Voters with college degrees are increasingly supporting Democrats, with Republicans now doing better among those without college—a big reversal in recent decades. Joshua Zingher finds that college-educated Americans are more liberal on social issues and that more educated Americans are moving furthest toward Democrats when surrounded by other educated people. White voters are flipping fastest by education but the trends are present across the electorate. Will Marble finds that white college graduates are now more liberal across economic, social, racial, and foreign policy issues. Less educated white voters have increased the importance they place on non-economic issues, polarizing the electorate on these issues.
Matt Grossmann: What explains the diploma divide? This week on the science of Politics for the Niskanen Center, I’m Matt Grossmann.
Voters with college degrees are increasingly supporting Democrats, with Republicans now doing better among those without college, a big reversal in recent decades. The pattern is even starker in geographic terms, with more educated areas growing very democratic. What explains this flip in voter partisanship and voting? This week, I talked to Joshua Zingher of Old Dominion about his political research quarterly article, Diploma Divide. He finds that college educated Americans are more liberal on social issues and that more educated Americans are moving furthest toward Democrats when surrounded by other educated people. White voters are flipping fastest by education, but the trends are present across the electorate.
I also talk to Will Marble of the University of Pennsylvania about his new working paper, What Explains Educational Polarization Among White Voters? He finds that white college graduates are now more liberal across economic, social, racial, and foreign policy issues. Meanwhile, less educated white voters have increased the importance they place on non-economic issues reaching parody with more educated voters. Dave Hopkins and I are finishing a book on the education divide, and we think it’s among the most important developments in contemporary politics. So it’s critical to understand when, where, and how it develops. I think you’ll enjoy the latest research on this trend, starting with Zingher.
So tell us about your new paper on the Diploma Divide. What did you find?
Joshua Zingher: What I found in this paper about the diploma divide is that our level of educational attainment says a lot more about how we vote than it did in the past. And to be specific about that, people with college degrees and or graduate degrees have become much more democratic over the last couple of decades, but in particularly the last six to eight years. And people without a college degree have become somewhat more Republican, although the change is less pronounced. And we see evidence of this both at the individual level if we look at how individual level voters behave, and we also see evidence of this at the aggregate level. So if you look at counties and or states, you see the aggregate level of educational attainment in these places strongly predicts voting patterns in whatever geographic unit that you look at.
So people are well aware of racial divides in voting behavior. But I think many ways in the coming decades we’ll have a second, if not even more important, divide in the electorate and that’s over educational attainment.
Matt Grossmann: You also look at some issue area differences based on education and you find some differences kind of across the board, but they do seem to be some relatively large differences on what we consider social issues or cultural issues. Do you see that as the driving trend?
Joshua Zingher: There’s certainly a big gap in the culture war issues between people with college degrees and not. So we think talk about things like abortion or we talk about things like immigration. Gay marriage is less of a hot-button issue, but we still see a divide when it comes to gay rights along educational lines. We do see some polarization between college graduates and non-graduates on economic issues. And there’s some evidence that college graduates have moved left on economic issues over the past decade. You also see divides along educational lines, a lot of racial issues too. So if you think of debates we’ve seen over Black Lives Matter, over police violence towards minorities, all these types of things. What we think about the policy responses to these social problems is strongly conditioned by our education too. So, the divide between degree havers and non is present across a whole bunch of different policy domains.
Matt Grossmann: So you mentioned you also find this at the geographic level. So one way of interpreting that is that we just have a lot of degree holders in an area, and so we’re just aggregating this individual level effect. But I noticed that even before any kind of interaction there remains just an effect of living in a more college educated area. So how do you interpret that? Is this partly a cultural context effect of living in a college town versus a rural area? Is education having an effect at kind of the aggregate level as well as the individual level, or are there just two ways of looking at the same?
Joshua Zingher: Yeah, I’d say the evidence in my paper is certainly more suggestive than conclusive of any particular mechanism here. But I strongly suspect that there’s a pretty strong contextual effect educational attainment based on the composition of where somebody lives. Jonathan Rodden in his book Why Cities Lose makes this point that college graduates or once pretty evenly distributed across the country. And now we’ve seen that pattern change a lot where college graduates have clustered in a lot of the big metropolitan areas. And so we see a big educational split between urban and rural areas in a way that we didn’t before. When you get a lot of people living together with similar educational backgrounds I think it reinforces whatever those particular set of cultural values that those groups hold. And so I suspect there is a strong contextual effect here based on where you live. People that live in highly educated areas adopt a different set of cultural values regardless of educational levels than people that live in less educated or more rural areas.
So, I think there’s pretty strong evidence for contextual effect. You probably can see this in the aggregate level voting data too. But yeah, people with college degrees that live in heavily educated areas are systematically more democratic and more liberal than people without a college degree, people with a college degree that live in other types of places.
