The emerging consensus is that Donald Trump won the 2016 election by divisively appealing to voters’ views on race and immigration. But Justin Grimmer and Will Marble find that Trump gained votes over Romney among low-education white voters, largely independents and moderates, who had centrist views on race and immigration. In contrast, John Sides and Lynn Vavreck find that the 2016 campaign activated voters’ attitudes on race, immigration, and identity, making them more important in driving voter decisions. A large all-star panel reviews the central debate over the 2016 election and its implications for the 2020 campaign ahead.

Studies: “Who Put Trump in the White House?” and Identity Crisis
Interviews: Justin Grimmer and Will Marble, Stanford University; John Sides, Vanderbilt University; Lynn Vavreck, UCLA


Grossmann: This week on the science of politics: did racial views drive Trump voters? For the Niskanen Center, I’m Matt Grossmann. The emerging consensus is that Donald Trump won the 2016 election by divisively appealing to voters’ views on race and immigration, but a new paper argues that he instead gained vote share and turnout among those with moderate views.

As we gear up for 2020, what are the lessons of the 2016 election? Today, I am joined by an all star panel in a special larger, conversational edition of the podcast. John Sides of Vanderbilt and Lynn Vavreck of UCLA are two coauthors, with Michael Tesler, of Identity Crisis, the book that set the standard for explaining 2016.

They find that the campaign activated voters views on race, immigration, and identity, making them more important in driving voter decisions. Justin Grimmer and Will Marble of Stanford University are the authors of the new paper, “Who Put Trump in the White House?”

They find that Trump gained over Romney among low education white voters, largely independents and moderates, who had centrist views on race and immigration. Lynn, John, Justin, and Will all join me to discuss the similarities and reconcile the differences.

Justin, can you start by explaining your basic approach to analyzing voting blocks and how it differs from the prior analyses?

Grimmer: Sure. Yeah, so the paper comes from a sort of fundamental view about how we understand who contributes to a candidate’s victory. And Will and my view is that we want to understand where the votes are coming from. And this differs from a big literature which has studied activation.

And the broad idea with activation, and Lynn and John know much more about this than me and can correct me if I’m wrong, the idea with activation is what we’re studying is the relationship between the vote choice of individuals who turn out and attitudes that those individuals hold.

So for example, we might look at a correlation between something like the attitudes towards various minority groups among white working class voters and their vote choice. And using that, we might get a sense of how particular attitudes correlate with a choice in one election or another.

And this is important and it actually turns out to be an input to a lot of what we’re doing in the paper that Will and I are doing, our paper. But a key difference is that activation actually is only weekly informative or sometimes not informative at all for where the votes come from for a candidate.

To get a sense about why that would be the case, consider the following example. Suppose we have a voting block that’s constituted of just one individual, so we have one person and this person is an ardent Trump supporter.

It turns out that the relationship between this person’s vote choice in their characteristic, say this ardent Trump supporter, is going to be extremely strong. But if we wanted to have an understanding of where Trump’s votes come from, this one individual’s choice is not going to matter all that much for understanding where the Trump votes come from.

So instead, what we also have to include, in addition to the activation, we also have to include things like the turnout rates of various groups and crucially their composition in the electorate. What proportion of the voting eligible population holds particular characteristics?

The reason this is important is because we show in our paper that I think surprisingly Trump receives fewer votes in a smaller share of the voting eligible population among people with high levels of racial resentment than Romney. And also, we show that Trump’s increase in votes, net votes, that is places where he either loses by fewer votes or wins by more votes.

The biggest gains for him come among individuals, white individuals, who are moderate on immigration rather than individuals who are more opposed or conservative on immigration. And so taken together by taking this broader view of where support comes from, we reach a little bit of a different conclusion, just a little bit.

A different conclusion about the basis of support for Trump, trump’s victory seems to be coming from moderates and from low socioeconomic status whites.

Grossmann:  So, Will, is it fair to say the headline is that racial attitudes and immigration have been overemphasized in our analysis of the 2016 election and if so, what do you think has been under-emphasized?

Marble: Yeah, I think that’s right to some extent. It’s clearly true that Trump tapped into and campaigned on these racial and anti-immigrant sentiments. So in one sense, it’s true that a big story in the 2016 election was where are these racist attitudes and sentiments that Trump tried to tap into?

But on the other hand, when we look at actually where the votes came from, I think it is fair to say that that his support didn’t come disproportionately from people who hold very racially conservative or anti-immigrant attitudes as Justin mentioned.

When we apply our approach to various co-variates that we could look at in the population, what we see under-emphasized is the extent to which Trump won over moderates and independents. So if you look at how he did among people who identify as Republicans, he actually did worse than Romney did in the sense of getting fewer votes than Romney did among people who self identify as Republicans.

And instead, he actually picked up quite a few votes among people who as independents. And you see a similar pattern if you look at a standard ideology measure, he did worse among people who identify as conservatives and better among people who identify as moderates and even among people who identify as liberals.

