By Eric Levitz | Intelligencer | March 3, 2021
In the United States, every season is campaign season. Four months after America last went to the polls, Democrats are still refining their autopsies of the 2020 race and already governing with an eye toward the 2022 midterms. Meanwhile, on the other side of the aisle, Republicans are trying to figure out just how firm Donald Trump’s grip on their party really is — and debating whether that grip should be stronger or weaker.
To gain some insight into these matters, Intelligencer turned to our favorite socialist proponent of ruthlessly poll-driven campaigning, David Shor. A veteran of the 2012 Obama campaign, Shor is currently head of data science at OpenLabs, a progressive nonprofit. We spoke with him last week about how his analysis of the 2020 election has changed since November, what Democrats need to do to keep Congress after 2022, and why he thinks the Trump era was great for the Republican Party (in strictly electoral terms).
What are the most important things you’ve learned about the 2020 election between the last time we spoke and today?
What’s changed since November is that we now have individual-level vote-history data in a bunch of states. And we also have a lot more precinct-level data. And people have had more time to run surveys. So the picture has gotten clearer.
One high-level takeaway is that the 2020 electorate had a very similar partisan composition to the 2016 electorate. When the polls turned out to be wrong — and Trump turned out to be much stronger than they predicted — a lot of people concluded that turnout models must have been off: Trump must have inspired higher Republican turnout than expected. But that looks wrong. It really seems like the electorate was slightly more Democratic than it had been in 2016, largely due to demographic change (because there’s such a large partisan gap between younger and older voters, every four years the electorate gets something like 0.4 percent more Democratic just through generational churn). So Trump didn’t exceed expectations by inspiring higher-than-anticipated Republican turnout. He exceeded them mostly through persuasion. A lot of voters changed their minds between 2016 and 2020.
At the subgroup level, Democrats gained somewhere between half a percent to one percent among non-college whites and roughly 7 percent among white college graduates (which is kind of crazy). Our support among African Americans declined by something like one to 2 percent. And then Hispanic support dropped by 8 to 9 percent. The jury is still out on Asian Americans. We’re waiting on data from California before we say anything. But there’s evidence that there was something like a 5 percent decline in Asian American support for Democrats, likely with a lot of variance among subgroups. There were really big declines in Vietnamese areas, for example. Anyway, one implication of these shifts is that education polarization went up and racial polarization went down.
In other words, a voter’s level of educational attainment — whether they had a college degree — became more predictive of which party they voted for in 2020 than it had been in 2016, while a voter’s racial identity became less predictive?
Yeah. White voters as a whole trended toward the Democratic Party, and nonwhite voters trended away from us. So we’re now somewhere between 2004 and 2008 in terms of racial polarization. Which is interesting. I don’t think a lot of people expected Donald Trump’s GOP to have a much more diverse support base than Mitt Romney’s did in 2012. But that’s what happened.
Does the available data give us any insight into why? Do you have any sense what was behind the large rightward shift among Hispanic voters?
One important thing to know about the decline in Hispanic support for Democrats is that it was pretty broad. This isn’t just about Cubans in South Florida. It happened in New York and California and Arizona and Texas. Really, we saw large drops all over the country. But it was notably larger in some places than others. In the precinct-level data, one of the things that jumps out is that places where a lot of voters have Venezuelan or Colombian ancestry saw much larger swings to the GOP than basically anywhere else in the country. The Colombian and Venezuelan shifts were huge.
One of my favorite examples is Doral, which is a predominantly Venezuelan and Colombian neighborhood in South Florida. One precinct in that neighborhood went for Hillary Clinton by 40 points in 2016 and for Trump by ten points in 2020. One thing that makes Colombia and Venezuela different from much of Latin America is that socialism as a brand has a very specific, very high salience meaning in those countries. It’s associated with FARC paramilitaries in Colombia and the experience with President Maduro in Venezuela. So I think one natural inference is that the increased salience of socialism in 2020 — with the rise of AOC and the prominence of anti-socialist messaging from the GOP — had something to do with the shift among those groups.
