vision 2020

David Shor’s Postmortem of the 2020 Election

Party’s over. Photo: Tayfun Coskun/Anadolu Agency via Getty Images

For the Democratic Party, the 2020 election was a devastating victory. Joe Biden won a larger share of the popular vote than any presidential challenger since 1932. The Rust Belt shed some red, while the Sun Belt trended blue. Democrats retained control of the House — and still have an outside shot at eking out a Senate majority.

And yet the party’s activist base is licking its wounds, while its elected officials claw at each other’s throats.

The reason for this odd state of affairs is simple: Barring an unlikely triumph in Georgia’s Senate runoffs in January, the 2020 election will go down as a down-ballot disaster for the Democratic Party. Even as Biden won the popular vote by as much as five points, Democrats saw their House majority shrink, most of their top Senate challengers fall flat, and their attempts to flip state legislatures come up empty. What’s worse, these losses appear attributable to trends that raise ominous questions about the party’s electoral prospects going forward: The most xenophobic Republican president in modern memory made large gains with Latino voters, while white rural America continued its steady rightward march. Most Democrats weren’t prepared for these disappointments.

But David Shor was.

A veteran of the 2012 Obama campaign and the Democratic consultancy Civis Analytics, the 29-year-old data scientist currently advises liberal super-PACs on advertising. Back in July, he told Intelligencer that Democrats were bleeding support from Hispanic voters, that the combination of education polarization and the decline of ticket-splitting was creating big problems for Democrats in the Senate, that Biden would likely need to win the popular vote by about four points to secure a comfortable Electoral College majority, and that the presidential race was likely to end up much tighter than public polls were suggesting. (On the other hand, right after the Nevada caucuses last February, Shor did tell me that Bernie Sanders would be the Democratic nominee, so sometimes his data does lie.)

Most recently, Shor spoke with Vox’s Dylan Matthews about his theory for why the 2020 polls were so inaccurate: In short, voters with low levels of social trust don’t answer surveys (but do tend to vote Republican these days), while liberal voters cooped up during the pandemic and George Floyd protests started answering pollsters’ calls at exceptionally high rates.

Intelligencer caught up with Shor this week to discuss the 2020 results, why progressives might be in for a lost decade at the federal level, and the socialist case for ruthless moderation.

How would you summarize the upshot of the 2020 election results in three sentences or less?

There was a uniform swing of roughly 1.5 percentage points toward us, relative to 2016. Education polarization, which is the gap between how college-educated white people vote and how non-college-educated white people do, continued to grow. And Hispanic voters swung against us by a lot. Exactly how much is unclear. But potentially by as much as 12 points. And the sum total of all the things was that we were able to narrowly win the Electoral College. So that was the big picture at the presidential level.

Pretty sure that was more than three, but we’ll let it slide. So, while unseating an incumbent president is cool and good, it seems to me that this election also produced a lot of very bad news for Democrats down-ballot. 

Yeah. Going into 2020, we faced large structural biases in the House and in the Senate and the Electoral College. Now, it looks like the Electoral College gap is significantly larger than it was in 2016, and there’s really no indication that that gap is going to narrow over the course of this decade.

Because of education polarization, or regional trends, or —

It’s educational polarization. As a whole, the Midwest continues to trend against us, and I think this election in a lot of ways was a natural continuation of the trends that happened in 2016. The public polling suggested that a bunch of non-college-educated whites were going to come home to the Democratic Party, and while that may have been true in a few discrete places, nationally that did not happen. In fact, the gap got worse. And, broadly speaking, the worse that you do with non-college-educated whites, the more structurally disadvantaged you are in the American electoral system. And this is especially true now that all politics is national. We spoke about this last time but, there’s been this 20-year — or arguably 40-year — trend toward less ticket-splitting on the Senate, House, and state-legislative levels. That continued. The ultimate correlations between presidential vote share and Senate vote share was higher in 2020 than it was in 2016.

Interesting. After a week of wringing my hands over Susan Collins’s landslide reelection in Maine, I’d begun to assume otherwise.