Matt Grossmann: So how should we interpret that interaction? Is there something kind of amplifying the actual effect of education or is it partly due to educated people who’ve chosen to live in educated areas? How do you see that cross level?
Joshua Zingher: Yeah, that’s a good question and I’m not sure I have a conclusive answer for that. My strong suspicion is we can think of cultural aggregation effect where you get similar types of people living in a similar type of area where people start to take on more uniform attitudes to match the other people around them, becomes self-reinforcing. And we’ve seen big increase in geographic polarization too on top of everything where different types of places in the United States, like all over rural areas looking increasingly like each other when it comes to voting and urban areas look increasingly like each other when it comes to voting. And people think about divide along red state and blue state lines.
If you want to think [inaudible 00:07:56] red state, look at Louisville, Kentucky. Kentucky not a place we typically think of as a bastion of liberalism, Louisville is extremely liberal. Or if you can think about some places, is in upstate New York. New York’s a blue state, but rural upstate New York is quite Republican. And so you see these splits between urban and rural areas and more educated and less educated areas regardless of the types of states that these places are in. Large metropolitan areas look pretty similar in a lot of ways regardless of the state they find themselves in. Louisville, Kentucky looks a lot more like Chicago in terms of the voting patterns than it does the rest of Kentucky. And you can say the same thing about downstate Illinois looks a lot more like rural Kentucky than it does Chicago. And I think these kind of contextual effects matter a lot. Measuring those in a conclusive way I think is tough.
Matt Grossmann: You also explore several potential causal pathways between education and voting, including party identification first to just show that it does seem to be changing patterns in both party identification and voting, but you also do a little bit to show that it might be having an effect through party identification as well as voting. In the long-term patterns it sometimes looks like the voting changes first. You have the Trump voters and anti-Trump voters changing, and then by the time the next election they’ve kind of gone all the way over to that side. How do you see that relationship and do we know anything about whether education is operating through changing people’s partisanship or maybe whether even there’s a reverse effect where it changes voting first?
Joshua Zingher: Yeah. This is, I think, an important question. We’ve gone almost 10 minutes and this is the first time Trump’s name has come up in this whole discussion. And I think all of the patterns that we see with education and shifting voting patterns along educational lines, that foundation was in place before Trump came along, but Trump kicked it into overdrive. And you see a lot of places like I’m in Virginia, and at one point in time not that long ago, the northern Virginia, the DC suburbs anchored the Republican Party in the state. Many of the states like Western and rural areas were very democratic leaning. Those patterns have flipped on their heads. Where now Henrico County in suburban Richmond, and then places like Loudoun County, Alexandria and the Northern Virginia suburbs heavily highly educated, the voting patterns flipped before the registration by and large. Which suggests people might’ve made some different choices at the ballot box and then later on updated their party identification.
That might work somewhat differently than we typically think about causal models of voting behavior and political science, going back to the American voter where you think social attachments come first, partisanship follows, and then ultimately voting decisions follow from there. I think the reality is probably messier than that. And certainly the evidence I uncovered, it showed some reciprocal effects like education shapes partisanship. We see a direct effect of education on voting behavior, but you also see an indirect effect because it shifts partisanship too. And so both the longstanding decision and the election specific decision, you see some evidence of change there. But saying the chicken or the egg question with partisanship and voting behavior, I think you can tell a story that from each perspective.
I think at least my personal take on the evidence is that we’ve seen voting decisions change before partisanship. But now many of the places that defected from the Republicans away from Trump in 2016 are now just pretty solidly democratic. If you think about the example, the district in the Atlanta suburbs in Cobb County that Jon Ossoff ran for in the special election in 2017 and narrowly lost, that district flipped in 2018 and is now just was not particularly strongly contested in the next couple of elections and is pretty much just a safe democratic district now. And think Cobb County, once upon a time, that was a Newt Gingrich’s base of power. So you think about how much that place has changed, and we see a lot of places around the country that look like Cobb County.
Matt Grossmann: You conduct your analyses among all voters and then just among white voters. And we know that the attention in the media has been on these changes among white voters, and you do find some stronger relationships among white voters. But it does seem like there’s at least some evidence that educational polarization may be beginning, or at least that there’s been a flip away from higher educated voters leaning Republican also among non-white populations. So to what extent might we be in the middle of those relationships also appearing among non-white voters? And I guess why and why not would we expect those patterns to be different among non-white voters versus white voters?
Joshua Zingher: Oh, I mean, if you look at public opinion and if you look at the actual issues, the divide on educational lines extends to all groups. Where it gets messier or where it gets less clear is actual partisanship and voting behavior. We definitely see evidence of a diploma among Latinos and Asian-Americans, where more highly educated Latinos and Asian-Americans are more democratic than less. African-American partisanship and voting behavior is much more uniform than either of those two aforementioned groups. So it’s harder to make a case for a diploma divide there, at least in terms of voting behavior, simply because there’s a lot less variation.