As for the story with Hispanics overall, one thing that really comes out very clearly in survey data that we’ve done is that it really comes down to ideology. So when you look at self-reported ideology — just asking people, “Do you identify as liberal, moderate, or conservative” — you find that there aren’t very big racial divides. Roughly the same proportion of African American, Hispanic, and white voters identify as conservative. But white voters are polarized on ideology, while nonwhite voters haven’t been. Something like 80 percent of white conservatives vote for Republicans. But historically, Democrats have won nonwhite conservatives, often by very large margins. What happened in 2020 is that nonwhite conservatives voted for Republicans at higher rates; they started voting more like white conservatives.
And so this leads to a question of why. Why did nonwhite voters start sorting more by ideology? And that’s a hard thing to know. But my organization, and our partner organizations, have done extensive post-election surveys of 2020 voters. And we looked specifically at those voters who switched from supporting Hillary Clinton in 2016 to Donald Trump in 2020 to see whether anything distinguishes this subgroup in terms of their policy opinions. What we found is that Clinton voters with conservative views on crime, policing, and public safety were far more likely to switch to Trump than voters with less conservative views on those issues. And having conservative views on those issues was more predictive of switching from Clinton to Trump than having conservative views on any other issue-set was.
This lines up pretty well with trends we saw during the campaign. In the summer, following the emergence of “defund the police” as a nationally salient issue, support for Biden among Hispanic voters declined. So I think you can tell this microstory: We raised the salience of an ideologically charged issue that millions of nonwhite voters disagreed with us on. And then, as a result, these conservative Hispanic voters who’d been voting for us despite their ideological inclinations started voting more like conservative whites.
Are these problems with Democratic positioning or with “disinformation”? Obviously, Joe Biden didn’t campaign on police abolition and worker control of the means of production. So there was a disconnect between the reality of the party’s platform and how it was perceived. Closing that gap, through a “Latino Anti-Disinformation Lab,” appears to be a focus of Democrats’ postelection efforts to fix their problem with Hispanic voters. Does that make sense as a path forward?
I’d say this: The decline that we saw was very large. Nine percent or so nationwide, up to 14 or 15 percent in Florida. Roughly one in ten Hispanic voters switched their vote from Clinton to Trump. That is beyond the margin of what can plausibly be changed by investing more in Spanish media. And I don’t think a shift that large can be plausibly attributed to what was said in WhatsApp groups or not buying enough in YouTube ads. I think the problem is more fundamental.
Over the last four years, white liberals have become a larger and larger share of the Democratic Party. There’s a narrative on the left that the Democrats’ growing reliance on college-educated whites is pulling the party to the right (Matt Karp had an essay on this recently). But I think that’s wrong. Highly educated people tend to have more ideologically coherent and extreme views than working-class ones. We see this in issue polling and ideological self-identification. College-educated voters are way less likely to identify as moderate. So as Democrats have traded non-college-educated voters for college-educated ones, white liberals’ share of voice and clout in the Democratic Party has gone up. And since white voters are sorting on ideology more than nonwhite voters, we’ve ended up in a situation where white liberals are more left wing than Black and Hispanic Democrats on pretty much every issue: taxes, health care, policing, and even on racial issues or various measures of “racial resentment.” So as white liberals increasingly define the party’s image and messaging, that’s going to turn off nonwhite conservative Democrats and push them against us.
When you say that white liberals are to the left of the typical Black Democrat on racial issues, how much does that depend on the definition of a racial issue? For example, one policy fight that often pits the interests of white liberal Democrats against those of the Black working class is housing and school integration. There are a lot of highly educated, white, liberal areas — full of “Black Lives Matter” lawn signs — which nevertheless oppose affordable-housing projects or school-redistricting plans that would bring less wealthy, less white students to their kids’ classrooms. The white liberals who oppose efforts to end de facto segregation may know the enlightened answer to abstract questions about the nature of racial inequality, but I’m not sure that puts them to the left of nonwhite voters on racial issues, properly defined.
Yeah, no, absolutely. White liberals do give more progressive responses across a wide battery of traditional racial resentment questions like, “Do you believe that the reason why African Americans can’t get ahead is due to discrimination or due to other factors?” But I think it’s important to put “racial resentment” in quotes whenever you talk about it. I’m not claiming that white liberals are somehow less racist than people of color, to the extent that question even makes sense. And I do think if you asked about affirmative action and inclusionary zoning, rather than these more abstract questions that political scientists use for measuring racial resentment, you could find a different breakdown.
