Transcribe your podcast
[00:00:30]

Welcome to another episode of Conversations with Coleman. If you're hearing this, then you're on the public feed, which means you get access to episodes one week late and that you'll hear advertisements. You can become a supporter at Coleman Hughes, dawg, and gain access to the private feed, which has no ads, gives you access to episodes one week early and it gives you an extra podcast episode every month. If you want access to the private feed but can't afford it, please email admin at Kohlman Hughes Dog and we'll help you out.

[00:01:02]

My guest today is David Shaw. David Shaw is a political data scientist, he worked for the Obama campaign and was described in New York magazine as Obama's in-house Nate Silver.

[00:01:15]

You may know David's name because he was fired in June over a tweet that was widely viewed as innocent. The tweet cited research showing that riots, unlike nonviolent protests, tend to move voters to the right. I wanted to talk to David about the circumstances of his firing, but for legal reasons, David can't talk about it publicly.

[00:01:39]

Fortunately, he is an extremely interesting person outside of his cancellation, and he's the perfect person to talk to about the lessons of the twenty twenty election. David was really a joy to talk to and a fountain of information, and I recommend you all follow him on Twitter. We talk about the basics of polling, we talk about why the polls under predicted Trump's performance in both twenty sixteen and twenty twenty. We talk about David's skepticism of the so-called Scheid Trump voter effect, used as a tool to understand why people voted for Trump.

[00:02:17]

We talk about why Trump's share of the minority vote increased between 20, 16 and 20, 20. We talk about what the twenty twenty election has done to the idea that demography is destiny. We talk about the implications of getting rid of the Electoral College, the surprising change in how superPAC money backs politics, how the protests and riots affected public opinion. And we talk about the future of Trump ism without Trump. One final note here, David's job has been to get Democrats elected, so he is a partisan, not in the pejorative sense of the word, just in the descriptive sense.

[00:03:02]

I say this only because he often uses the words we and us in this podcast to refer to the Democratic Party. And as I made clear at one point, I'm an independent, so I don't speak that way. In any event, I think Democrats would be very wise to listen to him, so without further ado, David Shaw. OK, David Shaw, thank you so much for coming on my podcast. A pleasure to be here, honored to be invited.

[00:03:34]

So we have a lot to talk about, given that it's November 13th, that we're speaking just a little over a week after the election and before we get into everything, and I'm really excited to have this conversation, I think you're the perfect person to help me and and my audience pass what is what we've gleaned from the election so far, what models of the American electorate turned out to be right and which turned out to be wrong and what we can expect from the constantly shifting coalitions of the two parties going forward and what we've learned.

[00:04:13]

What surprises and what expected results, we've we found out a week and a half ago, but before we get to everything, can you just give people a summary of your background, how you got into political polling and analysis and consulting work that you do? Yeah, sure, absolutely. I'd say I I was really lucky I was a math major in college and, you know, kind of as I was going through studying pure math, I kind of realized, though I don't think that's what I want to do with my life.

[00:04:50]

And this was in two thousand seven. And so the data science didn't really exist yet. And there was like a lot of a lot more low hanging fruit just in terms of applying data to anything. And I was pretty into politics. So I started a blog that very, very few people ever ran. But me and my friend, we ran it for a couple of years doing like Nate Silver style things. And then I applied to work at the Obama campaign in 2012.

[00:05:15]

And then I was part of the analytics team there. I built their electrical forecasting system. And after that I started the analytics team at the Obama campaign, started a company called Civis Analytics. I was part of the I went over there with my boss and it's kind of part of the founding team there. And then, you know, Civis played a pretty big role, I think, in a lot of ways. They've been like the IBM of political data science in a lot of ways.

[00:05:40]

And I worked there until earlier this year.

[00:05:43]

So before we get into the election and what we've discovered about the electorate, I want to just. Rehearse a few of the most basic points about polling that you probably haven't thought of in a decade but are worth reminding people, there's I think many people feel like they themselves have never responded to a poll.

[00:06:09]

And in view of the failed predictions of polling in twenty twenty, in twenty sixteen, or at least the apparent failures, it might be, it is worth going over. Very simply, why it's possible to learn anything of value from a poll of five hundred people say in a country of three hundred fifty million and maybe talk about the concept of randomisation and why that's important and how pollsters achieve or fail to achieve that. Sure, I think that I don't want to talk of polling.

[00:06:45]

We'll get into it. I think people have a right to be suspicious of polls and a lot of ways. But just to talk through statistically, this underpinning of polling, this idea that when you if you talk to you call a thousand people out of the three hundred million people that are there, the underlying theory is that if you take a sample and you ask them who they're voting for, that basically even though you're not talking to everyone, that's a lot of people.

[00:07:14]

And really kind of the margin of error for talking to a thousand people is very low. And I think looking at it purely in that simple example, I think if you went out to a store and if you and if you are if you walked outside and you looked and you just kind of counted how many people there were, and you did that every single day, a thousand times, and you average the results, I think that you'd end up feeling relatively confident.

[00:07:39]

And so that's that's kind of the idea is just that sampling does work in theory. I would say, though, that I think the real problem with polling is that even though it's theoretically possible to learn a lot about a large group by talking to a small number of people, subsample of people in a group, and that's kind of the underpinning of a lot of modern society. The problem is that that really relies on this equal sampling assumption that you have hundred million phone numbers and you're calling everyone and everyone is just kind of into everyone's chance at answering is exactly the same.

[00:08:17]

You need all of that to basically be true in order in order for a lot of this stuff to work. And we don't live in that world. The reality is that some people are substantially more likely to pick up the phone than others, some in ways that are very well understood. Older people pick up the phone, more women pick up the phone, more white people are more likely to answer the phone. And when you put that all together, you know, older white women over the age of sixty five, something like one hundred times more likely to answer the phone than Latino men under the age of thirty four.

[00:08:51]

That's a lot of bias. And if it was just those things, we could correct for it. But it turns out that what drives people to answer phone calls are not driven by a whole host of factors that are very hard to control for whether it's their trust, how nice a person like how agreeable they are or how much trust they have in the system. Or it could be that there are Democrats and Democrats are unusually enthusiastic. This is the news that it happened this week.