Yeah. There were exceptions, but overall, correlation was higher. And then in the House, just about every Democrat in a very red district lost, whether it was Colin Peterson or Joe Cunningham. Same thing on the state-legislative level. So there’s a systematic decline in ticket-splitting going on. Twenty years ago, heavily white, non-college-educated states like Montana were already wildly overrepresented in the Senate. But we could recruit a charming liberal and win. And that doesn’t seem to be true anymore.

So what is to be done?

So the path for avoiding a kind of lost decade for progressive change was to win a trifecta in 2020, and then pass a variety of laws — independent redistricting, adding new states, other changes to the rules of the game to make it easier for us to overcome these biases (though we never really had a path to fixing the Electoral College). And there’s still time. I think the Georgia races are very winnable, and everyone in your audience should give their money. Even with a bare majority, I think the party’s self-interest in structural reform is clear enough to force some action.

But I think there’s a very high likelihood that we’re going to end up in this dark scenario where it’s impossible for us to pass laws, or have a functioning government, for the next two years and then head into a midterm election, which, on average, the incumbent’s party gets about 47 percent of the vote in. The very best midterm for an incumbent party in the last 30 years was 2002, when Republicans got 51.9 percent of the vote.

Due to the nature of our maps, and the decline of ticket-splitting, it’s not even clear to me that in that best-case scenario — where we replicate the GOP’s post-9/11 performance — that we don’t lose the House. And if we’re going into 2022 and 2024 with 48 Senate seats, it’s going to be much harder than people realize to claw that back. But if we can’t reduce the structural biases that have appeared in the last ten years by changing the rules of the game, we will have to make the hard choice of changing our party so that we can appeal to these non-college-educated voters who are turning against us.

Education polarization is not only a decades-long trend but also one that spans most of the postindustrial democratic world. The Democratic Party can’t just shrug off its cosmopolitan voting base. It’s always going to be the coalition more associated with cities and liberal professionals. So how much can Democrats really change their party to appeal to those voters?

Yeah, it’s tough. Parties don’t have as much control over who their voters are as many think. One of my favorite examples is Jeremy Corbyn. When he came into the leadership of the Labor Party, I think he legitimately wanted to steer the party in a more working-class direction and move it away from this perceived Blairite shift toward college-educated people. And despite that, he presided over the largest increase in educational polarization in Britain’s history. He ended up flipping one of the richest parliamentary districts in the country and losing the most working-class ones. And then he did it again in 2019.

So there is a real extent to which these parties aren’t always consciously choosing the coalitions they end up with. There isn’t necessarily a guru behind the scenes of the Democratic Party who’s just been turning a giant dial away from the side labeled “Working Class” and toward the one labeled “College Educated.” But that said, I think we shouldn’t be nihilist about this. I think if you look at educational polarization by country over time, there is this general trend where it seems to be increasing almost everywhere. But there are also lots of breaks in the pattern.

Barack Obama, contrary to what a lot of people would expect, actually presided over a depolarization by education. He did better, in relative terms, among non-college-educated whites than John Kerry did. Part of that was the Great Recession. But I think a lot of it was message discipline on Obama’s part. His campaigns focused on economic issues as much as possible and avoided taking hot-button stances on things like foreign policy or immigration. I think he saw that as core to winning.

So, there are these structural forces that are pulling center-left parties away from non-college-educated voters (which we talked about last time). But it’s not an immutable law of nature that can’t be reversed. And we have a responsibility to try to swim upstream because, ultimately, that’s what’s going to determine whether or not we win.

On the determinants of winning, there’s a lively debate within the Democratic Party right now about what went wrong down-ballot in 2020 and what to do about it going forward. Several moderate House Democrats, who represent light-red districts, argue that the party suffered from its association with unpopular left-wing demands like “Defund the police” and ideological labels like “socialism.” They seem to suggest that all party members must distance themselves from radical social movements. Progressives, meanwhile, have argued that the Black Lives Matter protests actually aided Democrats by driving a surge in nonwhite voter registration — which, when combined with the work of organizations like Stacey Abrams’s New Georgia Project, and canvassing efforts like those spearheaded by Ilhan Omar and Rhasida Tlaib — spurred an increase in nonwhite turnout that was indispensable to Biden’s victory. They further maintain that the Democrats’ problems with rural white voters stem from the party’s tendency to “shy away from conversations about race”: You can only neutralize white racial resentment by directly confronting it. Instead of keeping “issues of economic justice and racial justice in separate siloes,” the party must reframe racism as a greedy elite’s strategy for “dividing and conquering” workers. Finally, the left argues that the party must increase its investment in organizing infrastructure, using the so-called Reid machine in Nevada as a model.

