the midterms

David Shor’s (Premature) Autopsy of the 2022 Midterm Elections

Photo: Jeff Swensen/Getty Images

The Democratic Party entered the 2022 midterms saddled with a historically unpopular president and an inflationary economy. And yet the party is exiting them as the election’s clear winner. In fact, Democrats put up one of the best midterm performances for an in-power party ever.

David Shor believes that he knows why. The founder of the Democratic data firm Blue Rose Research, Shor has gained prominence in recent years by evangelizing for “popularism,” a theory of electoral politics that emphasizes the importance of adopting poll-tested issue positions and exercising message discipline, among other things.

The debate over how to interpret the 2022 midterm is just beginning. Votes are still being tallied. But Shor believes that the completed results tell a coherent story. I spoke with him about the role that abortion played in Democrats’ success, the burdens that Donald Trump imposed on the GOP, and why he believes his party owes its success to persuasion rather than base mobilization.

What’s your nutshell summary of what happened in this midterm and why?
I want to preface by noting that it’s extremely early. But I’d say that the No. 1 most salient fact about this election is that Republican turnout was very strong relative to Democratic turnout. You can see this in a host of different data sources. Whether you’re looking at administrative data on early voting, or the AP VoteCast exit poll, or ecological regressions off of the county level results, it’s just really clear. It’s hard to get an exact number. But, back of the envelope, it looks like the electorate was about 2 percent more Republican than it was in 2020. Republicans literally outnumbered Democrats, according to the AP’s VoteCast. And yet Democrats still won.

And they won for a few reasons. First, Democrats won independent voters, which may be the first time that a party that controlled the presidency has won independents in a midterm since 2002. Second, they got a lot of self-identified Republicans to vote for them. And third, they did those things especially well in close races. The party’s overall share of the national vote is actually going to look fairly bad. It looks like we got roughly 48 percent of the vote. But that’s because Democratic incumbents in safe seats did much worse than those in close races.

In districts that the Cook Political Report rated as “likely” or “solid” or “safe” for the Democratic incumbent, Democrats’ share of the vote declined by 2.5 percent relative to 2020. In districts that were rated as “toss ups” or “lean Democratic,” however, our party’s vote share went down by only 0.4 percent compared to 2020.

I think that tells us a couple of things. It suggests that Democrats did a good job with resource allocation; we spent in the right races. But it also illustrates the power of message discipline. Democrats in competitive districts aired more ads than Democrats in safe ones. And they also were much more careful about which messages they amplified with those ads and which issues they chose to embrace.

Relatedly, before this year, ticket splitting had been steadily declining cycle after cycle …

By “ticket splitting,” you mean voters backing one party for president or governor and then the other party for lower offices?
Yeah. So 2018 had much less ticket splitting than 2014, which itself had much less ticket splitting than 2006. That decline is one of the most robust trends in American politics. It’s been going on since the 1980s at least. It’s a little early to say, but looking at governors’ races, it seems like ticket-splitting rates increased relative to 2018. It’ll take some time before we can really say that for sure when it comes to the House or Senate. But given the historical trend, it would actually be remarkable if ticket splitting didn’t decline, let alone if it went up.

The fact that Republican turnout outpaced Democratic turnout means that Democrats couldn’t have won without persuading swing voters. But that doesn’t mean that base mobilization was unimportant, right? Keeping your party’s turnout in a midterm only 2 percent lower than in the previous general election is actually pretty impressive, isn’t it? 
It’s definitely true that the drop-off from 2012 to 2014 was a lot larger than what we saw from 2020 to 2022. I think the bulk of that shift probably derives from the changing nature of our coalition, though. We do much better with college-educated whites than we used to, and much worse with non-college-educated whites. Even though that causes an enormous amount of structural problems, when it comes to midterms, it does have some turnout benefits, since college-educated white voters turn out for midterms more reliably.