[00:09:17]

And so adjusting for all that stuff is extremely hard. And I think that people are very reasonable. I think their intuitions that that this is something that is hard, that you can just like dial a bunch of people and then say something about everyone else. I think that that skepticism is founded, know we do the best we can. And I think that on average, to me, the surprising thing is that polls are ever right at all. And I think they they do have a somewhat reasonable track record.

[00:09:45]

They usually get things through within three or four points, which is nice. But but it is an extremely hard thing and people mess it up all the time. Yeah, so the polls in twenty sixteen and twenty twenty, both under, predicted Trump's success. So the question is why?

[00:10:04]

Where was the same mistake made twice or different mistakes made, or is there some third explanation? Yeah, I think it's worth decomposing what happened into two different stories in twenty sixteen, the national polls were not particularly wrong. I think the final average was three point four and the final result was two point two and margin. So not not terribly different. But on a state level, the reason why Trump's win was a surprise was because basically all of these states in the Midwest saw these giant swings toward Donald Trump that the polling didn't see coming and that I guess not just in the Midwest, really, if you look state by state, the biggest predictor is that states that had a lot of noncollege educated whites, but particularly in the Midwest, or at least particularly outside the South, saw very big polling errors.

[00:10:57]

But the national numbers were about about right. And so in twenty eighteen people, the national numbers were also about right. But the same polling errors actually occurred. Again, if you look, Democrats really underperformed expectations in these working class white states like Ohio or Michigan or West Virginia is the most extreme one. And then we flash forward to twenty twenty. And this exact same set of biases happen again. People make these spots and you can just see that really the polls were wrong in relative terms and basically exactly the same places for three cycles in a row.

[00:11:31]

But there was another phenomena that did happen, which is that in twenty twenty, the national polls were not just about right. The national polls were pretty wrong by historical standards. It looks like we're going to end up with something like a Biden plus four and a half, maybe five point race in the national popular vote once every all the votes are tallied. And that is not what the average, but what the national polling average was. You can see this pretty down the line, not just in the presidency than the Senate and the House, where everything was roughly two to three percent more Republican than everyone predicted in a relatively uniform way, a little bit more in Republican states, but fairly uniform.

[00:12:17]

And so these you had these two things happening simultaneously and they had a similar, but they were two different phenomena with two different causes. But I think that in a lot of ways they come from the same phenomena. And just to talk about that, when people talk about polling error, I think people mis focus on what actually makes polls wrong. People really like to talk about turnout. They like to talk about undecided voters. They like to talk about what we call in the industry movement.

[00:12:48]

The polls are wrong because everybody change their mind. At the last second, the undecideds broke our way or one side really mobilize their base and the other side didn't. And that's actually usually not why polls are wrong. All of those things are possible. And I forgot also, people have this Trump hypothesis that people are lying to pollsters. As far as we can tell, that's actually a very rare thing.

[00:13:08]

It usually does not happen actually worth dwelling on for a second, because I've heard a lot of people believe that. So what's the before you continue, what's the evidence that that's actually not the case?

[00:13:19]

There's a couple of pieces. The first is that we can one thing people like to do for this is they to get at this social desirability as they look at are there differences between how people answer online versus on the phone that people there's we can reinterview people multiple times and see how consistent they are. And the other thing is that we can match survey respondents to the voter file. And something that we can see is that really the vast majority in the voter file, we have things like party registration and we can see donation records.

[00:13:52]

We can see this is a very strong relationship between what people say and what people end up doing, both in reinterviews and with external information that we have about them. The problem isn't that Trump voters weren't you were lying. The real issue is that they just weren't picking up the phone. And that really gets to this core problem of why polls are actually usually wrong. It gets to this thing you asked me about a couple of minutes ago, which is that when you call a bunch of people polling working is built on this idea that the people who answer the phone are the same as the people who don't answer the phone once you control for the things that you wait on.

[00:14:36]

And traditionally, pollsters have really waited. And when I say weighted, I mean, if you survey a bunch of people and the sample is 70 percent women because women are more likely to pick up the phone. And, you know, from the census that the population is 50 percent women, you kind of apply a weight to the women's weight. The response is down and you wake the men up and then things represent the overall population. And so basically, if there are things that you aren't waiting on that are influencing someone's chance of picking up the phone and those things are correlated with who they're going to vote for, then your poll is.

[00:15:12]

To be wrong, and so that's why pollsters do do waiting. But historically, they've only waited on a very small number of factors, age and race and gender. And most recently, after twenty sixteen, they started to do education. And the problem with it is it's very hard to know when you do this process. You simultaneously have to ask people to think or have the information on the survey respondents. But you also have to know the true value and that's easier for a gender and the voter file.

[00:15:41]

There's a gender you count with. The census will tell you. But if you want to know, even for education, as you people with college degrees are more likely to vote than people who are. So even if the census tells you that 30 percent of people have a college degree, it's very hard to know what percentage of likely voters have a college degree. That's a very hard problem. But as a result, pollsters have waited on a very small number of things.

[00:16:04]

And most polling error comes from when something is predicting a response that they weren't controlling for the industry term for this is certainly not bias. I think people find it a little non-intuitive when I say this, but the important point is that response rates are very low now. It used to be going back to the nineteen forties that Gallup could call people and you would get an 80 percent response rate. And I think that's because the world was a lot more boring back then.

[00:16:35]

So someone good people would get a call and go, oh my God. A researcher wants to know my views on contemporary events and the world isn't like that anymore. Now, going into twenty, twenty sixteen, which was the last year that my former employer, the phone calls, we were looking at zero point eight percent response rates. And so if you're in a world where Democrats are picking up the phone at one point four percent and Republicans are picking up at one point one, that will actually generate a very significant bias.

[00:17:03]

They'll be very hard to control for. And so that's where most of this bias comes from. Instead of just briefly, what happens when I think you mentioned it when we spoke before this podcast, but what happens when the responding population differs from the non responding population on a trait that's completely invisible but still correlated with political party like a big five? Personality traits, say people who respond, are more likely to be to have some kind of personality, to be neurotic.

[00:17:36]

And that correlates is to me anyway. Is that is that a permanent limitation? Well, that's the whole problem. Then your poll is wrong. And that's what's so horrifying about what we do, is that you really need to control for as much as you can if you want your polls to be right. And that's that's something that's very hard.