So I think the under-emphasized story is probably the extent to which Trump picked up people who we might think of as kind of swing voters in order to improve over Romney’s performance in 2012.

Grossmann: So, Lynn, that sounds different than the findings in your book Identity Crisis, but how much is really similar and different to what you found?

Vavreck: I think that it sounds different, but the first thing to keep in mind is that any time you’re talking about evaluating an election outcome, and it can be a big election outcome like 2008 or a really narrow election outcome like in 2016, you’re almost always at the presidential level talking about the voters who swing using the term that Will just used.

So I think the first thing that’s helpful is to think about the difference between the swing voters and what we would think of as the base or the core. So that base and core, they’re given that name because they’re always voting for that party. They’re reliable partisans.

So when we’re talking about elections flipping, we’re talking either about composition of the electorate that Justin mentioned earlier, how that’s changing. If one party totally stayed home, that would be a story. That usually doesn’t happen, so we’re usually talking about compositional changes here and there for both parties.

Or we’re talking about a set of people who are routine voters who are moving from election to election, party to party. So that’s a super interesting story, has often been hard to tell because we haven’t had great data on turnout over time and it’s very hard to get that consistent in terms of geographic constituencies year to year.

So these guys have done a great job to try to isolate those problems. And it’s hard to know cause we don’t know individual level vote choice even when we have that turnout data, but it’s a super important question. But it’s a different question than the one that we’re talking about in Identity Crisis, which is given that you have people who are likely to turn out or have turned out, how do they think about their vote?

Or sort of what’s the recipe for their vote choice and how does that compare in 2016, for example, to what that recipe was in 2012 or what it was in 2008? So we’re looking at the factors that are important in elections over time. And what we’re saying is that regardless of whether you’re or your core person, in 2016, a very important ingredient in that recipe is your racial attitudes.

And it doesn’t have to be on the conservative side, right? It could be very liberal racial attitudes leading you to vote one way or the other, so it works in both directions. But that ingredient is playing a larger role in ’16 than it did in ’12 or in ’08. And then we’re saying that other factors are not seeing those kinds of changes in their impact.

And so it’s a different story, it’s tied more toward what the candidates are talking about and how they frame the election and sort of what the election … how people are thinking about it, how that manifests in their vote choice, but both of these pieces come together I think to tell the story of 2016.

Grossmann: John, one other way to put them together might be that a voters might not have extreme attitudes on immigration and racial attitudes, but they still might be important or more important to their vote choice in in 2016. How do you see it?

Sides: I would say three things. I think the first is that if we’re thinking about why it is that Trump won fewer votes from people with more conservative attitudes on race or … part of the story we have to focus on is who was changing their views on those issues?

And what we show in our book and while a lot of surveys have shown is that the people who shifted their views, the reason there are fewer people with conservative views on immigration and race, is because Democrats changed their views. And that happened to some extent in 2016 itself, it certainly has happened since 2016.

So to look just at racial attitudes and say that Trump won fewer votes than Romney, I think we have to ask a different question about the politics that produced that change. I don’t think the voters that have driven that change are in any sense swing voters, they are Democratic voters who have shifted their views on those issues.

So they really were never up for grabs and so to say that Romney won more of those votes than Trump did isn’t going to necessarily tell us a lot about the swing voters themselves. The second thing has to do with the who his swing voters were. So this is where our data is particularly advantageous because we have panel data.

And we’ve talked to the same respondents in 2012 and 2016 and we know who the swing voters are because they’ve told us that they are by telling us that they voted for Obama and then telling us four years later that they voted for Trump. And so I think what we are able to show in our analysis is the factors that are distinctively correlated with those shifts are factors that have to do with race and immigration more so than other factors.

And that those distinctive relationships are different than in some other recent elections, so we really think that they are unique to 2016. So in that sense, I don’t think it’s quite accurate to say that the swing voters were moderate per se.

Vavreck:  Well, John, can I just ask you a question? So it could be that that increasingly conservative racial attitudes are related to making this switch, but that really could even be at the middle to low end of the overall quote unquote like racial attitudes scale.

So if seven is a conservative and one is a liberal, people could be moving from two to three or three to four and still be on the liberal side like they don’t … you know what I’m saying? Like it doesn’t necessarily have to be a move that’s happening at the conservative end of the scale.

Grossmann: Well, I would say extremism does not necessarily equate to salience or importance in the voting decision. People could be somewhere in the middle, but have conservative to moderate views on those issues.

Vavreck: I mean, that would be a way that both of these stories are right.

Sides: To me, the word extremism is sort of irrelevant. I think for our story about the election, we don’t use that word. It’s not really an interesting way of thinking about voter’s attitudes in this regard. Let me say what I think is true, if you want to understand why Trump wins, part of the story is about turnout, agreed.