Again, it’s still really early, but right now it looks like there was a lot more drop-off in turnout among Black Democrats than among white ones. If you look at turnout by county in New York, for example, you see that turnout rates in predominantly white counties outside New York City vary within a pretty narrow band. But then in predominantly nonwhite counties in Brooklyn and the Bronx, you see turnout fall off the chart completely.

But I think the key thing is the overall turnout environment in 2022 was actually quite similar to the turnout environment in Virginia last year. Yet we lost then and won this year. And I think that the difference between those two cycles is that this time we won independents by quite a large margin.

Some pundits have suggested that a large increase in youth turnout propelled the Democrats to victory. I take it you disagree?
If you look at county-level data, the single strongest predictor of how much turnout dropped from 2018 to 2022 was the proportion of voters that were under the age of 35. In other words, turnout in America’s oldest counties surged while turnout in America’s youngest counties declined. It’s just hard to square the idea of a surge in youth turnout with administrative early-vote data, county-level data, and exit polling all showing that the electorate was substantially more Republican than in 2020.

If the turnout environment was broadly similar to Virginia’s last year, what explains Democrats’ success? Inflation has only gotten worse since November 2021. And Joe Biden is even less popular now than he was then. 
I think that the No. 1 answer is Dobbs.

The Supreme Court case that overturned Roe v. Wade.
Yeah. Years ago, in our first interview, I talked a lot about this concept of issue ownership or party trust: There are some issues that voters reliably trust Democrats on, and some they trust Republicans on, across election cycles. So, if you ask people, “Which party do you trust more on abortion or health care or crime?” you see a pretty common pattern that’s stable, both over time and really throughout the Western world. Center-right parties reliably get higher marks on immigration and crime, the center-left on health care, etc.

Looking at those sorts of metrics, abortion used to be a relatively neutral issue for Democrats. We track party ownership of 33 different issues. And in 2020, abortion was middle of the pack. It wasn’t an issue that Democrats dominated. But the Dobbs decision changed that. Abortion suddenly became our second-best issue basically overnight.

At the same time, its salience massively increased. Before the Supreme Court’s decision in Dobbs leaked, voters in our data ranked abortion as the 30th-most-important issue out of 33. After the leak, it jumped to 12.

Doesn’t that development challenge your model of politics a bit? In the past, you’ve argued that it’s largely futile to try to change swing voters’ policy preferences, at least within the time frame of a campaign. Your critics on the left argue that your view is unduly fatalistic: Rather than changing their policy positions to appeal to the median voter, Democrats should try to reframe the policy debate in a manner that gives them an advantage. 

In this case, Democrats did not change their position on abortion. Indeed, to the frustration of many pragmatists, they declined to hold an “up or down” vote on codifying Roe, but instead pushed a maximalist bill that would leave fewer restrictions on abortion than the pre-Dobbs status quo. And yet, despite this absence of moderation, the politics of abortion changed overnight. Which might suggest that Democrats can rapidly remake popular opinion on an issue if they only find the right way to politicize it. 
I think it’s important to emphasize that what happened with abortion is extremely rare. It’s very rare for party ownership of an issue to shift this rapidly. And I think it really boils down to this concept of “thermostatic” public opinion.

So, the president’s party almost always does poorly in midterm elections. That’s a very consistent pattern going back to the 1930s. And we see a similar phenomenon overseas. In local elections in England, or regional elections in France and Germany, the party with national power tends to do less well.

I think the best explanation of this comes from a paper by Joseph Bafumi, which basically found that voters like to balance out policy change. They just have a very strong sense of status-quo bias and loss aversion. And as a result, they react negatively to dramatic changes in policy. So when policy moves left, they move right. And when it moves right, they move left. Just as when the temperature goes up outside, you move the thermostat down, and vice versa.

You can see this in polling of whether the government has a responsibility to provide universal health care. Support for universal health care went up during the Bush Administration, then down as Obama tried to push Obamacare, and then up again when Trump tried to repeal Obamacare.