[00:17:53]

And just to talk through these two different biases, you know, first, the one that's appeared that appeared in twenty, sixteen, twenty, eighteen, twenty, twenty, and then the second one that appeared just in twenty twenty. The first one after twenty sixteen. We are polls were wrong, just like everybody else's. And we did a lot of soul searching to try to figure out why. I think a lot of pollsters were like oh the polls are fine, slight movement, whatever.

[00:18:15]

And we really wanted to be honest about the fact that we were wrong and that our measurement apparatus wasn't working. And so we spent a lot of time and one of the trying to understand what what went wrong and one of the things that really came out was this idea of social trust. It turns out that's that's a social science term, little academic, but it gets operationalized in this question of do you believe that people can generally be trusted or do you think that people should keep to themselves?

[00:18:42]

And something that we that we saw and this is something that actually researchers have known for a very long time, is that people who say that people can't be trusted are substantially less likely to answer phone surveys or really any kind of survey, that their areas also have lower census response rates. And a lot of ways that's very unsurprising. But that used to not matter. This bias has always existed. The question is like, why? Why did that start making the polls be wrong in twenty sixteen?

[00:19:10]

And the answer is really interesting, which is that until twenty sixteen, social trust used to not be correlated with partisanship once you controlled for age and education and all these other things. But then political coalitions change over time. Know the story of twenty sixteen was that a very large number of non college educated white people swung toward Republicans in a very large number of college educated white people, swung toward Democrats and those non college educated white people who swung against success were overwhelmingly had much lower than average levels of social trust.

[00:19:46]

And so if you look actually among non college educated voters, non college whites who say that people can be trusted actually swung toward Democrats slightly, while the ones who say that people can't be trusted swung heavily against us. And so public polls were surveying this like high trust the universe. And as a result, it or in particular this high, high trust subsample of non college educated whites. And there's like a lot of other trust as one dimension. There's a lot of other ways these groups are different.

[00:20:16]

Their occupational divides to being a teacher is that even though you don't have a college degree, there's a big difference between being a teacher's assistant versus being a construction worker. We found that working in an office as opposed to not working in the office was like a very strong predictor among non college graduates for swinging toward Trump. And so that polls were surveying this group and getting the wrong answer. And that's why these biases were the largest in these states that had the most non college educated whites like the Midwest.

[00:20:47]

Media bias is one of the great problems facing our democracy. It doesn't matter how smart you are or how much time you spend consuming news, if your news diet is unbalanced, you are very likely developing a false picture of the most important news stories of the day.

[00:21:04]

To combat this problem, I've recently been using something called a ground news, ground news is both a website and a smartphone app that collects the most important stories of the day, along with the various articles that cover that story and then sorts the articles by their political bias in a user friendly way. So, for example, I'm recording this on November 21st, when the biggest story is that a federal judge threw out President Trump's lawsuit requesting that the results of the Pennsylvania election not be certified.

[00:21:36]

So I can click on that story and then get a collection of links sorted by political bias. I can then check out how the left is covering this story and how the right is covering it. But the most important part of this app is a feature called Blind Spot. If your news diet is unbalanced, then every day there will be stories that you simply don't see. There are whole topics that the right is not interested in covering and likewise for the left.

[00:22:04]

So, for example, today on Ground News, I can see that the left is more or less ignoring the fact that ISIS launched rockets into a residential neighborhood, killing eight people and injuring two dozen more in Afghanistan. And I would guess the left is not so enthusiastic about stories like this because they are hard to square with the narrative that jihadist violence is an understandable reaction to American imperialism. Meanwhile, the right is barely covering the fact that covid-19 cases in the US have surpassed 12 million today and the virus seems to be spreading with a renewed vengeance.

[00:22:49]

This obviously does not make America or the Trump administration look very good. So the right is not so interested in it. So that's the kind of thing you can learn every day with the ground news app. This is a great tool to have if you're interested in having an accurate picture of reality, so I'll put that link in the description and you can all try it out.

[00:23:17]

I want to pick up on the last point you made about the different types of people that are, quote unquote, non college educated, and you mentioned that if you work in an office or you're a teacher's assistant, that marks you differently in an important way than being a non college educated construction worker. And what that says to me is when we talk about educated versus non college educated, we're using education as a proxy for culture, for some kind of cultural divide.

[00:23:52]

And I was recently speaking to someone who was wondering because superficially you could see, OK, college educated whites go Democrat, non college educated whites go Republican for Trump. Isn't it just the case then that the more educated you are, the more sort of accurate a model of the country you have, the more likely you are to vote for the party? That's better. I mean, that's obviously I'm an independent, but that's something a partisan might think or just a rational person might think based on that split.

[00:24:31]

But it seems to me the divide is. It is not even really about education so much as it's about the culture that you absorb in an environment like an office, a college or a school. Does that seem right to you? I think that's 100 percent right. We talk about in social science, this concept of socioeconomic class and, you know, the idea there being that measuring class is very, very difficult. Education gets you some of the way there.

[00:25:03]

But if you are a NYU dropout in Williamsburg as a biracial, you have very different values than if you're a nurse that happens to might be classified as having an advanced degree. And so I think that when we look at education, that that is a marker for a culture. Something that's really hard is that there are other ways you can cut this, too. You can look at social trust and trust in institutions, which I actually think is also highly correlated with education and is also honestly probably a class or a cultural marker.

[00:25:34]

But then there's the spicier thing you can look at is something called racial racial resentment. In the end, he obviously has a meaning in English, but in political science it has a particular meaning where you ask like a certain class of questions in surveys, you can't just ask people, are you a racist? So I think it'd be fun to try. I don't know how many people would say yes. But the idea is you can't say that. But clearly, racism as an idea exists in some way.

[00:25:59]

There are people some people are more racist than others in the abstract. And so people ask these questions that you try to measure to try to get at that. And so examples of this are asking people, how likely do you think it is that there are a lot of white people who are having trouble finding jobs because non-white people are getting them instead? Or do you think that white people have enough say in how the country is getting run or more classical?