Part of the story is about these shifts among voters. When you look at white Obama voters in 2012, many of them have conservative attitudes. I’m not going to say extreme, I don’t know what that means in this context, they have conservative attitudes on race. I mean, I can pull you data from our book.

Almost half of white Obama voters as of 2012 said that they didn’t think that blacks have gotten less than they deserve, almost half said that blacks should work their way up without quote unquote special favors, almost half favor the death penalty, about a third of them had views on immigration that seemed fairly out of line with Obama’s views and certainly with the Democratic party’s views as of 2016.

They said that illegal immigrants were a drain on society, they wanted to make it harder to immigrate to United States. So it’s not about being extreme, it’s just about having views on racial issues and immigration issues that were out of step with where Obama was and where the Democratic party was.

Why that becomes relevant is then once those issues are activated, once they become more strongly associated with how you vote because that’s the issues that Trump and Clinton talked about so much in 2016, it pulls these voters from Obama to Trump. And that’s what helps him win a substantial number of votes in crucial battleground States, particularly in the Midwest.

And that’s what gives him an electoral college victory. So again, you don’t have to have a conversation about extremism to tell that story. You just need to know what these voters believe relative to what the candidates believe. And you can see why they’re mismatched sort of behavior in 2012 became much better aligned with their racial attitudes in 2016.

Grimmer: So the first, we’re using extreme in a sort of statistical, not in a normative sense. So when we’re saying extreme, I mean people at the tail ends of the distribution. So very much in line with with what John’s saying, I don’t want to cast judgment on particular views that people are having as being sort of like out of the norm or something like that.

So I just wanted to add for that as a point of clarification. Then to think about this issue of flippers, people who switch from Obama to Trump, I think it’s a really interesting issue. But I think it’s also the case in that it’s important to contextualize this flipping behavior.

It turns out that Trump actually performs just about as well as Romney did with former Obama voters. And another way to put that is that, based on CCS data, the number of votes that Romney obtained from people who report voting for Obama in 2008 is about the same number of votes that Trump obtained of people voting for Obama in 2012.

One of the big differences though is that Clinton was less able to retain Obama support in ’16 relative to ’12 and that was one of the big things that we see in our data. And so that sort of flipping I think provides some important context for understanding the appeal that Trump perhaps had to these to the white former Obama voters.

Sides: So I said this to Justin on email, so I’ll just say it again publicly for anybody listening. That fact is true, but the politics of why those switches occurred between ’08 and ’12 and between ’12 and ’16 are very different. And I think that’s why it’s important for understanding what we think went on in these two elections.

Obama lost votes in ’12 relative to ’08 because the conditions in the country were not as favorable to the Democratic party, right? We had an incumbent Republican presiding over a financial crisis and a recession, Obama wins handily in 2008. By the time you get to 2012, right, the conditions are favorable to Obama’s reelection, but not nearly as favorable as they were.

And so of course, he loses some votes, relative to what they got in 2008, he loses votes among all groups of people, of all demographics. He loses votes in almost every state in the country relative to where he was. Yeah, it’s a uniform swing in many respects or close to uniform swing.

When you get from ’12 to ’16, it may be the case, right, that the number of shifters is roughly equivalent, but now the shifting is not uniform. It’s not uniform across demographic groups because better educated whites shift to the Democratic party, white with less formal education shift to the Republican party, et cetera.

It’s not uniform across states, Clinton does better than Obama in Texas and Georgia and other states that have lower percentages of whites with without a college degree. Clinton does worse than Obama in all these states which have lots of whites without a college degree, so that would be Midwest and Iowa and New Hampshire and a bunch of other States.

So I think, again, I’m not disagreeing with anything that Justin just said except to say that I think if we want to think about not just the quantification of the number of shifters, we want to think about the reasons why the shifting happened, what that tells us about the election. I think the shift from ’08 to ’12 is very different in character, even if it’s equivalent in number, than it was between 12 and 16

Marble: Just a brief follow up on that, and I want to reiterate that there is a big chunk of agreement between what we’re saying, one of the things that I think is really difficult when thinking about compositional shifts, and John characterize this, this shift in racial attitudes primarily the movement coming from Democrats, is that once we have this sort of shift happening, it becomes really difficult to make an across election comparison on the activation regressions. And so this is actually a point nicely made by Andrew Engelhart in a QJPS paper. The idea is that once we have folks who have different partisan identification sorting across the spectrum, it’s hard to disentangle a particular attitude being a stand-in for your group membership and an attitude exerting some causal effect on your vote choice. And so it becomes really difficult to look at 16 and 12 and say, “Oh well the correlation is stronger and therefore the racial attitudes mattered more for people who reach this… who turn out to vote and are in the voting booth.” And so I think those sort of comparisons just on a statistical level are very difficult to make apples to apples.

