Transcribe your podcast

Rationally speaking, is a presentation of New York City skeptics dedicated to promoting critical thinking, skeptical inquiry and science education. For more information, please visit us at NYC Skeptic's Doug. Welcome to, rationally speaking, the podcast, where we explore the borderlands between reason and nonsense, I'm your host, Julia Gillard. And with me is today's guest, Professor Brendan Nyhan. Brendan is a political scientist at Dartmouth College and a returning guest. He was on, rationally speaking a few years ago now.


I invited him back on the show now because like many people, I've been obsessively checking Twitter in the weeks leading up to the election and in the days following and out of all the voices on Twitter, I've just found Brendan to be one of the most careful and lucid commentators on what's been happening. So I actually reached out to Brendan right after the election to see if he could be stubborn as a last minute guest that week. And to his credit, he said he wanted a little more time to sort of carefully go through the data and figure out what happened, which, you know, I respect.


That's very in keeping with the spirit of rationally speaking. So but I'm glad you could join us. So welcome back to the show, Brendan. Thanks for having me. So today we're going to be talking about teasing out the different explanations for sort of what happened for why Trump won. And that'll involve like taking a look at the polls, you know, why or whether the polls were wrong. And just sort of more broadly, what kind of updates we should be making to our models of how the world works and to how democracy works, etc.


, from what's just happened. That's a lot to cover, but we'll we'll do the best we can. So, Brendan, as I'm sure you're well aware of, there have been a bunch of theories that have been popular trying to explain why Trump won the election, ranging from, you know, it Trump's fans felt left behind by the economy to Trump's fans are racist to Trump's fans or sexist to you know, Hillary was just unusually uncharismatic to the fake news, etc.


. And it seemed to me that all of these stories surely explain at least some of what happened. But I feel like it's a little lazy to just stop there. And I'm wondering if any of these stories seem to you to have sort of a particular amount of explanatory power like more than the others, or if or if we just can't tell?


Well, there's no there's no one answer or one explanation, as you suggested, as you were even asking the question. There's no we can't pinpoint one thing. That is the reason Trump won. And all of the stories after the election that suggested this is the reason Trump won, I think are misleading. That's not how elections work. And if you know, about one hundred thousand votes had flipped the other way, we'd be telling completely different stories now.


So it's important. It's important to be careful here and to think about. What we know about elections and to what extent this election corresponds, so in a lot of ways, this election obviously has been incredibly surprising, very different from elections in the past. In other respects, this election was very similar to 2012 and it's very similar to a lot of past elections. Most Republicans, despite all of the weirdness of this campaign, voted for Donald Trump.


Most Democrats voted for Hillary Clinton and it was a relatively small shifts in vote, shift in vote relative to 2012. That made a difference. That's ultimately what it came down to.


I was just going to say you could flip that around and say, you know, the fact that things sort of played out the way a normal election does is in itself surprising because Donald Trump is such an unusual candidates like that is sort of the surprising fact.


Yes, I agree. I agree on that completely. Yes. Yes. So if you had told us two years ago that Trump would do what he did, most people would have thought there'd be much more crossing of party lines than we saw in this election was a demonstration of the power of party identification, not just to convince people to vote for the candidate of their party, but the power that the that opposition to a disfavored candidate, the other party has to keep people from crossing party lines.


I think that's an underappreciated part of the story.


Right. And so you're you're saying that the breakdown was relatively similar to previous years with sort of more normal candidates on the Republican side? But was the was the like were the demographics of support for the Republican candidate this time also relatively similar to past elections?


No, the the. Most important, change in the composition of support.


For Trump, compared to 2012, for instance, which is the most immediate comparison is that he performed according to the exit polls, which are the only data we have now more poorly with college educated white voters than Mitt Romney did.


But he traded off a lower vote margin among folks in that group who turned out to vote for a higher margin among white voters who don't have a college education. That is a larger group as a percentage of the population. And they're concentrated disproportionately in those Rust Belt states that ultimately ended up making the difference in the election, the ones where we were the most surprised at the support for Trump when the data actually came in.


That's right, right. And is there any explanation for that fact, that sort of unique fact about the composition of Trump supports that seems to have explanatory power? I realize we're sort of in, you know, just like analyzing correlations territory. So maybe you can't say anything about it, but you may also have a hunch.