[00:26:27]

One is, do you think that discrimination is the reason why African-Americans haven't been able to get ahead the same way that Italians and Jews and other other other disadvantaged groups have? I didn't I don't make up these. I don't make up this word. These aren't claims for me or questions I'm raising. These are just standard academic questions. And so if you create an index of these, you ask a couple of these different questions and you average them. Something that really came out in twenty sixteen was that if you statistically control for racial resentment, then education becomes totally uncorrelated, you know, low racial resentment.

[00:27:00]

Working class white people swung toward Democrats and high racial resentment. College educated white people swung toward Republicans. And so that does capture a lot of the story.

[00:27:11]

The only problem there is that a lot of people, a lot of critiques people will make of racial resentment as a concept is that racial resentment itself is kind of a class marker. Like if you're a highly educated person, you're taught. I'm not supposed to say that black people should bring themselves up by their bootstraps or whatever. Yeah, you're like socialized against it. And so they're like it's a really tricky thing that if you look at twenty sixteen, they're a bunch of different things, you know, another one is openness to new experiences.

[00:27:47]

You know, you talked about psychometric traits, openness to new experiences. Seems like it was very highly correlated with swing against Trump. People who have very high levels of openness trended democratic and people with low levels of openness. And Republican Obama Trump voters as a group were a very were more openness than any other combination. But again, that is something that's super correlated with education. And so it's this really annoying thing where you have all of these similar but distinct things, openness to new experiences, racial resentment, education, socioeconomic class, trust in institutions.

[00:28:21]

And they're all correlated with each other and they're all clearly part of the same story. But it's very hard to disentangle exactly what caused what.

[00:28:29]

I want to dwell for a moment on this notion of racial resentment, and I've often wondered what exactly we're measuring when we ask someone a question. Like what? When you say, do you agree that black people should work hard like every other group or something like that? And I came across this paper by Riley Carney and Ryan.

[00:28:51]

So I think Harvard researchers were you may have seen this where they. They substituted a random Eastern European group like Lithuanians in these racial resentment scales and substituted them in for African-Americans and found that they got the basically the same responses from people of both parties. So people who would say black people should work hard like everyone else, were just as likely to say Lithuanians should work hard like everyone else. And what they surmise from this is that alleged racial resentment scales are measuring is not properly seen.

[00:29:33]

It shouldn't be seen as racial resentment, but rather some kind of deep instinct for procedural fairness and meritocracy.

[00:29:42]

How you come across that or what do you what do you know?

[00:29:44]

I haven't I haven't read that specific paper, but I do think I would push back against the claim. And I would I do want to make clear, well, why am I am sympathetic to a lot of these academic critiques that exist, racial resentment as a concept. I do want to make clear that I think that, you know, I mean, I don't want to feel that same racism. I think racism clearly exists. I think some voters are more are more motivated by racially charged things than others.

[00:30:19]

And I do think that racial resentment, even if it doesn't on an individual level, capture everything I think does is related in a lot of ways. I'd like to talk, but I think that, you know, the paper you just brought up does really highlight some of the how a lot of this stuff works. I think it's I think that racial resentment is a lot less cartoonishly evil. It's it's more subtle and complicated. And I think that a lot of liberals will give it credit for.

[00:30:46]

And so the one thing you could say in response to your paper is that without any context, it's totally true that a lot of a lot of conservatives will care a lot about procedural fairness for some random Lithuanian group, but. There's a reason that conservatives and actually care a lot about procedural fairness relative to liberals, and I won't I won't ascribe that entirely. But I think if you step back and you look at where these what areas of the country are high racial resentment in which areas are not, I don't think it's a coincidence that the white people with the highest levels of racial resentment live in the Deep South.

[00:31:23]

Or if you map out racial resentment, I don't think it's surprising that it's areas where Trump did the most well, did the best in the primary, or that on a geographic level, it's correlated with places where people make more. I think clearly racist search terms and so when they're Googling racial slurs and stuff like that. So I do I do think that this is a construct that's real. I do agree that it's it's more subtle than people give it credit for.

[00:31:52]

But I do think that racial resentment as a concept, this is important, I guess, at a high level. I think it shouldn't be surprising in a lot of ways. If you have a situation where you have a candidate who campaigned on saying a bunch of racially inflammatory things early in his campaign and then got a bunch of people who seemed to score highly on racial resentment, questions to swing toward him. I think that. You know, racism has something to do with it, even if the word racist is hard to define and racism is hard to define.

[00:32:24]

I don't wanna get into a semantic fight, but I do think I don't think it's true that all of these working class white people in southeastern within northern Wisconsin trended against Democrats because they were inspired by Donald Trump's devotion to fairness, procedural fairness. I don't think that's right. So, I mean, I guess this would be a good time to segway into the what seems to be, you know, given everything you said and I don't want to dwell on it for too long, but it is surprising that Donald Trump has gained a certain amount of traction in the Hispanic and black electorate.

[00:33:07]

If it is the case that his appeal is based on racism or partly based on racism, it's at minimum curious why why that's the case. There has to be some way to that that should either diminish the notion that racism is his appeal or else we need some more complicated theory that accounts for why his share of the people of color is increasing. So what is your view, given what you just said? Can you first describe what we know, what is turning out to be true about how much his share of the minority vote is increasing and how you square that circle?

[00:33:50]

Absolutely. It's a great question. I think if you just to go through the history, historically, the trend for really the last 40 years is that every single cycle white people had been getting more conservative relative to the country overall voting Republican at higher rates while non-white people were getting more Democratic. That happened consistently every single election from nineteen eighty eight to twenty twelve and twenty sixteen. It was actually the first break from this trend. It was, I think, one of the underappreciated aspects of the election at the time.

[00:34:24]

I think a lot of people wrote it off as Obama not being on the ballot anymore, but Donald Trump did, I think something like two, two to three percent that are among African-Americans. And how well he did with Hispanics is something that people still argue about. But my impression, from what I can tell, is that at least in battleground states, he did slightly better with Hispanics than Romney. And I think that that just stopping there, not for the next four years, I think that says something very interesting that when about what racial resentment is and how people how people perceive it.