Vavreck: So Justin, so two things. I think that when I think about things like this, I always want to say out loud, to anyone who might be listening, “Panel data, panel data, panel data.” Because I think that that’s a small step toward getting us better information about the things that you just talked about. But if I’m hearing you correctly, and I just want to ask you to clarify on this, even panel data, what you’re saying is that even panel data isn’t enough to be able to make that last part of the story about sort of telling the story of the election, because what you’re saying is that obviously, compositionally there are problems if a whole different set of people are turning out in addition to the small set that we have panel data on and we’ve missed that even with our panel data. But you’re saying that, as people’s attitudes become just a proxy for their party membership, we lose the ability to really tell whether it’s activation or just more people coming into that group. But the panel data does kind of help us there, right?

Grimmer: Yeah. So I think I said this in email as well, so I think the panel data gets us very far. It turns out to be much harder than I had anticipated to figure out who is an Obama Trump voter. It turns out just to be a pretty vexing question. So the panel data helps quite a bit. The thing that I think can make it a little difficult, particularly when we think about running regressions, is that if we have folks who are on the very far end of the spectrum that we’ve created, particularly extreme views, again, not normative, but just empirically on the end, and if that’s a rare group with very clear voting behavior, they prefer one candidate over the other, then it becomes this signal characteristic and it becomes very hard to understand the causal effect of that vis-a-vis just an indication of their general preference in politics.

Grossmann: Well, I want to bring you in. We’ve been talking mostly about differences across white voters, but there was a big turnout story related to minority voters comparing Obama and Clinton, and you also document some that Clinton got less share among African Americans and didn’t get the anticipated bounce among Latino voters. So is that a big part of the story, and how does that relate to our discussion?

Marble: So I think the answer is yes and no. Yes in the sense that we and other scholars have documented that there is a drop in turnout, especially among African Americans that ended up… meaning that Trump saw more votes from blacks than Romney did in 2012 so he actually increased his vote share among minority voters. That said, the increase overall is almost predominantly due to white voters.

Throughout our paper we focus primarily on white voters, given the emphasis on racial and immigration attitudes that we’ve been discussing and the fact that they constitute the majority of the voting population and the majority of the shift between 2012 and 2016. So I think in that sense, even if turnout had been significantly higher among minority groups, it still probably would not have been enough to offset the gains that Trump made among white voters.

Of course, election outcomes are overdetermined and there’s a lot of things… Every vote counts so losing fewer votes is the same as gaining more votes in terms of actually winning elections. So in that sense, of course it matters, but I think the primary story that we document is mostly movement among white voters. And as we’ve mentioned, what we think of as white, moderate voters, especially those who are low socioeconomic status.

Sides: We absolutely think that changes in the composition of the electorate made a difference in 2016. Although I agree with Will that it’s hard to say that, without those changes, Clinton would have won. Those counter-factuals are really hard. I think people focus a lot on the drop in black turnout relative to 2012, and I think understanding a little bit more about why that happened is useful.

I think our argument is essentially it was almost inevitable that blacks would not turn out as strongly for a white democratic nominee as they did for Barack Obama. And one of the things that we thought was particularly interesting when we looked at surveys, and looked at the attitudes of African Americans, was they were of course favorable to Clinton overall and they were integral to her victory in the primary, but they weren’t as favorable as they were toward Obama.

For example, in 2016 surveys, 76% of African Americans had a strongly favorable view of Barack Obama. Only 34% had a strongly favorable view of Clinton. Similarly, there were surveys early in Obama’s term, and then in 2016 that asked whether Obama cared about African Americans. How much? And whether Clinton cared about African Americans. Back in early 2009, 78% of African Americans said that Obama cared a lot about African Americans, but by the fall of 2016 only 42% said that Clinton cared a lot about the views of African Americans. So I think part of what the challenge that Clinton was facing, and part of the reason why the composition of the electorate shifted, was just that it was difficult for her to generate the same level of strong enthusiasm from African-Americans that Obama did.

Grossmann: Lynn, one of the oddities of both of your papers is that you find that, or the paper in the book, is that you find that immigration and racial attitudes are mostly moving leftward over this period that we have identified as this identity crisis that moved people to Trump. Obviously, Trump ran a campaign full of racism and some symbols, but what was happening on the Clinton side? Is it possible that Hillary Clinton just ran further to the left of the public on these issues? or heightened them more than the public was ready to move, in an attempt or an expectation that it would both increase her vote among Republicans turned off by those views, Trump’s views, and increased her minority share?

Vavreck: Yeah. I think this is a really complicated question. I’ll try to just say a few things about the way I think about what was going on with messaging on racial attitudes. I don’t think it was the the Clinton campaigns intention to run a campaign that was focused on the characteristics and traits of their opponent and then pushing back on the things that the opponent was saying about race and gender. I think that that’s a byproduct of who the opponent ended up being and that it ended up being Donald Trump, and that he ended up talking about those things and just sort of based on his past and his history. All candidates come with constraints. His are a lifetime of being in the public eye and saying all sorts of inflammatory things about women and non-whites. And so that is one of the areas that the Clinton campaign decided they could differentiate themselves on.