So the theory is: The reason why the party that controls the presidency does poorly in midterms is that voters are trying to balance out policy change by creating divided government. And I think what’s really unique about this midterm cycle is that Republicans created a radical policy change — and one that was quite unpopular — without controlling the presidency or the legislature. And that allowed Democrats to plausibly run as the party that was going to make less change than the opposition, which is a super-unusual situation.

I think that conclusion is supported by several pieces of data. For one thing, in our ad testing, messaging about reproductive rights tested very well. For another, Dobbs had an immediate impact on election outcomes. If you look at special elections before and after the decision, Democrats did much better in the latter. And then also, the number of primaries held before the Dobbs decision was roughly equal to the number held after it. And the percentage of primary voters who were Democrats increased by about 2.7 percent after Dobbs.

So I think Dobbs was really the major factor. But there was no guarantee that Democrats would capitalize on the political opportunity Republicans gave them. They had to really press the issue. And if you look at the actual ads that candidates put out, or the speeches that they made, or what they were talking about on TV, their campaigns were very focused on abortion, very focused on prescription drugs, and a bunch of other popular issues that people actually care about.

I think some critics of “popularism” worry that it encourages Democrats to be more timid legislatively than they actually need to be; or at least more timid than they should be, given the substantive benefits of many reforms. And I feel like these election results strengthen their case. Given that Democrats had extremely narrow congressional majorities, they actually used power fairly aggressively. They enacted a $1.9 trillion COVID-relief bill, the largest climate bill in history by an order of magnitude, and canceled $10,000 of student debt for tens of millions of borrowers through a controversial exercise of executive power. And yet they nevertheless put together one of the strongest midterm performances for a presidential party in history. 

Certainly, Biden has enacted a bolder, more progressive agenda than Bill Clinton did between 1992 and 1994. And yet Biden’s first midterm went much better than Clinton’s. So it seems like it is possible to get voters to tolerate dramatic policy change.
So, if you had told me Biden’s current approval rating a year and a half ago, I wouldn’t have been surprised. Presidents tend to see their approval rating decline over their first years in office. But the path to Biden’s current approval rating has been much different than I would have thought. The classic story, if you look at Bill Clinton, or Donald Trump, or Barack Obama, is: A president comes into office with a trifecta, he tries to pass legislation, that legislation turns out to be very polarizing and unpopular, and then he becomes unpopular. And that isn’t what happened here. It’s true that a lot of the decline in the Democrats’ share of the vote in generic ballot polling happened in October 2021, when reconciliation debates were in the news. But that was also when gas prices started to spike.

All told, the policy backlash to the things that Joe Biden did was much smaller than under previous presidents. I think that reflects the fact that Biden really picked a policy agenda that was very economically focused, and that didn’t necessarily play into people’s fears of big government. The Affordable Care Act really did substantially change how roughly 20 percent of the American economy worked. And there were lots of people who were really worried about how changes to health-insurance laws would affect them personally. I think Biden’s policies were kind of consciously designed to avoid triggering loss aversion.

Another difference was that, unlike Obamacare, the final reconciliation bill was kind of a phantom package; it didn’t really exist until moments before it passed. So it was hard for Republicans to attack any specific policy, because they didn’t actually know what was going to be in the bill and what wasn’t. And I do think it really probably does matter that the things that ended up being in the reconciliation bill were all pretty easy to administer, and also dealt with issue areas where people trust Democrats.

Although I do think there may have been one exception in terms of thermostatic backlash to policy change. The American Rescue Plan’s expansion of the Child Tax Credit did seem to trigger a backlash. That was probably the single policy I was most excited about. And both Democrats and Republicans expected it to be a political winner. But as the money actually went out, it really did become a lot less popular. When you looked at polling for whether or not the enhanced CTC should be extended, it was something like 60 percent against.

But I think it’s important to understand why voters opposed it. There’s this really strong sense within the Democratic Party that the CTC was unpopular because it was giving money to poor people. Which I think is a holdover idea from debates about welfare reform in the 1990s.