African-American public opinion, on the other hand, is way more complex than the voting behavior suggests, and you can certainly see evidence of a diploma divide in opinion. It just hasn’t manifested at the ballot box in the same way. But, when you think about the growth in Asian American and Latino populations over time and political diversity within these groups, yeah, I think this is going to be a big source of political diversity among Latinos, among Asian Americans and probably any other group you can think of.
Matt Grossmann: But it does seem like this is important for sort of how interpreting the divide, there’s at least one story that says that this is mainly about racial issues among lower educated white voters. We might not expect the same relationships among racial attitudes for minority voters, but maybe we should. To what extent should we expect that the same things that are driving this divide among white voters.
Joshua Zingher: Yeah, so I think abortion is a great example of why we should expect this to matter across the board and not just for whites. If you look at public opinion about abortion, it’s not particularly racially polarized. There are some differences between ethnic and racial groups, but the differences aren’t particularly pronounced. Now, if you look at abortion attitudes across educational groups, all of a sudden we see pretty stark evidence of polarization where support for legal abortion goes up as a function of education, and consistently up. So if you look at high school versus some college, or just bachelor’s versus postgraduate degree, support for legal abortion increases at every step along the way there. And so that, I think, is evidence, I think plenty of people think abortion is an important political issue other than white people. And if that’s the pattern that we see across groups along educational lines, I think that’s pretty solid evidence that we should expect the diploma divide to extend to other groups besides whites.
Matt Grossmann: So we do have some new kinds of evidence on the effect of higher education that try to take advantage of causal inference strategies, or do before and after higher education and actually look for change attributable to higher education in short order. But on the other hand, we have, after 2016, people who are 40, 50 years out of college changing their views as a function of something about their college education, but maybe not the college experience directly itself. So what do you think about that? Should we expect there to be an effect that is apparent from the beginning to end of college, or is this about the networks, the communities, the professions, the class that we end up in?
Joshua Zingher: Yeah, I think I much strongly lean to that second perspective. I don’t think there is much about the actual educational process that I would shift views to dramatically to the left. I strongly suspect this is much more closely related to the types of networks you have, whether or not you stay in your hometown or not. I think one of the underappreciated facts of college, and at least in the majority of cases, it forces people to move out of their hometown and be around with a bunch of other people that they ordinarily would not be. And this experience, I think, is probably where that profound difference comes.
And maybe you could even think about this deeper level where you have some traits like openness to experience, people that are willing to leave and go to this new place might look systematically different than those who choose to stay behind. And so whether it’s education, like is there some book that people are reading in the classroom, or a set of books, or a set of views they’re exposed to, I think the evidence is much more powerful that this is about a network effect, this is about leaving home. This is about living around different groups of people than you would ordinarily be exposed to.
And certainly we do see differences in political attitudes across college majors, but there’s also self-selection that goes on with that, which makes the inference problem that much harder. English majors are more liberal than engineers, but people choose which majors they go into, they’re not randomly assigned, fortunately. So you start thinking about that set of issues, I’m much more strongly come down on that second camp.
Matt Grossmann: So another way of looking at it is counterfactuals, do the parties have anything that they can do? On the one hand, you mentioned that there was a lot of complaints about Hillary Clinton’s campaign that Joe Biden’s campaign tried to rectify, and yet where he gained was the same kinds of places that Hillary Clinton had gained. On the other, the fact that this is, in part, about party positions and movements suggests that there might be things that the parties could do, either moderate their opinions or change their relative emphasis. How do you see it?
Joshua Zingher: Yeah, one of the fascinating things about the 2020 Election was just, I think the thought behind nominating Joe Biden to post Trump was like, this is someone who grew up in working mainly working class areas, Scranton, Pennsylvania, and then in Delaware, and had strong connections dating back decades with organized labor. And this was the guy that was going to be able to speak a common language with the white working class people without college degrees, a place where the Democratic Party has really struggled over the last couple of decades, and this was going to be the path to victory.
Well, it turns out the path to victory was through places like Oakland County, Michigan, and I think Dane County, Wisconsin, where Madison is, and these places that are not what you think of traditionally working class at all, but very affluent, very highly educated suburban areas, and that begins to think… And I think a lot of things that Joe Biden has done as president, even if the perceptions don’t take, looking at the Infrastructure Act and the CHIPS Bill, and these types of things were very… It’s a very new deal Democrat, I think, approach in some ways where you focus on these big public works projects and building up domestic industry. He’s not perceived that way at all by anybody, and I’m not sure why that’s necessarily the case, maybe these party brands are hard to shake.
One of the things I think that makes Biden unique is, yes, he is very old, but he remembers a time when the Democratic Party stood for some very different things than it does now.