Yeah, I mean, I would say what we have now, our hunches, we have seen that pattern to be to be clear, that pattern of disproportionate support among non college whites and underperformance among college educated whites is one we saw for months in the polling. It's not new. It didn't materialize at the last minute. The reason the outcome was surprising is that Trump performed better than we expected based on pre-election polls. But the question, you know, we can talk about the question of why the polling was a little bit off.


Yeah, I hope to get to that. Yeah. But, I mean, you know, a more fundamental question is just why that the support looked like that in in general rather than the way the polling did, how well the polling did it, forecasting the vote. And and I think. What we saw was that Trump's appeal disproportionately to noncollege whites in.


The message he offered, the things that made him unusual, made him more appealing, he promoted a kind of white identity politics that resonated with some noncollege whites even as it turned off some college educated whites. And that trade off isn't a bad one if you hold most of the Republicans in your column electorally. Right. So it turned out to work out pretty well for him. There's a story here about the kinds of messages the parties offer, and he offered a different one.


And so he he shifted. He pivoted that Republican coalition just slightly in that that may have helped make the difference. So an alternative explanation that's pretty popular in my circles is that the reason that he appealed more to the non college educated white voters was not so much about white identity politics as it was about a frustration with either political correctness, like the language of political correctness on the from sort of the educated liberal class and or the like, smugness or or disdain that they felt from liberals, which, you know, I've certainly seen anecdotal evidence of this.


But at the same time, I'm I'm often kind of skeptical of explanations that, like this thing that has annoyed me for years just turned out to be the reason why Donald Trump wasn't right, because these explanations often come from people who themselves are frustrated with the language of political correctness. And so I like I discount that a little bit. So, yeah, my question, I guess, is whether you think there is any reason to prioritize the white identity politics narrative over the like, you know, backlash against political correctness narrative, or do you think those were the same thing?


I think I think they're more related than than distinct. I agree with your instinct that we should be wary of putting our own stories into the mouths and minds of Trump voters. The idea that they're closely following the details of speech codes on college campuses and things strikes me as silly. Most people don't pay attention to that stuff that, you know, there's a narrow group of people who read conservative publications who might be familiar with those sorts of things. But I think the average person just doesn't know.


Now, there might be a broader sense in which people feel a kind of resentment toward, you know, coastal elites. And and that might be part of the kind of identity politics that Trump promoted. Right. I mean, even the geering of journalists at his rallies had a function like that. That was a part of the show that he put on was literally to point and yell at the journalists at his rallies. So I do think it's part of this this overall message.


You know, I should just say again, though, it's important to be clear about the extent to which Trump was or was not surprising, the composition of his support. Was surprising, the fact that he held his party together, given what he did, was surprising, but he ultimately will end up. Near where we would have expected, given the conditions of the country are using the so-called fundamentals models that people in political science used to try to forecast elections when they know nothing about the candidates just using things like how high approval of the president is, how fast the economy is growing, and how long the the party in power has held the White House key those.


Those factors predicted a very close election, and so the the outcome was not a surprise in terms of those factors, those factors suggested the Democrats would face a very tough race.


It's just that when Trump came in and did what he did, people thought he would underperform. He had the highest unfavorable ratings in the history of modern presidential campaigns. Shouldn't that mean he would over perform as underperform, rather it would turn out? The answer seems to be not as much as you think.


Right. You also, along with a few other commentators, have been following Kevin Drum, have been pointing out that a lot of these like root cause, explanations about people's anxiety about the economy or their resentment of of, I don't know, the growing status, relative growing status of minorities relative to white people. You know, there's no one's really giving evidence that those those root causes have gotten worse in the last few years. So, you know, if they haven't gotten worse, if they've sort of, you know, maintained the same the same rate, then why, you know, is it actually fair to use those as explanations for Trump's win or is it just that they are explanations?


It's just that no one before Trump took advantage of those underlying root causes?


I think that's I think that's the key Trump. Appealed to people's racial resentment. He appealed to resentment of other minority groups and and so he often his support was more strongly correlated with racial resentment than Mitt Romney's. His support was more strongly correlated with negative feelings towards Muslims than Mitt Romney's. And so those it's not clear to me that those resentments are stronger than they were in 2012. The Republican Party just wasn't leveraging those. Mitt Romney was not appealing to those feelings in a way that Trump did.