When we actually tested arguments for and against the CTC in our polls, it was super-clear that people didn’t object to the law covering poor people, but really opposed sending money to middle-class and upper-middle-class people. I think there is actually a lot of sympathy for the poor in the American electorate. If you look at Gallup polling, going back to the ’70s, voters’ views on whether we should have “a bigger government that does more with higher taxes” or a “smaller government that does less with lower taxes,” it moves around a lot thermostatically. But at the moment, most people disagree with this social-democratic vision that a middle-class person should be getting help from the government.

Well, they might disagree with that as an abstract ideological premise. But in practice, the U.S. directs a ton of social-welfare benefits to the middle class. And voters jealously guard those benefits. Like, it would be political suicide to end the mortgage-interest deduction and reallocate all the savings to housing subsidies for the poor. And voters have little tolerance for cuts to Medicare or Social Security.
Sure. But one of the best messages that we’ve ever tested is the talking point “Social Security is not an entitlement: People pay into it and they’re just getting what they deserve.” There’s this really strong sense that middle-class entitlement benefits were earned. So people don’t see Social Security benefits the same way they see the Child Tax Credit.

If you look at party trust, there’s this very frustrating dynamic where people often trust Democrats on specific issue areas, like education or health care, but then agree with Republicans on the abstract question of the overall size of the federal government. And sending checks to people who make $150,000 a year really seems to trigger voters’ conservative intuitions about the role of government in a way that climate tax credits don’t.

But anyway. I think this is actually good news. People think that giving benefits to poor people is unpopular. But it’s not. Even many Republican voters have sympathy for the poor. So the next time we find ourselves with a trifecta, I think there’s a lot we can do to alleviate poverty without triggering a backlash. Because the conventional wisdom around the politics of means testing is wrong.

One place where Democrats didn’t fare well Tuesday night was Florida. Ron DeSantis and Marco Rubio absolutely routed their Democratic challengers in Miami-Dade County, which is a 70 percent Hispanic area that voted for Hillary Clinton by 30 points in 2016. Do you have a sense yet of whether Republicans made similar gains with Latino voters in other parts of the country, or if Florida’s Hispanic community is on its own distinct trajectory?
The overall story with Latinos in 2020 was that we lost nine points of support from them nationwide, about 14 percent in Florida, and then something like 30 percent in parts of the Rio Grande Valley. Thirty percent is almost unheard of as far as drop-offs in support go. Whenever you make any kind of scatter plot of county-level results, those three or four counties in southern Texas are almost literally off the plot. So, relative to 2020, I think those counties did swing something like 10 percent toward Beto. Which is relatively common in state-level races. And I think it reflects the fact that he’s from there and campaigned a lot there.

But with the exception of Beto regaining ground in the Rio Grande Valley, I think there’s a pretty clear story nationwide. We’re still looking at precinct-level data. But if you look at county-level regressions, it really does seem like support continued to fall in relative terms in Hispanic areas. Certainly support was down relative to 2016. But it’s possible that it was down relative to 2020. Although I would make the caveat that, at this point, it’s difficult to distinguish lower Hispanic support from lower Hispanic turnout. If in heavily Hispanic counties Hispanic turnout declined while white turnout went up, then we might see Democratic support falling in those counties, even if Hispanic voters shifted left. But if you look at the results in Virginia last year, and in special elections earlier this year, I think it’s all consistent with the idea that a significant percentage of Latino voters durably realigned in 2020.

Earlier, you mentioned that ticket splitting may have increased this year. I feel like that sits in tension with one of your core arguments in our past interviews: That Democrats really need to exercise message discipline, even if they represent deep-blue areas, because politics is now thoroughly nationalized. Which is to say Democratic candidates in purple-to-red areas have a much harder time distancing themselves from their party’s national brand these days. And so it is really important to moderate the Democrats’ national image and deny Fox News any opportunity to paint the party as radical by, say, amplifying the most radical ideas that Alexandria Ocasio-Cortez ever promoted. 