But I think the split on those abstract questions captures something real. In liberal circles, racism has been defined in highly ideological terms. And this theoretical perspective on what racism means and the nature of racial inequality have become a big part of the group identity of college-educated Democrats, white and nonwhite. But it’s not necessarily how most nonwhite, working-class people understand racism.
How do they differ?
I don’t think I can answer that comprehensively. But if you look at the concrete questions, white liberals are to the left of Hispanic Democrats, but also of Black Democrats, on defunding the police and those ideological questions about the source of racial inequity.
Regardless, even if a majority of nonwhite people agreed with liberals on all of these issues, the fundamental problem is that Democrats have been relying on the support of roughly 90 percent of Black voters and 70 percent of Hispanic voters. So if Democrats elevate issues or theories that a large minority of nonwhite voters reject, it’s going to be hard to keep those margins. Because these issues are strongly correlated with ideology. And Black conservatives and Hispanic conservatives don’t actually buy into a lot of these intellectual theories of racism. They often have a very different conception of how to help the Black or Hispanic community than liberals do. And I don’t think we can buy our way out of this trade-off. Most voters are not liberals. If we polarize the electorate on ideology — or if nationally prominent Democrats raise the salience of issues that polarize the electorate on ideology — we’re going to lose a lot of votes.
Don’t these ideological self-descriptions carry similar definitional problems as “racial resentment”? Most voters may not identify as liberals. But judging from opinion polls, most voters do reject the lion’s share of the conservative movement’s governing priorities. In Congress, a “conservative” is typically a lawmaker who supports tax cuts for the rich and funding cuts for Medicaid, while opposing a higher minimum wage and another round of stimulus checks. Those are all extremely unpopular positions.
It seems important then to get clarity on what these ideological labels do and don’t mean. If taken at face value, the data looks pretty ominous for Democrats: They’ve built a coalition premised on overwhelming support from these nonwhite groups, but that support was rooted in historically contingent social conditions — not substantive agreement — and now those conditions are eroding, clearing the way for an emerging “conservative” majority. On the other hand, if you look at the polling of the biggest policy debates in Congress over the past eight years, you might conclude that there’s a natural liberal majority in this country and that the GOP is the party whose coalition is an “unnatural” agglomeration of groups held together by accidents of history.
I agree with everything you said.I do think that liberals sometimes take the ambiguities of ideology too far. A lot of progressives insist that ideological self-identification means nothing. And we know that isn’t true. One of the big patterns of the last 40 years is that ideological self-description has become increasingly correlated with partisanship and increasingly correlated to views on issues.
But there is still a large universe of policy questions — mostly economic but not exclusively — where a large majority of the public agrees with us. A $15 minimum wage polls above 60 percent; that couldn’t happen without a lot of “moderates” and “conservatives” supporting the policy. What I take from that is: Ideological polarization is a dead end. If we divide the electorate on self-described ideology, we lose — both because there are more conservatives than liberals and because conservatives are structurally overrepresented in the House, Senate, and Electoral College. So the way we get around that is by talking a lot about progressive goals that are not ideologically polarizing, goals that we share with self-described conservatives and moderates. Even among nonwhite voters, those tend to be economic issues. In test after test that we’ve done with Hispanic voters, talking about immigration commonly sparks backlash: Asking voters whether they lean toward Biden and Trump, and then emphasizing the Democratic position on immigration, often caused Biden’s share of support among Latino respondents to decline. Meanwhile, Democratic messaging about investing in schools and jobs tended to move Latino voters away from Trump.
Is that primarily a function of the fact that the marginal Hispanic voter — the one who’s least attached to the Democratic Party — is to the right of the typical Hispanic voter? Like, it isn’t the case that a majority of Hispanic voters respond negatively to immigration messaging, is it?
No. I mean, Hispanic voters are more liberal on immigration than white voters. But I think that, for one thing, the extent to which Hispanic voters have liberal views on immigration is exaggerated. If you look at, for example, decriminalizing border crossings, that’s not something that a majority of Hispanic voters support. Pew’s done a lot of polling on immigration reform, and if you ask things like, “Should we deport the undocumented population, should we give them a path to permanent residency, or should we give them a path to citizenship?” citizenship only gets a little over 50 percent support among Hispanic voters. So I think liberals really essentialize Hispanic voters and project views about immigration onto them that the data just doesn’t support.