How well do either of these analyses line up with your own views?

I think it’s important for us to be clear-eyed about what happened in 2020. We’re not going to know exactly what happened until there’s more analysis of precinct results. But I think that the county-level data we have tells a pretty clear big-picture story. Which is that we won the presidency because, one, while we lost non-college-educated white voters, we kept those defections to a relatively low level, and two, a bunch of moderate Republicans who had voted for Trump in 2016 decided to vote for Biden this time.

Turnout was up, but it was up for both parties. According to Nate Cohn’s estimates, Black turnout was probably up by around 8 percent, but non-Black turnout was up by something like 15 to 20 percent. So we had the highest-turnout election in a century, and despite that, we still only won because a bunch of people switched their votes in our direction.

That basic pattern holds for Georgia specifically?

Yes. If you look at county-level returns in Georgia, it’s pretty clear that nonwhite voters, as a share of the electorate, decreased at a time when the nonwhite share of the state’s population probably increased. Relative to the electorate as a whole, nonwhite turnout fell. And then, among nonwhite voters who turned out, support for the Democratic nominee fell. That’s just not consistent with nonwhite turnout being the decisive factor. The only reason we won is that there were these very large swings toward us among college-educated white people in the Atlanta suburbs.

Now, you can argue for the counterfactual — maybe Black turnout and Black support would have fallen by more if we didn’t do the organizing. That’s possible. In practice though, the empirical effectiveness of all of these election practices are vastly overestimated by people in the industry. I think people in progressive politics have a love affair with the idea of mobilization, just because it seems to promise a way of winning without ideological compromise; all we have to do is turn out all of these non-voting Democrats and then we’ll win. We don’t have to make any changes or try to appeal to Republicans. But Democrats have invested tremendously in get-out-the-vote efforts over the past decade. And I think that it’s pretty clear that none of these interventions worked particularly well. It turns out the effects of canvassing are much lower than people think. The effects of sending get-out-the-vote mail is lower than people think. The effects of these large-scale organizing programs are much less than people think.

And then there’s the other problem with the mobilization theory: As education polarization increases, this truism that “if more people vote, we win” is increasingly less true. If you look at some of the county-level returns in Wisconsin, I think there’s very good evidence that the marginal nonvoters who came out this time — but not last time — leaned Republican. I think that as non-college whites become more and more conservative, the pool of nonvoters has gone from being an overwhelmingly Democratic group to relatively even. Those trends are going to continue. And frankly, if nonwhite, non-college-educated voters keep drifting away from us, it’s going to continue in another direction too.

On the efficacy of campaign tactics: What about television ads? You’ve argued that TV ad spending is so much more cost-effective than canvassing operations that the latter should be largely abandoned in federal races. Yet in 2020, the losing Democratic Senate candidates in key races spent mind-boggling sums on TV while doing little-to-no field due to the pandemic. Is that not evidence that you might be wrong?

So it’s very hard to estimate the impact of advertisements. But I think if you do a rigorous analysis of all of the different modes, it would still tell the story that television is generally the most cost-effective way to win votes. Now, the level of spending we saw this year is literally unprecedented, and so it’s worth studying our points of diminishing returns again. But we also have to be realistic about what ads can do. Ads can be more effective than canvassing and still less effective than people think. Inundating the airwaves with a perfect 30-second spot can maybe move things by 1.5 to two points. That’s realistically the scope of what’s possible. Which is a polarization issue. As people’s votes have started to align more with their cultural values, the efficacy of every campaign intervention has declined. But elections are also very close in the polarization era. So a 1.5-point swing matters.

If the most effective possible campaign intervention only nets you two points max, that seems inauspicious for Democrats making enough inroads with non-college-educated whites to compete in the Senate.

Right, so everybody in politics likes to debate the particulars of campaign tech and strategy. But elections are won or lost primarily on the basis of these broad structural forces and each party’s national brand. Which are related.