And so this is, to a certain extent, a supply side problem. If you want to think about it in economic terms, the can and the kid, there's a candidate out there who's making ethnicity more salient to voters and appealing to people on that basis. The party nomination process in part, is to help parties choose candidates who fit the ideology of their of their members.


But it's also a candidate screening process. And in this in this case, it failed to stop someone who would appeal to voters on that basis. It's not clear that a similar campaign would not have worked in the past. And so, yeah, I think the root I think the root cause framework is the wrong when people tell these stories where it was somehow inevitable in this year. And I just don't think that's the case, that the failure here. Starts with the Republican Party allowing Trump to win the nomination.


There have been candidates like him in the past and parties have successfully coordinated to prevent them from securing presidential nominations. The failure here by the Republican Party to prevent Trump from securing that nomination allowed him to make these appeals to the broader electorate with the power of party identification behind him. And that's what's so powerful and that's what was missing before. Once he wins the nomination, he is a Republican and the party tribalism is going to hold the overwhelming number of Republicans in place.


Yeah, I definitely like in the set of updates to my models of how the world works in general from this election. One big update has been I just had this kind of instinctive sense that like the people in charge, you know, vaguely speaking, just like wouldn't let something crazy happen if it was at all possible to stop that crazy thing from happening. And I feel like one major data point against that model is the fact that Trump got the nomination, and especially that, like, as far as I could tell and I talked to some people who had sort of talked to insiders in the Republican leadership, there wasn't any real coordinated effort to stop him either.


It wasn't that they tried really hard and failed. Is that also your impression? And like have you do you understand why they didn't try harder?


Well, depends how you define try harder. They're their model. Let me say who they is. The Republican candidates who are competing against Trump had a model of how the world works that turned out to be wrong. Their model was Trump would fade. I expected him to fade. All previous political history expected him to fade. Celebrity candidates who've never held elected office might pull at high levels initially, but ultimately they've tended to fall off as the party coordinates around candidates with broader appeal who have deeper support in the party and so forth.


Everyone, myself included, expected that to be the outcome, and so they were all waiting to pick up Trump's support. They were playing the game of I want to be the last person standing against Trump, not the one to go down in flames while taking him down with me.


And so, you know, it was a painful, slow motion coordination failure where even though at different times different Republican candidates opposed Trump, they never did so effectively or in a coordinated manner. It's very difficult. There's no way to make binding agreements to coordinate among political campaigns. And so you could see Rubio and Cruz and Jeb Bush and everyone else struggle with how to do this. Right. So if if Marco Rubio is going after Trump, your incentive is Ted Cruz is to say, look at these guys down in the mud.


I'm the statesman who's rising above and vice versa. And as we saw, that didn't turn out to be a very effective approach. Yeah.


So maybe the update for me should be more like don't instinctively model, you know, quote, the people in charge or the decision makers as one block, but instead pay attention to, like, separate incentives that that could cause coordination problems and prevent any one subset from acting, even if it would be best for the whole group if they all acted or something like that.


That's right. That's right. Parties are these parties are less coherent, organized than they seem. And the Republican nomination process had features that made this coordination problem even worse. They had accelerated their calendar compared to twenty twelve, which gave them less time to coordinate and organize against Trump. And they also gave disproportionate delegates to the winner of states above their vote share that was intended to move them more quickly to a nominee.


There was a belief that Mitt Romney had been weakened by the primary process in 2012 they thought would be better to coalesce around a nominee to win the White House in 2016.


As it turned out, that helped boost Trump towards the Republican nomination, even though he wasn't winning 60 and 70 percent of the vote, he was winning 40 percent of the vote. But he was typically the plurality winner and getting additional bonuses and delegates are helping him push him closer and closer to the nomination. So those rules all ended up mattering as obscure as they are. And even as it became clear what was happening, no one within the Republican Party actually had the power to stop it.


Despite the fact that almost no Republican elected officials endorse Trump, we have no precedent for his candidacy. If you look at the in the modern period of presidential nominations, no nominee looks anything like him. We've never had a candidate with weaker support among elected officials in their party and. And yet they they failed to stop him, people either sat on the sidelines or they failed to coalesce among the other candidates, and they ultimately let it happen, despite almost none of them expressing that Trump was their first choice.