Now, how we should campaign and what we should do once in office are different questions. Our immigration system is a humanitarian crisis, and we should do something about that. But the point of public communication should be to win votes. And the way that you do that is to not trigger ideological polarization.
What’s your (way too early) assessment of Democrats’ odds of retaining Congress after the midterm? What do they need to achieve, in statistical terms, to pull that off? And then, from a substantive point of view, are there things that they can do in office to make hitting those marks easier?
As a baseline, midterms are usually very bad for the party in power. In the past 70 years, the incumbent party has gained seats in the House and Senate maybe once or twice. The last one was in 2002. The regularity of how bad midterm environments are for the president’s party is one of the most striking findings in political science. Generally speaking, over the last 30 to 40 years, the party that controls the presidency gets about 47 percent of the vote nationwide. Add in the fact that the House already has a fairly substantial pro-Republican bias — the median House seat is something like three points to the right of the country overall — it means that in the base scenario, Democrats are headed for near-certain doom. If we replicate the GOP’s post-9/11, 2002 midterm performance, we have a chance. If we replicate the second-best presidential-party midterm from the past 40 years, we lose.
The good news is that there’s a strong case for thinking this time might be different. I’m not a macroeconomist, but it seems like Joe Biden might preside over a post-corona economic boom. Already, Biden’s approval rating is very strong. The best predictor of how a midterm is going to shake out is how popular the president is. So, for now, everything looks about as good as you could hope for.
But we have no margin for error. If we conduct ourselves the way we did after 2008, we’re definitely going to lose. And due to the way that our electoral system works, we really could be locked out of power for a very long time, just like we were after 2010. So that means the need for messaging discipline is stronger than ever. But keeping the national conversation focused around popular economic issues probably won’t be enough. Since the maps in the House of Representatives are so biased against us, if we don’t pass a redistricting reform, our chance of keeping the House is very low. And then the Senate is even more biased against us than the House. So, it’s also very important that we add as many states as we can. Currently, even if we have an exceptionally good midterm, the most likely outcome is that we lose one or two Senate seats. And then, going into 2024, we have something like seven or eight Democrats who are in states that are more Republican than the country overall. Basically, we have this small window right now to pass redistricting reform and create states. And if we don’t use this window, we will almost certainly lose control of the federal government and not be in a position to pass laws again potentially for a decade. In terms of putting numbers on things, I think that if we implemented D.C. and Puerto Rican statehood and passed redistricting reform, that would roughly triple our chance of holding the House in 2022 and roughly the same in the Senate. The fact that it’s possible to triple those odds is a testament to how bleak the baseline case is. So we need to pass those reforms and we need Biden to remain popular. If his approval rating is below 50 by the end of the year, we’re probably fucked.
Is there a tension between those two imperatives? In the past, I’ve heard you talk a lot about “thermostatic” public opinion — how voters tend to move right when Democrats are in power, and left when Republicans are in power, and generally display a bias toward the status quo and against policy change. Could adding multiple states to the Union, and changing the way that we go about allocating House representation — specifically in a manner that diminishes the influence of white, rural voters — spur thermostatic backlash? And if so, could maintaining Biden’s current approval, and implementing the reforms necessary for Democrats to stay competitive at the congressional level, present an irresolvable dilemma?
I can’t claim to know exactly what the electoral effects would be of doing these things. But all of the polling I’ve seen suggests that things like HR 1 and adding states are above water. They’re not as popular as a lot of economic issues, but they’re above 50 percent. Electoral backlash doesn’t typically come from doing things that poll at 53 or 54 percent. It comes from doing things that poll at 30 or 40 percent. And so I think that the downside of this stuff is low. I think the level of voter interest in procedural issues is low. If we lived in a world where voters punished politicians for playing procedural hardball, we would have a lot fewer Republicans in office.
And actually, in some ways, pursuing procedural reforms that don’t concern voters much — but which do get the other party all worked up — could be electorally beneficial. If you can get the other party to talk about something that voters don’t care about, that’s good. People don’t always think about media attention as a fixed quantity. But it is. To the extent that the coronavirus impacted the 2020 election, I think one positive political effect it had for Democrats was that whenever the media was talking about the coronavirus, they weren’t talking about Hunter Biden or immigration. And I think that kind of blocked Republicans from creating and inserting wedge issues. If Republicans decide to make 2022 into a referendum on independent redistricting, that will eat up space that could have otherwise gone to effective attacks. We should dare them to do it.