So the median voter in the presidential election is about 50 years old, watches about six hours of TV a day, and mostly gets their news from mainstream sources. And that means that, if you want to influence what this person believes, you’re probably not going to get them at the door or even through a paid message. They’re going to form their opinions based on how the media reports on and characterizes the parties.

You can see this in Georgia. Nothing really unique happened there. The state behaved as you would expect given that we had already bottomed out with non-college-educated whites and had room to grow with college-educated whites, who were alienated by the GOP’s Trump-era brand as conveyed by mainstream reporting. And we know that this is a national, structural phenomenon — and not primarily a product of state-level actions or micro-targeting — because the gains we made in the Atlanta suburbs were nearly identical to gains we made in similarly educated counties in other parts of the country.

So how do you change a party’s national brand?

This does get to your earlier question and to this very real tension that exists right now in the Democratic Party. Voters are now determining their opinions about parties in a unified way and not reading about individual local candidates. There’s arguably less local news. But people’s consumption of local news has definitely decreased, while their consumption of national news has increased. So it’s hard for candidates in redder areas to differentiate themselves from the national party than it used to be. This is part of why ticket-splitting is declining.

And that does create some awkward trade-offs. Like, it is now true that what a left-wing congressperson in a deep-blue district says will get transmitted adversarially by the Republican media, and to a significant extent by the mainstream media, to people who disagree. And those people won’t say, “Oh, this left-wing congressperson, well, he’s crazy. But Max Rose? He’s dope.” They’re just going to say, “Oh, Democrats support socialism now, because there’s this one socialist congressperson.”

I think the reality now is that whenever any elected Democrat goes out and says something that’s unpopular, unless the rest of the party very forcefully pushes back — in a way that I think is actually very rare within the Democratic Party currently — every Democrat will face an electoral penalty. And that’s awkward. But I think it’s a natural consequence of polarization and ticket-splitting declining. I think progressives try to get around this awkward reality by saying, “Well, Republicans are going to demonize us no matter what we say or do.” But I don’t think that kind of nihilism is justified. What they say actually does matter. Parties and candidates that say less controversial things, and are associated with less-controversial ideas, win more elections.

I think that the only option that we have is to move toward the median voter. And I think that really comes down to embracing the popular parts of our agenda and making sure that no one in our party is vocally embracing unpopular things. I know that sounds reactionary. But moderates don’t have a monopoly on popular ideas and progressives don’t have one on unpopular ideas. There are a lot of left-wing policies that are both popular and transformational. Worker co-determination. A federal job guarantee. There’s still a lot we can do.

And we also still have a chance to limit how much we need to compromise by winning in Georgia and then passing sweeping structural reforms. But if we don’t, then the reality is that the median voter who gets to determine Senate control is going to remain a non-college-educated 55-year-old in a pretty Republican state who voted for Donald Trump. Probably twice. That’s who we’ll need to win over in order to govern.

Bleak as it is, I feel like you might actually be overselling what’s politically possible under your model of the world. The last time we spoke, you argued that the reason why some congressional Democrats vote for unpopular right-wing policies — like bank deregulation — was that business exerts a lot of cultural influence. Although the “median voter doesn’t want to deregulate banks,” that voter also “doesn’t want a senator who is bad for business,” you argued. Wouldn’t that same dynamic apply to things like a federal job guarantee or worker co-determination? Forcing large corporations to put workers on their boards might poll well. But wouldn’t the Chamber of Commerce spend a lot of money telling voters that it’s “bad for business”?

It’s a good question. I think the reality is that capital is able to get its way on some issues but not others. We’ve seen states adopt $15 minimum wages across the country. The fact that an idea attracts industry opposition does not mean that you can’t implement it without facing voter backlash. It just means that that idea needs to have a lot of public support. There are low-salience issues like banking regulation, where it’s really hard to get large public engagement. It might be true that once Democrats get into office, Democratic ideas are going to become less popular, due to the thermostatic nature of public opinion. So all of the things that are currently polling in the 60s may end up polling in the 50s. But that just means that the bar has to be higher and we have to restrict ourselves to very popular things.

And if you look at what that formula for popular left-wing ideas entails, it means not having large-scale middle-class tax increases. If you look at the experience of the last eight years, whether it’s Illinois rejecting progressive taxes in general or whether it’s the complete failure of single payer in Vermont — which was followed by the Republican takeover of the governorship — I think that there’s this clear story that, you know, the median voter is 50 years old and has a mortgage and doesn’t have a ton of appetite for radical change. So you need to be really careful about what radical changes you pursue.