[00:34:56]

When Donald Trump dialed up the racial resentment, what happened was that a bunch of college educated white people with low levels of racial resentment moved to Democrats and non college educated white people with high levels of racial resentment moved to Republicans. The non-white vote didn't move very much or swung a little bit against Trump. And I think that's something that liberals didn't expect. I think a lot of liberals were really hoping that kind of openly xenophobic language would cause increases in Hispanic turnout.

[00:35:26]

And I think it's clear that didn't happen at least, and definitely not increases in support. So when you flash forward to twenty eighteen, you you see this racial polarization continue. White people relative to twenty sixteen swung something like five or six percent toward Democrats and non-white people as a group swung I think about a point against them, a point or two against Democrats.

[00:35:48]

And this was like a meaningful thing. Stacey Abrams would be a governor if she did as well with African-Americans as Hillary Clinton did. And we lost the Florida Senate race because we improved ground with white white voters, particularly rural white voters, but then lost ground with Hispanic and black voters and ended up losing a very close race because of it. So that was pretty important. And then now we flash forward to twenty twenty in twenty twenty. It seems like, again, the African-American vote has shifted toward Republicans, though it's unclear exactly how much, but probably something on the order of one to two percent.

[00:36:23]

While the Hispanic vote has swung tremendously toward Republicans, polls did kind of see maybe a third of that coming. I think going into into it, some people were expecting a three or four or five percent shift against toward Republicans. But so far, county level results really indicate that there was a much larger swing. There's a county in Texas that where I think it's about a county that voted for Democrats literally as long as it had existed for over one hundred over one hundred years.

[00:36:55]

Clinton won it by 30 points and it now went to Trump. If you there are just these incredible swings in most concentrated areas where there seems to be a lot of Hispanics. And it is one of the major reasons we lost Florida was incredibly poor performance in Miami-Dade. And so there's still a lot we don't know about the nature of this Hispanic divide. It's something that we're going to have to wait until we get more precinct results for and voter file results for.

[00:37:20]

But it does seem like in pretty broad swaths of the country, whether it's South Florida or South Texas or Houston or large swaths of the northeast in Hispanic neighborhoods in Massachusetts, that there's been a pretty broad double digit decline. It very easily could end up being as large as 14 percent. And that that's seismic. It's it's one of the largest shifts in racial. Voting that's really happened in decades in one cycle. I don't know, I think that's super interesting.

[00:37:49]

And then at the same time, white voters trying to draw Democrats. So in a lot of ways, Donald Trump has ended this multi decade trend of racial polarization and kind of reversed it. And I think that's something that I personally did not see coming going into this. But I think it's clear at this point that it's happening. Yeah, nor did I. So there was some I want to come back to that in a moment, but there was some initial polling result.

[00:38:15]

I think that right the day after the election that suggested that the LGBT vote and the Muslim vote had also swung towards Trump a few percentage points or perhaps more. Do we know more about that now? Did that turn out to be robust? You know, it's not something I had looked into. So I don't I don't want to comment too much. I think it's definitely true. People talk about the Muslim, but usually with regard to most Muslims in the US are Asian, actually South Asian.

[00:38:48]

There's a lot of evidence that, like Vietnamese Americans, swung pretty had a pretty substantial swing toward Trump. Obviously, Trump campaigned in the South Asian community, though, usually more with Hindus and Muslims. So, you know, I'm not sure, but I wouldn't find it surprising either way. So I want to come back to this notion of racial resentment in light of the the minority vote trending towards Trump because, you know, intuition as well as personal experience would tell me that.

[00:39:20]

Suggest to me if you were to give a quote unquote, racial resentment scale to the median. Ghanaian or Nigerian immigrant or Central American immigrant or Vietnamese immigrant or Pakistani immigrant, that they would you would see results very similar to, quote unquote, racially resentful white. Noncollege educated working class. And this is not to say how conditional on that being true. It seems to me there are a few things to observe. One is that, as you said, there's there's a never ending semantic debate over what exactly something like racism means, which is not useful ground here.

[00:40:16]

But then there's there's the fact that it seems when I read about racial resentment in the media, it seems like it's framed in such a way that this is something this is a uniquely a white, if not a sin than embarrassment, rather than a multiracial human phenomenon that might not usefully be moralized. It might make sense to come up with some less morally charged language or language that could not easily be interpreted as you're a racist for thinking that this group should play by the same rules or so on and so forth.

[00:41:01]

And it seems like we're less tempted to do that in cases where minorities have those same opinions. And I can see how a white person might feel. Well, the fingers unfairly being pointed at me when my opinions or my instincts are widely are shared by a lot of minority voters as well. So what do you make of that? Yeah, I mean, I do want to stress that it is definitely true, I think it's definitely true that a substantial faction, non-white voters hold possibly racial racial resentful language, though, is to be clear, like a lot on average, non-white voters, lower levels of racial resentment.

[00:41:48]

If you go and ask, which I think is unsurprising, I think if you did it a different way you asked African-Americans about their feelings toward Muslims, it's quite possible that they might be similar to working class white people or Hispanics.

[00:42:01]

I know. Yes. Like towards immigration, you you often find nearly identical levels of anti-immigrant sentiment among blacks and whites, for instance.

[00:42:12]

Yeah, polarization cuts against that. But I think within party, that's 100 percent true. And I grew up in Miami, one of the most dangerous places in the world. And I'm fully aware of the fact that white people don't have a monopoly on ethnic chauvinism. In some ways, it's definitely not true at all. And I think it is true that a substantial fraction of non-white people do agree with certain aspects of this worldview. What I think is really interesting is that historically, historically among Democrats, African-Americans had lower levels of racial liberalism is a different a different but related views on these things and white Democrats.

[00:42:54]

And what's really interesting is that after 2014 that changed. Yglesias Matthew Yglesias talks about the Great Awakening, that there was this really massive, unprecedented increase in racial liberalism among white Democrats. And at this point, white Democrats have higher levels of racial liberalism under most measures and black Democrats and definitely even Hispanic Democrats. And that's really interesting. I think that once you live in a world, once you have a scale where white liberals are scoring higher than nonwhites, then it does raise some questions that the limits of what what this stuff means and what the implications are for messaging.