We talked a lot about the rightward drift of Hispanic voters in 2020. But the other big change was a leftward shift among college-educated whites. Understanding the cause of that shift seems pretty important. If these college-educated voters were primarily rejecting Donald Trump, Democrats might not be able to count on their support in 2022 and beyond.
Yeah, it’s a great question. Let’s start with numbers: In 2016, non-college-educated whites swung roughly 10 percent against the Democratic Party. And then, in 2018, roughly 30 percent of those Obama-Trump voters ended up supporting Democrats down ballot. In 2020, only 10 percent of Obama-Trump voters came home for Biden.
So I think what this shows: There is a long-term trend of increasing education polarization here and in every other country in the West. But the fact that education polarization declined significantly in 2018 — when Trump wasn’t on the ballot — and picked up again in 2020 suggests that Trump is personally responsible for a significant portion of America’s education polarization. I think that there’s a really strong case that this transition was specifically about Donald Trump.
A lot of people theorized that we first alienated Obama-Trump voters during the fight over comprehensive immigration reform and that their rightward movement was already apparent in 2014. But if you actually look at panel data, it seems really clear that these people didn’t start identifying as Republicans until Trump won the GOP nomination. I think there’s a very strong empirical argument that Donald Trump was the main driver of the polarization we’ve seen since 2016. He just personally embodies this large cultural divide between cosmopolitan college-educated voters and a large portion of non-college-educated voters. Those divides take a lot of different forms: attitudes toward race, attitudes toward gender, opinions on what kinds of things you’re allowed to say, or how you should conduct yourself. And you know, as Trump became the nominee, and as the media made politics the Donald Trump Show for the last four years, that led to increasing political polarization on attitudes toward Donald Trump specifically. I think the reason why we saw less education-based voting in 2018 is that Trump was a smaller part of the media environment than he had been in 2016 or would be in 2020.
Looking ahead to 2022, and just thinking about the next four years, the big question is how much is Donald Trump going to shape media coverage of the Republican Party or the Republican Party’s own branding? And I don’t know the answer to that question. If Trump fades out of the spotlight, I’d expect some level of education depolarization, particularly if Democrats show ideological discipline.
That speaks to a question I’ve been mulling for a while. During the 2016 campaign, Vox developed this concept of “the Trump Tax,” which was a measure of the electoral penalty that Republicans were paying for picking the most unpopular nominee in polling history. Basically, it took a “fundamentals” model of how one would expect a Republican presidential candidate to perform, given economic conditions and other background factors, and then measured how much lower Trump’s support was than that. And yet, while Trump remained historically unpopular in office, he also helped the GOP increase its structural advantages at every level of government. So I’ve long wondered: Was Donald Trump’s unpopularity with the general public more detrimental to the Republican Party than his gift for deepening education polarization was valuable?
So, in 2016, Hillary Clinton got 51.1 percent of the two-party vote. Obama got 52 percent in 2012. In just about any other country, retaining 51.1 percent support would have been enough to keep power. But in this country, between 2012 and 2016, the Electoral College bias changed from being one percent biased toward Democrats to 3 percent biased toward Republicans, mainly because of education polarization. So Donald Trump is unpopular. And he does pay a penalty for that relative to a generic Republican. But the voters he’s popular with happen to be extremely efficiently distributed in political-geography terms.
Imagine Hillary Clinton had run against Marco Rubio in 2016. Rubio is a less toxic figure to the public as a whole, so let’s say he performed as a generic Republican would have been expected to, and Hillary Clinton’s share of the two-party vote fell to 49.6 percent. If she had maintained Obama’s coalition — if her 49.6 percent had the same ratio of college-to-non-college-educated voters as Obama had in 2012 — she would have won that election. And then, if you look at the implications that would have had down-ballot, especially in the Senate, Republicans would have been a lot worse off with a narrow majority coalition — that had a Romney-esque split between college and non-college voters — than they were with the Trump coalition.
So I think the Trump era has been very good for the Republican Party, even if they now, momentarily, have to accept this very, very, very thin Democratic trifecta. Because if these coalition changes are durable, the GOP has very rosy