Stepping back for a second: How do we know that candidates and parties can control what ideas they’re associated with in the public consciousness by adopting issue positions that poll well and disavowing those that don’t? Your analysis is premised on the notion that it matters a lot which issues Democrats choose to support. But that seems like a hard variable to isolate in the real world. And I’ve seen many surveys and focus groups in which voters betray wild misconceptions about what the issue positions of the major parties actually are. So do we have a strong empirical basis for believing that issue positions matter, outside the synthetic environments of a message-test survey?

Yeah. This is something that I’ve changed my mind about over the past five years, the extent to which issues matter. Something that people in our party say a lot is, “The problem with Democrats is they talk too much about issues and they don’t talk enough about values.” And I think that’s actually exactly backward. I think the problem is that fundamentally, swing voters generally don’t share our values and only and mainly vote for us because they agree with us on issues. Obviously, exactly how you market issues or values is a tricky, ill-defined thing. Obviously, voters aren’t policy wonks; they probably don’t know about the details of various tax credits or whatever. But they do actually have many strongly held opinions on things like, “Should gun regulations be stricter or less strict? Should health care be easier to get or not?” Voters do have opinions on these things. We do see this consistently in our hypothetical candidate experiments, where informing voters that the Democratic candidate agrees with them on certain issues moves votes.

But you can see it in the real world too. If you really look at which candidates overperform the fundamentals — which is to say, candidates who do better than you’d expect, given how voters in a congressional district voted for governor the year before, or how it voted presidentially, or whether it’s an incumbent — you see that moderates do better than ideological extremists. I used to not find this compelling. Because there’s an alternative, ideologically convenient narrative that says it’s just a question of candidate quality. There are incentives not to be an ideological outlier. And so maybe the most charming and talented politicians tend to follow those incentives and run as moderates. And maybe extremist candidates are inherently weird people. And maybe that’s all we’re catching.

But I think we have to acknowledge that if you pull up a list of Democrats who have outperformed their presidential race the most in the past 20 years, it is, generally speaking, a list of boring, moderate people. There are a few progressive candidates who legitimately have impressive records. Bernie Sanders is a great example. In 1990, he managed to win Vermont’s House seat two years after the state voted for Bush over Dukakis. And he did it as an open socialist. People really don’t appreciate how insane that was. And over the course of all his elections, he did have a track record of outperforming the fundamentals. But, for the most part, the list is overwhelmingly dominated by moderate politicians. To be clear, that isn’t just issue-taking. Part of it is also that moderate politicians are less threatening to capital, and so they’re more likely to get Chamber of Commerce endorsements. But stepping back, if you look at all the polling and all of the evidence, I think there is a story that taking unpopular positions really hurts. You can see it as a time-series story too. If you look at AOC’s initial polling, or Bernie Sanders’s polling in 2016, or Warren’s standing in the Democratic primary, they were all much more popular before they started embracing a bunch of really unpopular issues.

The best example though might be Donald Trump. His approval rating rested in a very narrow band, for basically this entire election, particularly among non-college-educated whites. There was only one time that shifted, in the course of his entire presidency, and that was when they tried to repeal the Affordable Care Act. His approval rating among non-college whites specifically plummeted. We were tracking Obama-Trump voters; this was the only time when they came home in large numbers. And the reason is that a lot of these voters agree with Democrats on Obamacare, and they were very angry about attempts to repeal it. (And then they stopped being angry because it failed and everyone in politics has a short memory.)

Finally, if you look at 2016, I think people want to tell this story about values. But ultimately, if you look at what the biggest predictors of voter preferences were … there were just a lot of voters who agreed with us about health care but not about immigration. Obama got something like 60 percent of those voters; Clinton got roughly 40. And I think the reason why that happened was that both parties talked about immigration. So I think it does actually matter.