[00:43:35]

And I think that when you talk about why it is that non-white voters just go back to that for a second, trended towards them, I want to be clear. I, I really mostly mostly don't know. But I do think there are some really interesting facts about the world. One thing we asked people, I think sometime in twenty, seventeen or twenty eighteen was we asked we asked voters, do you think that Donald Trump was racist? It's something that was pretty interesting is that when we look when we talk to black Trump voters.

[00:44:06]

Something like 30 to 40 percent of them actually did think that Donald Trump was racist, and I think that GetSet won, a lot of non-white people don't see the world as a. Crusade or jihad against racism in the same way that that white liberals do, that a lot of a lot of non-white people have had to live and work with racist white people their entire life. And the fact that there are racist white people in charge is less of a big deal on a Day-To-Day basis than a lot of white people might think.

[00:44:43]

And I think it does show that racial liberalism has kind of become a cultural marker among whites. And I think that or play it a proxy for this like within class or within culture, divide between kind of higher economic whites and lower socioeconomic whites. And so I think, you know, as higher socioeconomic status, educated white people have entered the Democratic Party, not a lower education. The whites have left the party. They have kind of become a larger share.

[00:45:15]

And I think they've really taken over a lot of the party's messaging, a lot of the party's branding. And I think that culturally, to get it, what you said with racial resentment, working class black people and working class Hispanic people and working class white people culturally have a lot more in common across a wide range of issues than they do with college educated white people. And so in a lot of ways, I think it should be unsurprising that as.

[00:45:41]

As American politics has turned into this war between high education whites and low education whites, that this has caused a little bit of racial polarization, I don't think that's so.

[00:45:54]

I guess this is a good, good time to go to this notion of demography is destiny, which is a phrase that has been thrown around, probably a phrase that basically means as the country, as the browning of America occurs, where the share of the population becomes less white and more Hispanic and Asian and African-American, that the Democratic Party will inevitably benefit from this trend and the Republican Party will inevitably suffer. So this is a phrase that's been repeated quite a bit in the past five to 10 years and celebrated on the left and feared on the right.

[00:46:38]

So can you explain this? This notion never really persuaded me, but because there seemed to just assume a static population, it assumes that the average black voter and the average white voters is going to have the same beliefs today as they will in 30 years, which seems crazy given the history of population change. But what do we know about that notion now, given the 20 20 result? Oh, I really like to tell the story of Florida in 2000, Florida was basically tied, obviously came up to a recount, and since then, population of Florida has doubled its non-white.

[00:47:21]

The non-white share of the electorate has doubled. And yet Florida is actually 20 years later. Florida is more Republican than it was 20 years ago. I think at any point in 2004, it would have been really easy to say, oh, we lost older voters, we won younger voters. And so in 15 years, we're going to win all these elections. And it's 18 years later. And clearly that has clearly, clearly there's a dialectic here.

[00:47:46]

There are these countervailing forces to talk about who practice what those forces have ended up being, even though younger people tended to be Democrats. At the same time, older people of twenty, twenty, twenty four Republicans. And actually, as the real shock and the reason why we didn't win Florida this time was that non-white voters trended Republican over the same period. So I think it really does go to show there is in political science, people talk about this concept that the median voter theorem, this idea that there are these countervailing forces and that party should expect to get about 50 percent of the vote over time.

[00:48:26]

And I think that's that's right. And I think the other point is even there are a lot of demographic trends that should help Democrats. In some ways, people are getting married much later than they used to. Birth rates are lower than they used to be. People are more are more secular. They're more educated. And these are all things now that really became Democrats. But at the same time, parties have shifted in response to those trends, like the Democratic Party is substantially below where it was 20 years ago.

[00:48:56]

And maybe if the Democratic Party hadn't left and we were still running people like Al Gore, who would be better. But maybe, you know, maybe then we would have had more of a chance for the parties to do these things more than people realize.

[00:49:10]

Yeah. So the same I've had the same skepticism or skepticism for the same reason of the notion that simply getting rid of the Electoral College, which I think might be a good idea, would inevitably benefit Democrats as well in the medium or long run, because I can't predict how under a new set of incentives, politicians and media would react and how voters would react to the knowledge that their vote now counts no matter where they live. So to me, it's a complete mystery how the Electoral College would affect things.

[00:49:50]

Do you have more insight on on that question?

[00:49:53]

I mean, I think that there are very real deep structural biases that go way beyond the Electoral College. If you look at the Senate, for example, in twenty eighteen Democrats won by wavier, they won by eight points nationally and they lost two Senate seats. The basic issue with the Senate, this really shows up with a bunch of other places is that rural places are overrepresented at every single layer of our government. And so if you look at our legislature, our state legislative maps that maps know, Democrats really would need to need to win by enormous amounts in order to win the Electoral College is a little less fundamental.

[00:50:35]

It's kind of a fluke right now that they're that the Electoral College bias is so large, though it's got larger this time in twenty sixteen. And I think that will probably stay large for the rest of the decade. But there's nothing inherent. Its bias is a little bit more in the Electoral College actually benefited Democrats somewhat in 2012. But I do think when you look at these biases, it is it is a very a very hard it is a real problem.

[00:51:00]

I think it does very meaningfully influence how many elections you'd expect Democrats to win. I think it's obviously unfair and a lot of ways it works. Franchising people who live in cities are disproportionately white. But, you know, it's interesting because in theory, it should be possible for no matter what electoral system you have, even if you had a system that didn't allow that, only let people in cities vote. In theory, some of the political scientists tell you that no matter what your electoral system is or how unfair it is, it should be possible to chase median voter.

[00:51:33]

Move your policy so that you win these back. And I think the really interesting challenge is that I think now it's become very hard for parties to do that. I think these college educated white people, the American electoral system is really built so that college educated white people can't have a coalition with them. And it's really built against a lot of ways. And so as education polarization has increased, it's really disadvantaged Democrats. But these same voters, these same college educated voters have a wildly disparate.

[00:52:07]

Amount of influence within the Democratic Party. They donate at much higher rates. They volunteer at higher rates. They're more likely to run as candidates and obviously they mechanically make up the staff work and all of these organizations. And so getting them to have less power. So or it's very hard for a party that is really fundamentally run by highly educated people. To put it its brand, to try to appeal more to older rural life. And we're going to have to try.