Matt Grossmann: We had an explosion of popular commentary and, to some extent, scholarship are really designed to explain the 2016 election, and we may not be entering our third election in a row, a presidential election in a row with Trump on the ballot. How should we assess the state of that research? Were we too focused on changes in one election that turn out to have been a longer set of patterns? Has political science successfully influenced the popular knowledge about this? Or did we learn something from the popular commentators that we weren’t paying enough attention to?
Joshua Zingher: So this is what I think people get wrong about Trump, and this is where I think a lot of the popular commentary gets wrong about Trump. Trump is viewed as this oracle, this working class white whisper, this idea of he has this special connection amongst working class whites that simply other candidates don’t. And I think that might be true to an extent, but I think what’s important to keep in mind is Trump did not noticeably improve the share amongst whites of the vote, amongst whites without college degree relative to Romney in 2012. Where the movement came from was whites with college degree leaving the Republican Party. And I think so much of the Trump commentary focused on the fact that, yes, he won the Electoral College in 2016, but don’t forget to include a giant caveat that he won with 46% of the vote, which is, you think lower than Mitt Romney’s performance in 2012, it’s lower than John Kerry’s performance in 2014, it’s lower than Al Gore in 2000.
It was a showing, you think about Trump having some sort of unique political power, I think is often the view. But in many ways, Trump and Trumpism is historically unpopular. And you think about an incumbent president not even cracking 46%, or not 47% of the vote, again, in a reelection campaign. Historically speaking, that’s a really bad performance. And then when you couple with that with what happened in 2018 and 2022, I think you have some pretty powerful evidence that there wasn’t any special magic there. It was an appeal to a narrow but intense slice of the electorate and that appeal has remained consistent. But barring some funny stuff in the Electoral College, it’s not enough to put together a nationwide majority coalition.
And so that’s one of the things I think that the coverage has gotten wrong about Trump. This wasn’t a movement that got 53% of the vote nationwide and won easily. This was, I think, in context, an Electoral College fluke in 2016, and then an incumbent president losing the popular vote by 7 million ballots in 2020. I think, in most cases, this is a pretty disastrous electoral record for political movement.
My personal comparison of Trump is this is William Jennings Bryan who happened to win one. We think about William Jennings Bryan being this democratic populist, agrarian populist, the Cross of Gold and bimetallism and all these issues. He never won anything, for the most part, at least in the presidential level. You could have a similar story with Trump, with the big exception that one time, 46% of the vote was enough.
Matt Grossmann: So what do you think we still don’t understand about this rising education divide and what should we be looking for in 2024?
Joshua Zingher: So what we don’t understand about the diploma divide? Here’s what people ignore about the diploma divide. Way more people have a college degree now than they did 20 years ago. And so you start thinking about, why is this shift so important? Well, almost 40% of adults have a college degree now. This used to be a very niche constituency, 10 to 15% of the adult population. Now it’s 40% and still climbing. So you talk about college graduates, how they behave, and when you consider college graduates turn out to vote at a rate that is much higher than the national average, even a bigger percentage of the electorate. And so you think about previewing 2024, I haven’t studied this in any type of rigorous fashion, but you think one consistent trend over the last couple of years is Democrats have really overperformed in special elections. And I think a big part of this is who is the type of people that show up in special elections?
Well, it’s the people who show up in any election from dogcatcher to city council to midterms to presidential elections, and this skews much higher levels of education than the people that vote in maybe a presidential election, or maybe they show up to a midterm. You want to think about a coalition of people that show up to vote frequently. This is college graduates. And if college graduates are predominantly siding with one party over the other, for many years I think this was the Republican Party’s strength, they had strength among college graduates and they turned them out every single election. And if those voters go a different way, I think that’s a huge part of the story. That’s a big portion of the population and a disproportionately large part of the electorate.
How that pertains to 2024, you start thinking about what’s the big drawback about college graduates? Well, they live in geographically clustered areas. Winning additional college votes in Boston or San Jose or some of these places doesn’t help you a bit in the electoral college. At least I saw from some data from L2, the voter registration tracking company, they’re saying non-college whites, they might be less, 44% of the population in shrinking. There’s still a majority in 23 states, including places like Pennsylvania and Wisconsin, Michigan, these places, where the Rust Belt states that are going to most likely swing the 2024 election one way or the other. So college graduates, they’re a large and growing segment of the population that’s increasingly democratic. They’re also not particularly strategically distributed across states, across congressional districts, this kinds of stuff, which blunts the impact.
Matt Grossmann: Zinger has tracked the key trends, but Will Marble sought to explain them with a model of issue voting that incorporates how much voters care about each issue area and their positions. He came away feeling there are different factors driving each side of the education divide.
So what are the main findings and takeaways from your new paper on education polarization?