Speaking of Bernie Sanders, there was one more thing I wanted to ask about with regard to mobilization. Maybe turnout didn’t win this election for Joe Biden. But didn’t it almost win it for Donald Trump? As you said earlier, in parts of Wisconsin, it looked like new voters leaned Republican. Some on the left look at that and say, “Hey, look, a right-wing pseudo-populist was able to remake the electorate by rousing disaffected working-class voters with an anti-system, anti-elite message. So maybe a left-wing, anti-system candidate can do that as well.” And then, if this left-wing candidate had more message discipline than Trump, maybe he or she could finesse the gap between these disaffected non-college-educated voters and the Democratic base of cosmopolitan urbanites. Plus, as you’ve argued, polls tend to undersample voters with low social trust — a group that I believe Bernie Sanders tends to do better with than most other Democrats. So are we sure there isn’t a left-wing populist mobilization path out of progressives’ impasse?

There are some questions about Texas and Wisconsin. But I think, for the most part, the rise in turnout was pretty steady. The electorate was roughly as Democratic as people expected it would be. The reason why the polls were wrong, not to get too much into my other interview, was really because of who answered them. There wasn’t this surge of Trump voters who came to the polls. That doesn’t seem to be true. It was actually just that liberals were really excited to answer polls at higher rates.

If you want to make a case about Bernie Sanders, in his general-election polling before the primary ended, he really did seem to do better than Biden with younger, non-college-educated white people, even as he did a bit worse overall. It’s an interesting question of whether the polls were underrepresenting the non-college-educated group. And I do think, in 2016, Bernie had very, very strong general-election numbers. A lot of people say, “Oh, it was just that people hated Hillary.” But, at the time, in 2016, Bernie Sanders was one of the most popular politicians in America. And over the next four years, that stopped being true. He came into the 2020 primary with lower net favorables than Biden. And I think it’s really worth examining why. I always defined the Sanders model, in 2016, as having a lot of message discipline in terms of talking almost exclusively about economic issues and trying to frame all issues as a series of conflicts between good and evil. I think that kind of politics has a lot of appeal to these low social-trust voters, even if it turns off people who are highly politically engaged. And I think you could potentially build a very interesting coalition around that politics.

But we should learn from what happened over the next four years. Which is that, frankly, the Bernie campaign decided that the reason why they’d lost was that they were overly white and focused too much on economic issues, and that they had to become more woke on a variety of different social issues. And I think as they made this left-wing turn, while also really doubling down on policies that involve very large middle-class tax increases, as that happened, his support declined.

We’ve covered a lot of bleak trends for Democrats here. But one promising one that I’ve clung to — and built a lot of my punditry around — is that the millennial and zoomer generations still look like the least conservative America has ever seen, in terms of their attitudes, policy preferences, and voting behavior. But there’s still some who insist that these cohorts will age out of their radicalism and the generational “emerging Democratic majority” will prove as illusory as the racial one. So what do you think: By 2030, will the zoomers start to save us?

One thing I always try to remind myself is that in 2008, you could have looked at a Florida exit poll and it would have told you that Obama was losing people over the age of 55 and winning people under the age of 29, and the gap would’ve been very large, and you would’ve thought, All right, in 12 years, that’s probably going to be worth something. Now, here we are, and Florida is to the right of where it was before. And so I think the dialectic has a way of asserting itself. There are a lot of countervailing factors. But I will say that right now, there are a couple of structural forces that are helping us. The first thing is that the age gap is larger now than it’s ever been. Young voters are more Democratic now than ever before. Then there’s the question of how persistent that is going to be. And I think there’s some heartening news there. A fairly large fraction of the gap between older and younger voters on partisanship and ideologies can be explained by younger voters’ lower levels of religiosity and higher levels of education. And those gaps aren’t going away. So the fact that the age gap is partly rooted in these really fundamental, generational characteristics should give Democrats some hope.

There isn’t slam-dunk evidence on the life-cycle effect — that people, as they age, and get married, and have kids, tend to get more conservative. But I personally would be very surprised if that wasn’t true, just looking at the impact of those variables on vote choice currently.

But even if it is true, it’s not as big of a problem as it was a generation or two ago. One big, really underappreciated social change is that, relative to 20 years ago, the average age of first marriage has gone up by almost a decade. Which is insane! And at the same time, fertility rates have dropped tremendously. I’m not going to comment on whether or not that’s good or bad in a nonpolitical context. But I think the reality is that Democrats are now going to claim a larger part of the life cycle. We previously were going to start having people turn more conservative at 26. Now that’s been pushed up to 34 or 36. That’s actually very meaningful in terms of votes and in terms of how much longer we can expect millennials and zoomers to stay overwhelmingly Democratic. Also, zoomers actually seem to be incredibly liberal, even relative to older millennials, so far.