Vavreck: A lot of I think what happened in terms of messaging there was just a consequence of who the Republican nominee ended up being. Now given that that’s where you end up… the way I look at this is, and we sat down for Identity Crisis and did a content analysis of the advertising messaging from both campaigns, and the Clinton campaigns advertisements were largely about Donald Trump. There were some ads about her policy positions, particularly on the economy and providing strength for the middle class and things like this.

And if you remember back to her announcement video, it was all about that. She was walking down the street meeting people who were starting new jobs and trying to buy new houses and she said, “I want to start a new job too, and live in a new house.” And they moved away from that pretty fast. I think that the original tagline was Everyday Americans. Or something like that, but the move away from that was not strategic because of, “Oh, the electrodes moving to the left, we should move left. We should try to lead on that. We should try to pull them.” It was in response to Trump and Trump’s rhetoric and his constraints. And so I think that is the most simple explanation and I wouldn’t really attribute any more to it than that

Sides: To add to what Lynn said, my best sense of the Clinton campaign’s theory of the electorate was that they needed to reconstitute and mobilize the Obama Coalition, and that coalition was understood in a particular way as being comprised of racial and ethnic minorities and younger voters and women, and it was viewed as a relatively liberal group on these and other issues. And so I think that her campaign was built to some extent to achieve that. Part of the reason why I think that that was already somewhat baked in was she came out even relatively early on, on issues like police violence against African Americans and she talked about it and in pretty unabashedly progressive terms, she talked about implicit bias and systemic racism. So I think that that may reflect some of the activist pressures within the party that preceded the 2016 election.

But it also reflects to some extent the changing views of democratic voters even before 2016, so democratic voters over all were becoming more liberal on racial issues and on immigration prior to 2016. We show that in our book.

To some extent the democratic coalition has shifted. I think the activist pressure had pushed her in that direction. And then if there was ever a question of whether she would not talk as much about those issues once the general election campaign got underway, then I think Lynn is exactly right. She did so very much in reaction to Trump himself and that I think helps solidify a campaign that had a lot more focus on race and immigration than even 2012.

Vavreck: Let me just jump back in for one second because it’s this conversation is when I really, really like to think about the things that Justin and Will are showing in this paper. So when you think about the rhetoric of these two campaigns, and then you take the finding that’s in their paper, like, “Oh, the swing voters are these people who are not in the tails of the distribution on these things. They’re kind of people who are in the middle and are swinging.” That just makes so much sense.

And both of these things can be true. The election can be structured. And the thing that Justin just said earlier, that if people are sorting into the parties based on these attitudes, and say they’re sorting into the parties… John just said this is starting in 04 and then 08 and then 12, this sorting is happening, and by the time you get to 2016 you’ve got people well sorted so that these attitudes about race and immigration, or any kind of identity inflected thing, whatever you want to think about, can be structuring the basis of the party vote in these elections.

But then the swing just makes sense. By definition, are going to be people who are not so steeped in those attitudes. They maybe can’t figure out where they belong. And you can see that data in Identity Crisis. We have a really long table, white Obama voters, half of them have more conservative racial attitudes in 2012, and a third have conservative immigration attitudes. And so I think the story of the paper in our book really comes together nicely, especially when you think about the long arc of the movement on public opinion on these issues.

Marble: If I could just jump in here, I think one interesting thing that has come out of this is if you think back to the 2008 and 12 campaign… and you’re right, Lynn and John write this in Identity Crisis, is that Obama ran on a populist economic message, especially in 2008 in response to the financial crisis. Even given these shifts on these more social issues, it seems to me, and Lynn and John, I’d be interested if you agree with this, that maybe Clinton kind of made an error in figuring out how to reconstitute the Obama coalition, which he brought together in large part due to this populous economic message and not necessarily due to this messaging on immigration and race that Clinton was backed into due to the fact that she was running against Donald Trump.

Vavreck: I’m going to use the word victim here, but I don’t really mean victim, but they’re both victims of their circumstance. In 08 Obama’s able to frame the election in that way because of what’s going on in the world, and it’s the same thing that happens to Hillary Clinton in 2016. The frame of that election, she’s maybe reacting more to it than setting it, but Trump comes in and he’s setting that and she’s going to be a player in that. So to say that that was an error is… yes. Could she have made a different choice? Yes.

I like to think of it like: would any candidate, any presidential candidate that we can think of in the last 50 years have ended up in the same position as her in 2016, running against Donald Trump, and made a different choice? Maybe one or two. I think most of them would have looked at public opinion, would have looked at the composition of the electorate, the state of play, and probably said, “My best play is to push back on this.” Obviously it ends up looking like a mistake because it didn’t work, but I don’t think that it’s fair to say that Obama read the tea leaves quote unquote better than she did.