Will Marble: I start by documenting what’s now become a familiar pattern of polarization by education levels among white voters over the past several decades, where we see that from the 1980s through around 2000, there were relatively small differences in voting patterns and presidential elections across educational lines, with college educated voters voting slightly more for Republicans. But then around 2000, you start seeing this divergence whereby college educated voters start trending towards Democrats and non-college voters start trending towards Republicans. And by 2008, it’s really clear in the time series that we’re witnessing a realignment, and it’s really only grown since then to the point where education is now one of the best predictors of vote choice.
So I ask whether there are particular issues that underlie this realignment, and so what I do in this paper is I compile dozens of policy questions from the ANES and the CCES to generate estimates of policy preferences that are comparable over time across several different issue areas. And then I develop a framework to interpret the correlations between individuals’ attitudes on these issues and their vote choice. And I come up with three primary findings. First, I find that attitudes are increasingly correlated across issues for both groups. But this is especially important for non-college educated voters where there was relatively little ideological constraint up until relatively recently. And now you see a very high correlation across issue areas so that there’s consistent polarization between college and non-college voters across these different issue areas.
Second, I find that the working class, those white voters without college degrees, sometimes I’ll use working-class as shorthand here, have long been more conservative on cultural issues, but we now also see that they’re more conservative on economic issues than college-educated voters as well. So we’ve seen that there’s an increasing liberalism on economic policy among voters, white voters with a college degree.
Finally, I find that the relative weight placed on cultural issues has increased for non-college voters. So for much of this time series, there was not a whole lot of correlation between vote choice and issue attitudes on cultural issues among non-college educated voters, which I interpret as the issue weight placed on those issues. But recently we see a convergence in issue weights so that we see that both college and non-college educated voters are basing their votes on cultural as well as economic issues. So in some sense, these conservative cultural attitudes have become more important for the vote choice of non-college educated voters. So these are the main findings that I present in the paper, and this provides a lens to interpret this growing educational polarization.
Matt Grossmann: So you say there’s been a flip in the relationship between education and voting, and then an increase in polarization. I know that we often look at this in terms of specific elections, although you’ve looked at it over a longer time period. So to what extent is there a sense in just the basic relationship and time series that there was some critical election and we do need to explain 2016 or 2000? Or is this just a long, slow pattern that doesn’t have as much to do with what the parties are doing in any given election?
Will Marble: So I would say that this is a trend that we see going back now about two decades, this realignment. It’s been accelerated certainly in recent elections, in the Trump elections especially. But I think to really understand this pattern, we need to go back in time and understand it as a broader realignment that spans elections. So I think that it’s not to say that what the parties are doing don’t matter, or that what individual candidates in elections do don’t matter, but I think that there are bigger trends that we need to understand to really put into proper perspective the realignment that we’re seeing. So I tend to think that election specific factors certainly are important, but these long-term trends indicate that it’s not just a one off critical election that’s explaining these patterns.
Matt Grossmann: So you look at these four different issue areas. So why don’t we go through each one of them and tell us what’s in it and what direction is the relationship between education and opinion going? And what’s the trend, if any, in the importance of those issues?
Will Marble: Sure. So these four issue areas that I group different policy questions into are based in part on previous research that groups issues into economic issues and moral or social values issues. I add to that issues of race, racial policy, racial resentment, things like that. Then the final category is foreign policy, which includes a hodgepodge of issues related to the United States place in the world, as well as some particular issues that cut between foreign and domestic policies, such as trade policy and immigration. So these are four broad issue areas that I think reflect a lot of the debate in contemporary politics. So let’s go through each one of them and think through what are the differences between college and non-college voters on each of them, both in terms of the attitudes they hold and the weight attached to each issue.
So first on economic issues, what we see is that for a long time span, college educated voters are more conservative on economic policy issues. These issues, the questions that go into this are primarily related to redistribution and spending by the government. What we see is that starting about 15 years ago, we start to see a convergence in the attitudes on economic policy so that we see college educated voters no longer being more conservative on these issues. In recent elections, they’re actually more liberal on these issues than working class voters. So we see this big shift in the orientation of the working class versus college educated voters on economic issues.
Then on these other issues, so moral social values, race and foreign policy, we see that college educated voters have long been more liberal on these issues. This doesn’t change a whole bunch over the time span that I study, which is 1984 to 2020. What we do see is that there seems to be an increasing importance placed on these issues for vote choice among non-college educated voters. So going back to the eighties and nineties, you see very little correlation between, say, non-college educated voters attitudes on cultural issues and their vote choice. In recent years, you see that the correlation between vote choice and issue attitudes on these issues is nearly identical between college and non-college voters.
So I develop a framework to think about how we should interpret these correlations, an issue voting framework. I show that when we look at the ratio of the coefficients, we can interpret this under this framework as the relative issue weight placed on these issues. So what we find is that there’s a convergence in issue weights, such that non-college and college educated voters are placing about the same amount of weight on both economic and cultural issues, which is really a big departure from past decades where cultural issues really only seem to matter very much for highly educated voters and seem to matter a lot less for non-college voters who basically vote to a much greater extent on economic issues.