So there’s reason for hope. But if we, for example, lose 13 percent of our Hispanic voters, that can suddenly wipe out the impact of a lot of these expected changes. Some demographic trends are good for us. But plenty aren’t, like the long-term decline of the Black church or the rising share of Hispanics who are Evangelical. We also still haven’t bottomed out with rural white voters in the Midwest, who still don’t vote the way they do in the South, but could. So I think it’s very important not to bank on structural demographic change to work out in our favor. I think there really was an attitude after 2012 that the declining white share of the electorate meant that Democrats didn’t have to bother persuading people, and they didn’t have to engage in the discipline of running on things that are popular and not on things that aren’t. And I think that attitude directly contributed to Donald Trump becoming president.

Unless you’ve changed your mind on this since last we spoke, you identify as a socialist. And yet your views on electoral politics — that the Democratic Party is more deferential to the left than its own political interests would dictate, that it needs to disavow any interest in raising middle-class taxes to fund social benefits and must cater to the sensibilities of the median (middle-aged, white, culturally conservative) voter — are antithetical to the views of most U.S. socialists, at least as represented by DSA and on Twitter. And your electoral analysis seems to stem from a broader pessimism about the working class as an agent for radical change. So — assuming you still identify as a socialist — what does that identification mean to you? And do you think that your empirical analysis of how politics works today is compatible with a belief in the possibility of socialism’s realization at some distant future date?

I still identify as a socialist. I think there’s a lot of academic arguments that we can have. But like I said, when you look at all the popular things that are out there, there’s a lot we can do. And it’s true that not being able to pass large middle-class tax increases is a big constraint. It means we can’t re-create Sweden, definitely can’t re-create things past Sweden. But I think that pretending that these very real political constraints don’t exist does not accomplish anything. When you think through the optimization problem of, “How do we enact the most left-wing legislation possible without running over these trip wires that will make the public turn against us,” one part of it is that there are things that poll badly but are low salience. Blue states could create sectoral bargaining boards, which would enable workers to bargain over minimum wage and benefit standards across an entire industry. I don’t know if that idea polls above 50 percent. But whether it does or not, it’s not the kind of thing that we lose elections over. No matter how mad capital was, it would be a hard thing to drum up opposition to, as long as it was implemented correctly. And then there are popular left-wing ideas. Like, as I’ve said already, mandating worker representation on corporate boards. One personal frustration of mine is that when Elizabeth Warren embraced that, and wide swathes of mainstream Democrats embraced it, they were met with a pretty lukewarm reception from the left.

And then there are also a lot of accounting gimmicks that are very promising. I will point out that we actually did finance a very large section of the ACA by nationalizing the entire student-loan industry. Not being able to raise middle-class taxes does not mean that capital gets everything that it wants. I’m sure there are banks that were very angry that we nationalized student loans. But ultimately, when the Democratic caucus had a choice between raising middle-class taxes or nationalizing an entire industry to close a budget gap, they did the latter and nobody blinked an eye. So I think there is a lot of room to maneuver within these trip wires. And if we could actually implement all of the things that are popular, it would be a pretty transformational change.

And I guess the other thing I’ll say, in the theory camp, is just that I think it is very important to grapple with the reality that a bunch of incredibly smart people in a bunch of Western European countries tried to go through the electoralist route of creating social-democratic parties, and trying to implement their ideas, and it’s really important to take seriously what happened. I think some people like to let their socialist theory end in 1918. But it’s actually very important to look at, like, what the Austrian social democrats in the 1970s went through and what kind of hard coalitional choices they had to make in order to actually build a modern welfare state.

So I think when you go through and take seriously some of the challenges that democratic socialists have faced, it’s very sobering. And you do learn that the U.S. is not the only country where it’s very hard to raise middle-class taxes. And that accounting gimmicks are important. Interest-group politics can’t be ignored. And you can’t just switch off cultural conservatism or pretend that it doesn’t exist.