s hard to define what the consensus is.
julia: My final thoughts: ambiguity and ambivalence are the watchwords of this political moment. Close elections, strong partisanship/polarization, new horizons in types of candidates. I’d also argue that both parties are moving on from the Reagan/New Democrat era, and so there’s murkiness about what the next sets of political debates (like in the general sense, not like podiums and moderators) will be.
natesilver: I also agree with Julia that Trump is extremely overdetermined. He’s unusual in so many ways. If he loses big — or wins big — we might be able to say that it had something to do with Trump, but we might not be able to say why, exactly.