[00:52:37]

But it's it's very difficult. You've talked in the past about the changing role of money in politics or the changing political valence of super PAC money in politics. Can you talk about that a little bit? Yeah, I think we we'd be clear from prison, I think getting money on politics would be a good idea. I think it's being it's worth being clear eyed about what the current situation is, which is that right now in twenty twenty, the Democrats outspent Republicans at almost every level of government, maybe not in state legislatures, but in the presidency, in the Senate, House.

[00:53:20]

We generally outspent opponents almost uniformly in all of our races. And that was that has two sources. One, the people who donate to the so-called small donors and tend to be overwhelmingly professionals and professionals as a group have trended tremendously toward the left. If you go to twenty twelve, I think small dollars is it or if you go back to twenty ten small donors as a group, I think gave a slight majority of money to Republicans, but now I don't know what the number was.

[00:53:53]

Twenty eighteen that number. It was more like seventy two percent. Super PACs started to exist in 2010. Those are unlimited spending, the super rich people and they used to be 70, 30 or 80, 20 Republicans. And twenty eighteen I think was the first year where super PACs spent more on Democrats than Republicans. The reason is that as education, polarization is increased. The effects have been largest with the highest educated, richest people. I'm sure that we still lose billionaires as a group or super, super rich people in the group, but they're much more democratic than they used to be.

[00:54:28]

And so structurally, we now live in a world. We now live in this world where it's not money and politics doesn't have an obvious partisan balance. Or if it does, it's one that helps Democrats. And we also live in this world where if you look at how primary elections go, left wing candidates like Bernie or Warren were able to outraise moderate candidates. And the reason for that is that the both of them have an electoral base of highly educated white people that donate a lot more than Biden's days of working class black people and working class white people.

[00:55:08]

And so I think this is creating a situation where I think the money in politics, to the extent to it, to the extent to which it's a factor, is moving the Democratic Party to the left and arguably making it harder for them to win elections as a result, because campaigns are chasing this median donor as opposed to a median voter. And that's that's causing a lot of it's probably causing problems.

[00:55:34]

OK, two more questions before I let you go. So one is about the the net effect of. Some are 20, 20 that will go down in history books. I'm speaking of the death of George Floyd in police custody, the protests that ensued, as well as the riots that ensued. I've heard two countervailing trends that both seem persuasive. One is that the initial surge in support for Black Lives Matter as the nation watched in horror as as George Floyd died beneath the knee of Derek Chauvin led to a huge upswing in support for and registered voters for the Democratic Party, which had a positive effect for Biden on November 3rd.

[00:56:27]

On the other hand, there is a trend of the nation watching in horror as cities burn. And being driven towards the right as a result. So what do we know about the net effect, the racial politics of summer? Twenty twenty on the electorate. That's a great question, and it's something that in some ways is hard to know, but I think if you look at the if you look at the public polling trends, I think you look at them properly.

[00:56:57]

I think that you could tell a story that there really only two things that happen in this race. This race was really incredibly stable. And the only two changes you really see were, one, a surge in support after the Lafayette Park incident where Donald Trump ordered the National Guard to clear out and tear gas protesters in front of the church.

[00:57:21]

I think I really I can say I can really pinpoint that. I think if you look day by day, it's pretty clear that that was the event. The polling go up. And then I think you can tell a story that toward the end of August complicate it a little bit by the presence of the RNC around the same time, but that some of the unrest in Milwaukee led to a decline, that basically it was spread over like a two to three week period, but basically completely undid the one to two percent surge after Lafayette.

[00:57:52]

Now, that said, I think polling is very hard in retrospect. I think given that the polls seem to have been overestimating, but I think there's a case you could make that maybe liberals started answering the phone more after the Lafayette Park incident or that or that they answered less after the after the unrest. So I don't know how much of those trends are real, though. I suspect the answer is probably most or at least some. But I think that those were really the only things that seemed to move polling.

[00:58:22]

It was a little bit after the first debate, but really everything was incredibly stable other than those two things. And I think that tells you that these events were very powerful in a lot of ways. They were the only things that matter in the entire arguably more than voting. So that's that's which econometric play seems to not have any relationship at all with friends. So I think that's really interesting. And I think it gets to how nonviolent. Well, it's just that there's a lot of power in organizing, but also a lot of risk and not having sufficient discipline.

[00:59:02]

But like, if you look at the horse race, if you look at polling, you know, there's basically this story that immediately after the start of the protests, there were these very large increases in various measures of some measures of racial liberalism, but also a lot of measures about BLM favorability or attitudes toward the police. And unfortunately, when people started calling it again at the end of the summer, a lot of those gains that evaporated. And so it does show, though, I think not not all the way down to where it was before.

[00:59:35]

So it does show that one nonviolent protest would be extremely effective. There's a lot of great evidence in that paper. But but also that property destruction and all of that can be incredibly electorally damaging. And there's a lot of a lot of risk.

[00:59:53]

Yeah, OK, so so last question. The question is about how to evaluate. Trump, what the success of Trump, the surprising success not only in twenty sixteen, but to some extent in twenty twenty, says about the future of Trump ism. There is a tendency to want and I notice have noticed this in myself as well, a tendency to want Trump to be a fluke and a desire to just get a John McCain or Mitt Romney type Republican next time, just from the interest of a person who thinks Trump's personality has made him a dangerous and unfit president.

[01:00:42]

So the question is given, given the fact that Biden did not crush Trump as was expected and as I would hope should have been easy given Trump's unusual flaws. It may just be worth re imagining his flaws as not liabilities, but as an asset. I have to imagine that. Republican presidential hopefuls, the lesson they're taking away from the past four years is not everything Trump did is a mistake. And we should go back to the Republican status quo.

[01:01:22]

It has to be something more like Trump's presidency for Republicans represents a half success, where all the ways in which he departed from the norms of the Republican Party and did everything possible to piss off the half the country, a lot of that actually might point a direction in a direction of success for the Republican Party. And they might just see three or four of Trump's character flaws as and his mismanagement of covid as the things to avoid. So the mold for them might be Trump minus the narcissism and perhaps minus the five worst scandals.