Grossmann: Now Justin, the other prominent explanation for 2016 has been about economic insecurity, but neither of these studies really find much evidence for that. We did have these huge amplified divides among white voters, by education and by urban rural residents. So what, if not identity or, or economics, what explains those?

Grimmer: Yeah. This is a great question, and one that I’d love for everyone else to chime in on as well. So the first thing about economic insecurity, one of the striking things that comes out when we do our analysis, is that things are just a lot better in 2016. Zip codes have much lower utilization of unemployment insurance, folks are back to work, folks are reporting that their household incomes are increasing. So in the sorts of measures that tend to proliferate in the literature about economic insecurity, it just turns out in 2016 people are much more secure than they were in 2012. I think that jives with our intuitive understanding of what was happening.

And so when we think about the pool of people that Trump had to appeal to that were in some measure of some economically insecure position at the moment it was smaller than the economic insecure folks in 2012. Now there could be some more global thing happening, something where folks may feel like they’re not using unemployment insurance because they are of a particular age where it’s hard for them to get back to work and so they’ve sort of been backed into being retired. Perhaps that’s one group of people who could be helping to explain this.

And we have some circumstantial evidence for this. I don’t think either Will nor I want to hang our hats on it completely, but we do find some of the biggest increases in support for Trump among white voters to occur among individuals who are in this 50 to 59 year old range, which is a thing that’s been written about a lot where folks who are laid off during the financial crisis and they have trouble getting back to work because they’re right before retirement so employers aren’t really excited about hiring them.

And we also find some evidence that folks who report that their employment status is disabled are more supportive of Trump. And again, when we think about the sort of disability status as someone who’s chronically unable to go back to work, but perhaps had previously been working. But I do not have the one-line, “This is the thing that explains the education divide,” and I’d love to hear what other folks have to explain it.

Sides: We took a pretty substantial crack at this and we’re more than willing to let the data speak, but we had no preconception about the relative role of economic insecurity versus racial factors, but we just couldn’t find much evidence that economic insecurity was a crucial factor here. Justin’s absolutely right that if you just go to the macro-economic level in 2016 it’s much better than 2012 and that’s true among all people. It’s not just the wealthy. Incomes are increasing among every income quintile. They’re increasing among every education group. Consumer confidence is up among every group. It’s at levels consistent with Reagan’s reelection in 1984.

People could not get it out of their heads that there was this just gloom and doom all across the nation. But, you ask people, “Are your financial circumstances better than they were a year ago?” And people increasingly say, “Yes”. “What about business conditions in the country?” “Yes, they’re better too”. “What do you think will be true in a year from now?” “I think they’ll be doing well”. “Is it a good time to make major purchases like a washing machine?” “Yes”. That’s what the longstanding consumer confidence measure was telling us.

When we went and looked at the election campaign itself and the role of these different factors, one of the nice things about the survey data is that we have these interesting questions which are not just the generic, “How are you doing financially?” But they asked you very specifically, “Are you worried about losing your job? Are you worried about missing a housing payment? Are you worried about missing a healthcare payment?” And I think if I understand what commentators mean by economic insecurity or economic anxiety, it would seem that these are exactly the indicators that you would want to ask.

And then basically, we found that these had very modest, if any relationship, to people’s choices in 2016. They had almost no relationship. They had a relationship that was no stronger than it was in 2012. And when we looked after the election in a different survey project, Rod [inaudible 00:40:18] and I, for the voter study group, we found that the people who express the most economic insecurity were Clinton voters, not Trump voters because unsurprisingly, economic insecurity is more prevalent among racial minorities than it is among whites.

Grossmann: Lynn, what about the possible role of just simple spatial voting? You all find that Trump was perceived as more moderate than the Republican party, and if not economic insecurity, that seemed to be about economic policy positioning. So is it possible that these were just independent and moderate voters who didn’t see a Trump as extreme as Romney for, say, on the economic policy issues?

Vavreck: Yeah, you kind of set me up here when you asked me have we underestimated the basic role of spatial voting? I feel like some days I’m the last political scientist in the world who’s willing to say, “Yes, every day. Almost always.” And I think about this a lot and I think that what Justin was saying earlier, that in an era where the parties are so well sorted, you take a snapshot of the electorate and you say, “Oh, look at the emotion. Look at the way that people are connected to their party.” It’s tribal, it’s emotional, it’s psychological. And you don’t know that. That masks, if that’s all you’re looking at, there are ways to know this, but when the parties are so well sorted on issues, a different person could take a snapshot of that same electorate, see all those same things, see that all Democrats believe X, Y and Z, all Republicans believe A, B, and C and say it’s spatial voting. People are voting on issue positions and policy positions.