So one interesting theme I'm noting in the conversation so far is explanations about why now are about like why it makes sense that Trump won this particular year, sort of don't have as much weight behind them. And there are certain sort of almost accidents that made him especially able to succeed when past candidates in Trump's reference class have failed. How does that it seems like there's some tension between that theme and the general theme that I keep hearing in the media about the rise of kind of authoritarianism and nationalism around the world.


For example, in Brexit, like a lot of people see parallels between Brexit, which was like this rejection of cosmopolitanism and this turning inward and Trump's rise and a lot of the rhetoric that made Trump popular. But doesn't that kind of contradict the like? Why no explanations?


Yeah, no, it's an interesting point. I think we're still trying to figure out the extent to which what we've seen here is parallel to what's happening in Europe. There does seem to be a common element in terms of the messages that candidates are using. You can think of what Trump is offering as a as a kind of a. cosmopolitan message of the sort that has proved to be successful in some contexts in Europe. The EU makes for an especially fruitful target in their case because it is this transnational institution that exercises quite a lot of power.


We don't have an equivalent institution here. Instead, Trump. Focused more on the threat from immigrants and supposed waves of Muslim terrorists coming to get us disguised as refugees, but that was all. Bundled in with a similar kind of message. I think there's something there. I think the danger here is to turn the occurrence of an event into the inevitability of an event. What, what, what, what, what Trump what Trump did might have been a five or 10 percent chance.


And the fact that it happens doesn't mean it was inevitable it happened. There's always some baseline risk of of a candidate like that. Right. We've only had, you know, forty four presidents. And in the contemporary period, our sample size is really small. So we actually don't even have a great sense of how well our party nominations system works in terms of screening out candidates who make demagogic appeals. There just isn't enough data to say something with a lot of confidence.


I think we in my field put too much faith in that system based on that small sample. We thought it was more robust to two candidates like Trump than it turned out to be. And then similarly, after he won the nomination, people went through a similar set of rationalizing processes about the general election that it too would be more, would prove to be more robust to Trump than it turned out to be. But that may still also be a kind of.


Unlikely event now, as I said to you, the fundamentals suggested a close election, so once you won the nomination, it was going to be a lot tougher. You know, it looked for a long time, like Hillary Clinton had an advantage, but not by that much. And the margin of error was relatively small, even though the models were saying 85 or 90 percent heading into Election Day, that is not 100 percent. And as we saw it, you know, it was.


Yes, OK, that's a great Segway to talk about the polls in the models, because on the one hand, I I'm usually the person saying, hey, people don't complain that the predictions were, quote, wrong because predictions are probabilistic. And if you say something has a 70 percent chance, you know, 30 percent of the time it's not going to happen. And that's completely in line with the model that made the prediction. At the same time, you know, some of these models like Sam Wangs at the peak were, you know, ninety five.


Ninety nine percent. Surely we should be making some updates now that those predictions made the wrong prediction that there's something wrong with the way that model is generated. And so, yeah, my question is, are there any updates you think we can make with relative confidence, given the probabilistic nature of all of this, about how much to trust these kinds of models or about which kinds of models to trust? Yeah, so I would I would say two things we should look at the polls and the models, the the the models depend on input from the polls.


There are ways they can try to account for error in those polls. But the reanalysis that I've seen that have been most thorough in the wake of the election suggests that the main problem is just the polls were off. And it's important to even be specific about how the polls were off the national vote. Hillary Clinton won.


The polls were not going to be are not going to be off very far in in how much they missed on the national vote. The problem is that in certain key states, the polling was unusually inaccurate. And in a close with a close enough map, that meant that the the forecast, the forecasted outcome was wrong. So we're still figuring out. I think why the polls were were off in those states, and it may it may be something about.


How how the polls tried to estimate who would actually turn out, it's a it's a complicated issue. Polling is very hard because, you know, no one answers the phone. And so we're trying to build representative samples in other ways. And those don't always work as well as we'd like. So one issue is what's going on with the polls? I don't think we know. And then the second issue that you were getting at with your question is, what about the models?