Matt Grossmann: Then real quick on just foreign policy and racial issues, did those match the trends in cultural issues or did they have unique dynamics?
Will Marble: Sure. I’m using cultural issues as a shorthand for both racial issues and these moral values issues. You do see a convergence in issue rates in race and civil rights issues as well as in foreign policy. Foreign policy has a lot of variation across elections, especially going back to the nineties, but it really has stabilized in the last four elections or so to the point where both college and non-college voters placed the same amount of weight on this. I think to some extent this is due to the choice of including immigration as a foreign policy issue here, which has a whole set of unique dynamics. My approach here was to try to think procedurally about what policy domain different issues would fall into. But of course, there’s valence issues or valence that is attached to immigration that is not quite the same as debates about NAFTA or debates about other trade agreements or the UN, things like that, that you would typically classify as foreign policy.
Matt Grossmann: So it seems like the popular debate about this is relatively consistent with your findings on the weight that folks are placing on these dimensions, but maybe less consistent with your findings about changes in attitudes. Is that how you see it?
Will Marble: I think that’s basically right. So there was this debate in the [inaudible 00:40:39] about what’s the matter with Kansas and this idea that low income voters were voting against their interests based on cultural values, opposition to gay rights and so on. I think that debate between Thomas Frank, Larry Bartels and others concluded that cultural issues didn’t seem to matter very much for low income voters or non-college voters. It was really the highly educated who placed a lot of weight on those issues. I think that debate happened a little bit prematurely in the sense that it happened right at the beginning of this realignment along educational lines. If you extend the time series forward from 2008 to 2020, you really do see that these cultural issues, LGBT rights, abortion, things like that, matter to a great extent for the working class as well as college educated voters. So I think in that sense, the popular narrative is right.
You see a similar thing on racial issues, especially in the Trump era where the Republican Party especially has been much more explicit in talking about racial issues, much more explicit about appealing to white identity. This is a trend that I see reflected in the data here as well. I think what is new in my paper, or especially new, is the findings on economics. I think this is less appreciated both in the popular debate and I think in the political science literature as well, that we see people with college degrees are increasingly expressing pro-redistributive views and pro-spending views in a way that they weren’t, say, three decades ago.
Matt Grossmann: So you focus most of your analyses on white voters, not including black, Hispanic and Asian American and other minority voters, but you do show the trends among non-white voters. It sure looks like there is a sign of a flip in the role of education among non-white voters. Although the current gap is not very large between educated and less educated non-white voters, it certainly looks like we could end up in a situation like we were in 2008 where we’ve sidelined it as this is about white voters right at the wrong time in the process when it might be continuing for minority voters. So how much can we really say that this is a phenomenon among white voters versus non-white voters?
Will Marble: Yeah, I think you’re right that it’s certainly possible that we’re seeing the beginnings or maybe middle of a similar realignment along educational lines among non-white voters. I think the analysis that I show here focuses on white voters first because a lot of the popular debates have been about the white working class and things like that. But I think more fundamentally, there’s still a lot less variation to explain among non-white voters. I think the sort of framework that I developed might not be as applicable to non-white voters. You might think that the role of racial issues, for instance, is much different among non-white voters than it is for white voters. The sorts of survey questions that get asked often tap into attitudes, antagonistic attitudes, toward racial and ethnic minorities. These questions just don’t seem to play the same role among racial and ethnic minorities as they do for white voters.
But when it comes to more traditional moral issues, especially as the GOP in recent years has turned back, I would say, to moral issues with the overturn of Roe v. Wade, these book bans, don’t say gay laws, et cetera. I certainly think it’s plausible that we could see some working class minority voters increasingly vote on these conservative cultural attitudes as well. So I don’t want to make any big bets about what will happen in the future, but I think it’s plausible that we might see a similar pattern if these trends of issue emphasis continue into the foreseeable future.
Matt Grossmann: So the way you set up your analysis means that we’re comparing people in the same party environment in each election, but certainly there are signs that the parties are changing the relative salience and the positions that they take on these issues over time. How much do you think that that is part of the story, as the parties have changed their positions or their emphasis? That is, would this have happened if everybody had talked about economic policy as much of the agenda as they had in the past? Would it be happening if the parties had stayed in their older positions? Or is there a liberal move across the board on some of these issues?