And so I’ve been spending a lot of time trying to figure out how we can sort that out given the landscape the way it is. And I think it’s a really important question because as social scientists interested in politics, we’re responsible for getting this right. So yes, I think is the answer your question. Everybody knows that 90% of Democrats typically vote for the democratic candidate in the presidential election. The same thing for Republicans. We saw a little bit of a shifting away from that, very small, in 2016. This election that was supposed to be so different. And so we have to pay more attention to why people are sorting into these parties, how much that has to do with policy. And I think that right now people are just very focused. I see this in the dissertations my students want to write, in the job market, candidates, the papers that people give when they interview for jobs, the people who are getting interviews, people are very focused on psychological attachment group identity. And I think that there’s a lot of work here to be done on the fidelity between policy preferences and candidates positions.

Sides: This is one of the things that Lynn and I actually don’t fully agree on. As she well knows, I tend to under own special voting almost every day and every election. I don’t think that voters have the information that they need to vote in the sense that we mean by the term spatial voting, where you’re processing candidates on a left right spectrum and choosing the candidate closest to you. But I do agree with Lynn that it’s very difficult to separate some of these things. And part of what makes it hard is that there are a lot of policies or issues that you could imagine in an abstracted ideological way, but there are also lots of issues which are obviously intimately tied to people’s own social identities and their perception of social groups.

So there’s a sense in which immigration is a policy issue. We can have debates about immigration as what policy we want the government to implement. But of course, people’s views about immigration policy depend a lot on their views of immigrants and the groups that they conceive of as being prevalent among immigrants, like Latinos. So I don’t want to say that it’s social identity all the way down, but there are lots of policy issues in which is going to be difficult for social scientists to sort of put them in a bucket that’s like, “Oh, this is spatial voting versus this is social identity,” when it’s entirely possible that the reason why voters voted for a candidate who had a position on an issue closer to their position on an issue, like special voting would suggest, is because their social identities led then to have that position on that issue. And then when the candidate articulated that view, they were willing to hop on board that candidate’s bandwagon.

Grossmann: Well, one of the factors we haven’t talked about much is gender, but obviously we had an election between a male and a female candidates with very big gender overtones and we had some movement among men to the Republicans and a lot more division on traditional gender roles and the kind of sexism or modern sexism indices. What do you make of that? Is that another case where we’ve looked at a regression after the fact and found something that that isn’t real? Or would an analysis of that show similar things? That these were people with, forgive me for saying, moderately sexist views that were attracted to Trump in this election and if that’s the case, how should we interpret that?

Marble: Yeah, that’s a great question. So I’ll start with if we look at the breakdown in where the votes come from by gender among white voters. Basically we find that there’s not a disproportionate increase among men for voting. Trump didn’t do better among men than he did with women relative to Romney. Basically he increases the number of votes among white men and white women by the same amount. Which I think is, when we’re looking among white women especially, kind of in line with historical voting patterns where white women tend to vote for Republicans, and it turns out 2016 wasn’t actually all that different on that dimension.

So to the question of how views on gender roles and sexism play into it, we don’t look at this too in depth in the paper. There’s other work out there, including an identity crisis, that shows that people who have, or men especially, who have more sexist views are more likely to vote for Trump than they were to vote for Romney, conditional on people actually turning out to vote.

And one thing we can look at with our data, or one thing that we have looked at, I should say, is a feeling thermometer measure on the ANES about how warmly people feel towards feminists as a group. And there we see a pattern that is somewhat consistent with the other findings in the paper, namely that Trump gained the most number of votes over Romney among people who feel kind of moderate towards feminists. People who rate them between 50 and 75 on the feeling thermometer, a hundred point scale. And then Trump lost votes relative to Romney among people who feel very warmly towards feminists, namely those who rate them at 75 to a hundred. So I think there’s probably more we could do in looking at other measures of sexism. But based on that, it seems like the pattern might be somewhat similar in that it’s not people who are extremely antagonistic towards feminists or extremely sexist who are giving Trump the most votes and the most increase in the number of votes he got compared to Romney.

Grossmann: So Justin, what should we expect in 2020? How much should we expect the divisions that materialized in 2016 to carry over? And any predictions about turnout versus vote choice changes in impacting the outcome?

Grimmer: Okay, so the, the one prediction that I hope will be right about 2020 is that we get to the third in the series, the gamble identity crisis. I hope we get a 2020 version. That’s the first thing I want to say. Second… Okay, so what do we know about what’s gone on in politics recently? So Trump has been a boon for political engagement. So he’s got a lot of people to pay attention in ways that I think have been difficult for them to pay attention before. And so if I had to take a gamble, I think turnout is going to be way up in 2020, both on the Republican side where I think there’s a lot of folks who are very excited about the things that the Trump administration has been doing, and with folks who are more sympathetic to Democrats who are concerned about the actions of the Trump administration. So would guess that we’ll see a pickup in turnout in a way that we didn’t see in 2016.