Now, clearly, the ones that were most confident seem to have been far too certain. So the the Huffington Post pollster and and the Princeton Election Consortium models were both at Clinton at over ninety nine percent. Now, one event doesn't falsify those either, but they appear to have been overconfident. On the other hand, the upshot model where I contribute had Clinton in about eighty five percent and they show that the the actual outcome is well within the range of reasonable probabilities, if you will, get the distribution of possible outcomes under their model.


So it was somewhat unlikely or a 15 percent chance, but by no means vanishingly small. So, you know, we're going to have to think about this more carefully and and but at the same time, always keeping in mind that eighty five percent is not 100 percent. The example I used in a talk I gave after the election, which came from a tweet somebody wrote, was imagining this being the third election where there's been a lot of emphasis on forecasters.


And basically this is eighty five percent is almost the equivalent of the chance of getting a one if you roelofs excited, die. Right. And we we rolled it twice and it came up not one. And then the third time we rolled in, we came out one. And everyone's like, well, that means data has no value. Right. And and that's the conclusion we have to resist. There's more to elections than than forecasting outcomes. But also one election does not undermine the value of all quantitative analysis.


So I've tried to kind of push people to be a little more precise in in in how and the extent to which they update and what they update about.


Yeah, I, I had a lot of conversations with people in the weeks leading up to the election because I think I was a little bit more stressed out than my average friend. And and people kept pointing at the models as a reason not to worry at all. And I kept feeling like, look, you know, even like there's many credible models that have have Trump at, you know, 30 percent or so, which is like, you know, if you thought you had a 30 percent chance of, you know, severe complications from surgery, you wouldn't be like, oh, you know, I'm not worried at all because I have a 70 percent chance things will go fine.


You'd be worried. And I think people just, you know, they sort of their brain passes that 70 percent, as you know. Well, it's greater than 50 percent. Therefore, Hillary is the favored candidate. Therefore, I don't have to worry. I don't I don't fully understand what's happening in people's brains. But I had so many people arguing with me that, like, look, there's zero reason to worry. Trump has zero chance in the face of all this evidence, I think.


Yeah, maybe our brains just round up probabilities to to want to around down to zero. Yeah.


We have to figure out how to communicate probability in a more intuitive way. The upshot tried this idea of a field goal kicker, basically the yardage of a field goal as a way of translating their estimated probability. So if it were a certain percentage, you'd say this is the the the probability of Hillary Clinton winning is the same as the probability of a field goal kicker making a field goal from this distance, which at least has the property of something that happens most of the time if you're talking about a reasonable yardage.


But people have many examples in their mind of it not happening.


Yeah, yeah.


Because well, actually, that's why I was thinking about the surgery case that I thought that was something we're intuitively people actually do know how to think about probabilities, but maybe they don't. Maybe that was the.


Oh, no, no, no. I don't think they do this. Is this an issue in in business, in government, in medicine? Probabilities baffle our mind and we just have a really tough time with them. Julie, can I add something, though, on the you know, the the why the the polls were wrong that we haven't talked about values.


So I've heard two stories after the election that are interesting. I don't know if they're true, but I think they're they're worth considering. The one I put less stock in is the idea that consistent with what your friends were telling you, people knew these probabilities and therefore didn't try as hard. So basically those probabilities are assuming everyone tries as hard as they would have in prior elections. They're calibrated based on past elections. So if if everyone knows the probabilities and adjust their behavior accordingly, does that somehow hurt Hillary Clinton's chances now?


I don't think the group she underperformed with are likely to be the one. That are obsessively consuming five thirty, she racked up huge she probably racked up pretty big margins in the kind of coastal cities that are the reason her national vote, total popular vote margin is is well over one point five million at this point. So I don't think that's why. But it's at least worth considering what those mean when there's widespread awareness of the forecast. Yeah, that the second point, though, is, you know, we haven't talked much about Hillary Clinton's strategy, but that's the counterpoint to this, right?


The issue terrain in a presidential election is defined by the interaction of the two candidates. And normally that's been fought on this traditional liberal conservative axis about the size and scope of government. So think of Barack Obama and Mitt Romney talking about how to address how health care and so forth. So presidential elections typically center on domestic policy issues for the most part, and those are typically defined by that divide between the parties on what the role of government is. What was unusual about this campaign was that not only was was Trump pivoting towards this anti cosmopolitan appeal and away from a traditional small government message, but that Hillary Clinton kind of went along with it.