[01:02:13]

So basically, my question is, do you expect that we are going to be dealing with Trump ism out of the Republican Party as as a direction or with a reversion to the status quo? You know, I would say predictions about the future, but I do think it's important to recognize Trump ism is I think that people really focus on a lot of there's a lot of psychoanalysis of Donald Trump that people do know. People really want a strong leader.

[01:02:44]

They really like he was on The Apprentice. But I think, you know, if you step back every single country in the West right now, whether it's France, Britain, Italy or Germany, you know, all of them have seen this phenomenon of anti immigrant populists coming out, promising not to talk a time about economic issues and then gathering and kind of having incoherent stances on economic issues and then gathering this base of working class people who used to vote for the left and others.

[01:03:19]

This has happened basically in almost every Western European country you have. And so there's something really fundamental here. And if you look at who the leaders of these people are, they're not all like Marine Le Pen. It's not like Trump. The head of it is like they're all pretty different. And so I think the personality aspects to why this coalition exists are are not essential. But what is kind of there is really culturally being cruder, trying to appeal to having these working class appeals, talking about being pretty stridently anti-immigrant, trying to dial up racial resentment, whatever that word means.

[01:04:00]

And the only difference is that it wasn't successful in these other countries because they have a different electoral system. But our electoral system, we are very heavily overweight, the views of non college educated white people. And so I suspect the GOP was rewarded from the Trump era by large, almost semi-permanent majorities at every single level, state legislative, congressional, Senate, and also a world, an electoral college world where they can win with forty, forty eight and a half percent of the vote.

[01:04:37]

I think that's going to be very tempting. And so I think that's the problem is that due to the nature of our electoral system, the Republican Party really has a very strong incentive to continue along this model. Something I like to do just to talk about numbers. Is Hillary Clinton got Barack Obama in 2012, got fifty two point zero percent of the Tea Party vote and Hillary Clinton got fifty one point one percent. And in any other country, you know, if you went from fifty two to fifty one point one, it would be fine.

[01:05:08]

It wouldn't really mean anything. But what happened at the same time is it the bias of the Electoral College went from being half a point biased toward Democrats to being like three or four points biased toward Republicans. And the reason that happened is that all of these swing states had a lot of non college whites who know the strategy really appeals to. And that's true along racial resentment, authoritarianism, all of these things heavily populated with class. And and so, you know, that that's why he was able to win was because this coalition has so much more power.

[01:05:38]

It wasn't about actually bringing in more voters. It was about changing the nature of the coalition. And that's going to be super tempting as a political party. I would if I I wish that my political party could win with forty eight percent of the vote, everything would be very different. And so I think it's going to be very hard for them to change. And that that's even just from a strategic perspective. I think if you look at the nuts and bolts, Republican primary voters are going to try very hard to finish.

[01:06:06]

I personally don't think Trump is going anywhere, which I think is scary. And I think that the only long term solution to fighting terrorism is fight. The reason why it works by trying to get noncollege white people to split their votes more. We basically it's awkward to say I just want to go back way back to we're talking about racial resentment is that there are a lot of white people with relatively relatively high levels of racial resentment, and there are so many of them that it's impossible to win elections without getting a certain fraction of the votes.

[01:06:41]

And realistically, Trump ism is going to exist as long as 80 percent of those people are voting for Republicans. If we don't return to the twenty twelve Obama era status quo of 60 percent of those people voting for Democrats, then the electoral, then we're just going to keep having the same problem again and again and again in this time. Luckily, Donald Trump is super narcissistic and personally, I'm unlikable in a lot of ways. And so we were able to eke out a win, but next time we won't be so lucky.

[01:07:12]

And so it's really imperative that we do what we can to win these voters back. Yeah, it occurs to me when he said these things, racial resentment and authoritarianism and all of these other things are correlated with class. There's a problem, a universal problem of social bubbles. But there's a problem when the class that is labeling all these things, deciding what to call this is the higher class. Right.

[01:07:41]

It's up to people like me and or whomever who went to the Ivy League to to label what that scale means and to moralize it. And however way it makes sense to me based on my upbringing in a sunny liberal Montclair, New Jersey. And that may look differently to other people. And of course, it does. At some level, it's impossible to correct for one social bubble perfectly. But I think this if there is one sort of shooting from the hip here, but if there was one piece of advice for Democrats to be Trump ism in the future, it's to be very aware of how one's existence in an elite bubble influences the messages that one finds compelling in a way that puts one out of touch with what the people you have to win over and who are winnable are going to find compelling.

[01:08:38]

Yeah, I think that's that's exactly right. It's just a very hard thing to do. That's something that we've seen in our tests. We actually my old firm, we did we tested a lot of ads. One of the things that was really clear was the more people in the office liked the ad, the worse it did a lot of the time. But my favorite example is in the 2016 campaign, there was this thing called the Mehrzad, the one of the most Sherzad of the campaign of the campaign.

[01:09:03]

And there's this little girl and she's in front of the mirror. And then Donald Trump is saying all these really horrible things and the girl starts crying. That's like an ad. And it turned out that that actually cost us votes in large swaths of the Midwest. You know, we spent millions of dollars on it, but there really were I personally am a liberal. I like the ads, but there were a lot of working class white people who are like, you know, Donald Trump is talking to me about real issues that matter.

[01:09:30]

And you are trying to guilt to be about some politically incorrect jokes that Donald Trump made. And I don't care. And so even though, you know, even there's a real issue, which is that the people who run, you know, run Democratic campaigns and I'm in this bucket of have really very different values, then a lot of swing voters. And so that's that's why I think it's super important to be clear headed and do your best to have good measurement and also to try to have more representation in a lot of concrete ways to people right now.

[01:10:06]

See it within the Democratic Party. Highly educated people have an enormous amount of power and that's highly educated. People always want everything that is that is actually how the world works. But everything we can do to dial that down will make us better at winning elections.

[01:10:21]

So on that note, David, thank you so much for your time. And can you point people in the direction of the work you're doing now to your Twitter profile or website if you have one? Yeah, I am.

[01:10:32]

David Shaw on Twitter. Last name is our David Shaw. And I have I have a lot of fun, professional stuff coming up and I'll be announcing there. Thanks so much for having me on.

[01:10:45]

Absolutely. It was a pleasure. All right. Take care. All right.