Will Marble: Yeah, this is a great question. I think it’s a good time to point out some of the methodological challenges with answering this question. So part of what I do in the paper is to try to develop this issue voting framework and really think through what we can infer about different components of vote choice. So I adopt this spatial voting, multidimensional spatial voting framework, and this framework decomposes vote choice into three components. There’s voters’ own attitudes, there’s the parties or the candidate’s positions across different issues, then finally, there’s the weight that different voters place on each issue. It turns out that it’s really methodologically difficult to disentangle changing party positions over time from changing issue weights. So if you do a simple regression of vote choice on issue attitudes, in this framework I show that the coefficient you get from that regression is a combination of both issue weights and the candidate’s platform divergence.
This is intuitive. You wouldn’t think that there should be a correlation between vote choice and attitudes if there’s no weight placed on that issue. Also, you wouldn’t think that there should be a correlation if there’s no difference between the candidates on that issue. So the upshot is that measuring these three components across time is really difficult. So the partial solution that I come to in the paper is to say that, well, if we look within an election, different groups face the same choice of candidates so that we can hold fixed the party position and then look at the relative, so we can interpret differences in these correlations or these coefficients as differences in issue weights. So here, this is the basis by which I talk about relative issue weights. So that’s a long-winded way of getting to my answer to your question, which is that it’s hard to know for sure what’s going on with candidates platforms because we don’t really have good data that measures platforms on the same scale that we can measure voters’ attitudes here, except in limited circumstances on limited issues and in a few elections. That said, when party… It does seem… This isn’t from the paper, but I think other research has pointed to the fact that parties really are taking out distinct positions, especially in this longer time span going back to the eighties and nineties on these cultural issues. And when they stake out these different positions, I think it’s likely that candidates are going to highlight those positions in the campaign and that could potentially influence the weight that voters attach to different issues. So it’s not something that I have that much data to say, “This is the reason that there’s this increased issue emphasis,” but it seems plausible that as parties are staking out distinct positions, they’re campaigning on these issues and it leads voters to increase the weight that they place on those issues as well.
Matt Grossmann: The other interview is with Josh Zinger who looks at some of the same individual level trends, but also looks at them geographically and tries to find some interactions there, and basically shows that it’s college educated people in college educated areas that are moving the furthest leftward and vice versa. And it is interesting here that in some sense, the press discovered this first more as a geographic pattern than an individual level pattern. It raises the question of whether this is about culture that people are surrounded by in educated areas or the kind of social networks that they’re in versus an individual level effect. How do you see those findings coalescing?
Joshua Zingher: Yeah, this was a really interesting paper and I think it’s indicative of the geographic sorting that we see along educational lines. We see that college graduates have become increasingly clustered in economically prosperous metropolitan areas, thinking like Boston or San Francisco, places like that. And so, what Zinger’s analysis, my interpretation of Zinger’s analysis is that this conditional effect whereby the effect of holding a college degree is greater in places with a high concentration of college graduates is that it tells us something about people who are in the off diagonals, people who don’t have college degrees but end up in places dominated by people with college degrees and vice versa, people who do hold a college degree but end up in a place where there aren’t that many people like them.
My hypothesis here is that college graduates are clustered in metro areas and these people tend to be especially liberal on cultural issues. I think there’s some evidence that people who self-select into cities and other metro areas have a more cosmopolitan orientation on these cultural issues. That would help explain why you see an especially large effect for college graduates is where there is a high proportion of college graduates. And then these people are also probably more liberal on economic issues as well. And this is a conjecture of mine, but I think some of my other work and other research suggests that there’s something to this. The idea is that people who are moving to these big metro areas often are moving away from family, friends, traditional cultural or traditional social institutions. And so, they may depend a little bit more on government spending or at least be more amenable to it because they don’t have the informal safety net to fall back on. And as I mentioned before, people who live in these metro regions also tend to be exposed to the externalities of inequality a little bit more acutely.
If you live in a very densely populated area, it’s very common to see people who are from a different social class from you and maybe feel negatively about the social situation that they find themselves in, which could lead them to be more willing to spend money, spend government money in particular to fix that. And then a final explanation for this could be neighborhood effects, as you alluded to. It could be that there are these liberal bubbles, for lack of a better term, and it becomes self-reinforcing where people move there and then they adopt some of the values of the places that they’ve moved to. And there’s some nice work by Greg Martin and Josh McCain suggesting that there is some causal effect on people’s attitudes when they actually move to a place. I think all of these explanations could help explain this really interesting pattern that those Zinger documents.
Matt Grossmann: When you think about education as having an effect on people’s political attitudes or vote choices, are you thinking about something that occurs at college that changes their views? Because on the other hand, lots of people flipped in 2016 who were 50 years out of college. So it could just be about changes in salience, but it could be that college puts you in different social networks, different communities, different professions, you have different surroundings, and that changes your culture. How are you thinking about the effect of education?
Joshua Zingher: Yeah, so in this paper I use education as a pretty crude proxy for class. I think there’s been a lot of popular and academic discussion about the white working class, and I think an imperfect proxy, certainly imperfect, but it’s one that we have available to us over a really long time span. I think certainly there’s partly a causal effect of college. You mentioned some of the research on this earlier, but I think a lot of this is also about the social environment that people who have college degrees find themselves in. I think I would say it’s a bit of column A, a bit of column B.