She thought she could peel off suburban women, Republican leaning or, you know, moderate Republican women, especially by emphasizing Trump's offensive statements and lack of qualifications for the presidency. That meant she didn't offer as strong a message about government. Standing up for you and being on your side as a Democratic candidate typically offers, including Barack Obama. And that's another reason potentially that that the that Trump may have over performed among those white folks who didn't go to college in the upper Midwest.


They may not have heard the message about who's on your side. And instead, it was Hillary Clinton's message about Trump made offensive statements and so forth. And qualifications may have played into this kind of cosmopolitanism versus nationalism divide that may have broken Trump's way in the end.


On the general topic of how to update models from surprising events, do you have any understanding of why the markets rebounded so thoroughly after like right after, I guess, the morning after Trump's win? I've read a bunch of explanations and I haven't found one that feels satisfying to me.


So one of one of the two authors of the main analysis of how the market moved during the campaign is down the hall from me here at Dartmouth Erickson Twits, who's an economist. He and Justin Wolfers, who's an economist at the University of Michigan, found that the market seemed to be moving quite significantly in overnight.


Futures markets downward. Whenever Hillary Clinton's fortunes in prediction market prediction markets fell during major campaign events, their estimates suggested that the market would be valued significantly less if Trump won. But then when he did, it didn't decline as much.


Well, it went it went really down. And then it went back up basically all the way.


Right. So that initial overnight, those overnight futures markets move negatively. But the when the market reopened, it's it's stabilized. And Justin Wolfers wrote a piece for the upshot where I contribute, trying to figure out what happened. He and Eric, I think, are still working on this. The theory they offer in this piece suggests and wrote is that the overnight traders are kind of guessing where the market will go. It's like a smaller, less thick market than the main market that's open during the trading day.


So these and so they may these folks who are reacting during the debates and so forth may not have been representative of the full opinion of the market as expressed after the eventual outcome of the election. But I think the honest truth is they don't know. Markets are even harder to parse sometimes than than voters. So I think I think we're still working on that one. Yeah.


I mean, it's very tempting to just say, you know, well, the markets are irrational now or I guess the markets were irrational back then when it when they reacted negatively to, you know, news that was good for Trump, but that I really don't want to lean too hard on. Well, guess they're just irrational theories because that sort of doesn't really allow us to to revise our model. It's like an easy out that I don't want to use unless, you know, all signs really point to it, I think.


Yeah, I agree. Oh, OK, well, we're just about out of time, but thank you, this was a pretty helpful post-mortem for me and I suspect for many of our listeners. So at this point, let's wrap things up and move on to the rationally speaking pick. Welcome back. Every episode we invite our guest on rapidly speaking to introduce the pick of the episode, that's a book or blog or article or something that has influenced his or her thinking in some way.


So, Brendan, what's your pick for today's episode?


Well, an article that I think your listeners might enjoy and probably find depressing is, is Craig Silverman's analysis at BuzzFeed of how widely shared fake news stories were and that platform during this election. You briefly mentioned this issue. I don't think it is, quote, the reason that Trump won, but this election did reveal that Facebook in particular is an important vector of misinformation. It's an issue our democracy is going to have to confront. Craig's article shows that completely fake stories are reaching millions and millions of people.


That's a problem for our democracy. Even if it doesn't change a single person's vote, it is contaminating the quality of democratic debate. It's making people less well informed. It's an issue we have to take seriously, and it's one that people can help address in their own daily lives by, for instance, reporting fake stories that they see on Facebook to try to get the platform to stop circulating them. So I think it's definitely worth people's time to check out.


That does indeed sound enjoyable and depressing in one package. Thank you, Brendan. I'm here to help. Well, we'll link to that on the podcast website as well as to your blog. And Brendan, thank you so much for for joining us on the show.


It was my pleasure. Thanks for having me. This concludes another episode of Rationally Speaking. Join us next time for more explorations on the borderlands between reason and nonsense. The rationally speaking podcast is presented by New York City skeptics for program notes, links, and to get involved in an online conversation about this and other episodes, please visit rationally speaking podcast Dog. This podcast is produced by Benny Pollack and recorded in the heart of Greenwich Village, New York. Our theme, Truth by Todd Rundgren, is used by permission.


Thank you for listening.