As far as divisions go, or hardened divisions, I think there’s going to be some really interesting questions about what happens with the ongoing demographic shift in American and how that manifests in this election. So in the run up to 2016, there had been this long discussion that there’s changing demography in America. White voters are becoming a smaller share of the voting eligible population and because then these nonwhite voters are more likely to vote for Democrats, this gives Democrats this huge advantage. This overstates in two ways, and it’ll be interesting to see what happens in 2020.

One, it overstates that nonwhite voters may, while they are in many instances unlikely to vote for Republicans, they may decide not to vote at all, which our analysis shows can be quite beneficial to Republican candidates. And second, as we see the white share of the voting eligible population decrease, it could be the case that white voters become a more reliable block for the Republican party. And so if those things persist, I think that we could see hardened divisions and we could see perhaps even bigger gap among the racial groups that we saw in 2020. And given the current patterns, I would expect to continue to see the education divide to manifest in 2020.

Sides: Yeah, to echo I think and agree with what Justin said, there’s no evidence that the divisions that seem more hardened now than they were four, eight, 10, 12 years ago. There’s no evidence that those are weakened at al. All of the survey work that’s happened since the 2016 election is showing Democrats and Republicans further apart on issues like race and immigration than they were even when the election itself took place. And again, a lot of that’s because of this shift among Democrats.

I also was wondering to see whether democratic leaders would shift their own thinking about these issues because of the interpretation of the election that Clinton messed up by being too liberal or by focusing on these issues too much as opposed to on the economy. And I don’t see any evidence of that either. If anything, the democratic party shifted further left on these issues because they’re reacting so strongly to Trump’s agenda. So Clinton took a lot of liberal views on immigration in 2016 but she didn’t say we should abolish ICE, or we should decriminalize border crossings, which are things that democratic presidential candidates have said from the debate stage.

Sides: So what we would want, if you thought we would get different alignments of voters or you would imagine that different factors came to the fore, I think that would happen if we had a recession or something that really brought the economy back to the front of the agenda. But for the moment it seems like the way that the coalitions of the parties look in terms of the voters as well as the strategies and decisions of the leaders. It looks like it’s kind of continuing to harden the same divisions on identity inflected issues that we documented in 2016.

Marble: If we have time, just jumping on one more thing there that John just touched on, I’m interested to see how well the democratic party will come together after this primary. One of the most interesting tables I thought an identity crisis showed that only 79% of Bernie Sanders primary voters ended up voting for Clinton in the general election, which I think is super interesting given that we have what seems to be shaping up to be a pretty divisive primary as well with kind of the moderate Joe Biden, Pete Buttigieg lane facing off against the more liberal Bernie Sanders and Elizabeth Warren. So I’m very interested to see whether there’ll be able to kind of come together and deliver the votes in the general election despite this divisive primary.

Grossmann: Lynn, you talked about Trump putting a Democrats in a box, or Hillary Clinton in a box, last time. How do you see that the current democratic candidates either under learning or overlearning the lessons of 2016 and is it going to matter who the nominee is and how they position themselves? Or are we seeing another election that’s going to be all about Trump?

Vavreck: Well, to tie this back to the original topic about the swing and who are the swing and what characterizes them, I think it does matter who the democratic nominee will be because that is going to affect the group of people that Will and Justin are exactly interested in talking about in this paper. And it’s going to be a different group of people if the nominee is Sanders than if the nominee is Pete Buttigieg. I’m just willing to bet money on that. There’ll be a different group of people. I wish we could see both scenarios play out and these guys could write their same paper for both of those scenarios.

I think that if Trump is going to be central to this, well you have to find some daylight between you and your opponent. So he will definitely play a role, but you can see how it’s going to play out. If you think about the four people who are front runners right now. Warren, she’s talking about corruption and she’s going to portray Trump as part of the problem. Corrupt, corrupt, corrupt. Biden, he’s talking about experience, gravitas, temperament. That’s probably the closest to what Hillary Clinton did in 2016, so you can think about what that might mean.

Buttigieg, I think, is probably a new choice for a new generation, time for a change, sort of evocative of that Kennedy sixties new frontier changing of the decade. There’s a lot of similarities between what Kennedy did in ’60 and what Buttigieg is doing in terms of messaging and framing the election. Trump is implicated in that, but much less so than in the Biden and Warren stories. And then Sanders is probably the candidate with the least Trump centrality to the messaging. It just is what it is. It’s ever so what always been, the country’s broken, we need to be thinking in a different way. And I think those are your four front runners right now going into Iowa. So we’ll see.

Grossmann: There’s a lot more to learn. The Science of Politics is available biweekly from the Niskanen Center. I’m your host, Matt Grossman. Thanks to Justin Grimmer, Will Marble, John Sydes and Lynn Beverick for joining me. Please check out Identity Crisis and “Who Put Trump in the White House?” And then listen in next time.