Higher education certainly exposes people to people from different geographic, cultural, economic backgrounds that might lead them to adopt different views on cultural issues, exposes people to new ideas which might lead them to question or reject more traditional values. But also importantly, there are different labor market outcomes between college and non-college educated voters, workers. And college grads as I mentioned earlier, are more mobile and this is likely to be important for attitude. I would say that it’s a bit of the causal effect of college that certainly plays into it. But the way I’m really thinking about it in this paper is thinking about education as a proxy for broader social class.
After the 2016 election, there was a very large increase in popular interest in issues related to this and related to the role of racial attitudes in voting, which I know you’ve worked on separately from this as well. How would you put the current state of knowledge there? Are we still too much reacting to the Trump era and the specifics of it, or have we learned more about these long-running trends that maybe we just hadn’t been paying as much attention to? And would you rate popular commentary and political science as converging here or still in pretty different worlds?
Joshua Zingher: I think one benefit of the 2016 and 2020 elections was to increase interest in these longer running trends. And I think the initial wave of research really looked a lot at these specific elections, trying to understand what was different about Trump. But more recent research, my paper included, has looked at taken a broader, more holistic look at these trends over time. And I think that public debates about these trends are somewhat right, somewhat always a little bit reductive. But I think there are a couple things it gets right. As we’ve discussed before, there’s this idea that cultural and racial issues are becoming increasingly important, and I think that’s true. And it wasn’t obvious going back to 2008 or before that that was actually happening. But I think with some benefit of a little bit more data, we see that that actually is true. A lot of the debates among pundits and strategists are about whether the Democratic Party should emphasize or de-emphasize cultural issues are reflecting some real trends in the data.
That said, I think a lot of this debate hinges on how much weight voters place on different issues, but this is a really slippery term and I’m hoping that by providing this framework in my paper, we can help think through a little bit more carefully about what we should infer from these correlations between issue attitudes and vote choice. And then finally, especially in 2020, we saw that Biden really won due to suburban counties with a highly educated electorate. You think about the county surrounding Atlanta as an example, the Atlanta suburbs. And these counties have swung left, and I think my analysis suggests that it’s not just cultural issues that explain this, it’s both economic and cultural issues. And these highly educated people tended to vote for a long time on cultural issues, issues of race and moral values, but now they are increasingly economically liberal as well. I think that helps explain to some extent why we’re seeing this shift at the macro level toward Democrats in counties that are doing pretty well economically, often in suburban centers of economic activity.
Matt Grossmann: So you, as we’ve said, methodologically took the party positioning out of your analysis here, but it still seems to have implications for those debates you were referencing about what the parties should do. And if I will put it somewhat crudely, it doesn’t seem to offer much hope for changing anything if what we’re doing is increasingly lining up with one side or the other across the board and making a voting determination based on that. That doesn’t seem like anything that’s likely to change, even if, say, Trump isn’t in the election or even if one of the parties decided to moderate on a particular issue attitude. Is that your sense as well?
Joshua Zingher: I think yes and no. I think that these coalitions are likely to be relatively stable, at least at the macro level going forward, because we are seeing this increasing correlation across a whole range of issues, where college educated voters and non-college voters are just staking out distinct positions across different issue areas. That said, we’re in an era of intense party competition for the presidency especially, and really small changes in candidate positioning or the weights that voters place on different issues could end up mattering a lot for these outcomes. It could be that even paradoxically, even though we are in this situation where there are really deep divisions in the electorate, there are still moderates out there. There are still people who are not set on one party versus another. And in that environment, really small changes to positioning could have large implications for election outcomes.
And then finally, I would say that I would push back against any notion, not that you’ve said this explicitly, but push back against the notion that what elites say doesn’t actually matter, because public opinion and elite activities are always going to be in a feedback loop, where voters are learning and updating their views from politicians and politicians are updating their stances in response to the electorates that they face. It’s a complicated back and forth, so I would really not want to say that these divisions imply that there’s nothing candidates can do to either shape public opinion or influence the outcome of the elections that they’re running in.
Matt Grossmann: There’s a lot more to learn. The Science of Politics is available biweekly from the Niskanen Center. And I’m your host, Matt Grossman. If you like this discussion, here are the episodes you should check out next, linked on our website. Higher Education, an Engine of Social mobility or a driver of Inequality. The politics of School from Home, Why the Baby Boomers Rule American Politics, How Marriage and Inequality Reinforce Partisan Polarization, and Is demographic and Geographic Polarization Overstated? Thanks to Joshua Zinger and Will Marble for joining me, please check out Diploma Divide and what explains educational polarization among white voters, and then listen in next time.