I don’t think Democrats did much to moderate their national brand since Biden’s election. And yet Democratic candidates in several states proved capable of winning lots of Republican votes by cultivating idiosyncratic political profiles. Kansas reelected a Democratic governor. Josh Shapiro and Gretchen Whitmer won lots of Republican votes in Michigan and Pennsylvania. And John Fetterman won by a larger margin in Pennsylvania than Joe Biden did in 2020, which was a more favorable year for Democrats. And Fetterman did so while suffering from disabilities that severely compromised his public-speaking skills. This despite the fact that Fetterman associated himself with Bernie Sanders, Medicare for All, and lenient criminal sentencing, among other left-wing positions.
I think it’s important to remember that 2018 featured the lowest levels of ticket splitting in a midterm ever, and by a large margin. The fact that things may have gone up a bit is heartening. But ticket splitting was still lower in 2022 than it was in 2014. We’re talking about a pretty small uptick, if there was one. So it doesn’t change the structural challenges that Democrats face. It’s still the case that people are judging candidates by their party’s national brand to a historically great extent. And it’s also the case that Democrats need to win in states that are more right wing than the country as a whole in order to compete in the Senate and Electoral College.

As for ideological positioning and message discipline, if you really look at what these candidates actually did, Fetterman took a lot of care to distance himself from a lot of unpopular ideas, including defunding the police.

He really focused on economic issues and reproductive rights. And, as I said earlier, the Democrats who won in close races all generally did that. And I think the importance of their message discipline is illustrated by the fact that they did so much better than Democratic incumbents in safe districts. So I really do see all of this as a victory for the idea that disciplined campaigns that try hard to appeal to the median voter can succeed. The key thing here is that these candidates won by persuading independents. This wasn’t a mobilization story.

A lot of Republican operatives are arguing that Donald Trump almost single-handedly cost them a red wave midterm. I think that clearly understates the role that the Supreme Court played. But how much do you think Trump hurt his party, whether through his endorsements in GOP primaries or just tainting the Republican brand?   
I think it’s obviously true that Trump’s endorsements made it harder for Republicans to win some key Senate and gubernatorial races. It looks like Fetterman is going to win by a comfortable margin. But in polling taken at the time of the Pennsylvania primaries, he was polling considerably below David McCormick. It’s hard to find a single instance where Trump didn’t endorse the less electable candidate.

Also, though Trump’s authoritarianism and election denial weren’t very salient to swing voters, I think they still helped Democrats. Elite concern about the GOP’s threat to democracy helped Democratic gubernatorial candidates outraise their opponents. And it also made it harder for Republican election deniers to raise money.

But again, I really just want to emphasize the importance of Dobbs. If you look at public polling, or special-election results, or primary turnout rates, every midterm indicator pointed to a red wave before the Supreme Court’s decisions. And that changed afterward. I really don’t think that was a coincidence.

Before this year’s midterm, you painted a pretty dire picture of where American politics was heading. Given rising levels of education polarization, declining rates of ticket splitting, a growing pro-Republican bias in the Electoral College, and the really unfavorable Senate map that Democrats will face in 2024, you suggested that the conservative movement had excellent chance of securing an unprecedented power by mid-decade. Specifically, if the GOP won in 2024, then Donald Trump, Ron DeSantis, or whomever they nominated might command more federal power than any U.S. president since FDR. Are you much less troubled by that prospect now than before Tuesday’s results?
The fact that Democrats just had a historically strong midterm performance has probably eliminated the very-worst-case scenarios. If this had been a typical midterm, we would’ve lost four or five Senate seats. A lot of this is luck. Even a very small shift toward Republicans, and we would have lost two or three seats. And we were also lucky that this year’s Senate map was favorable for Democrats.

But because we did so well, Republicans no longer have a path to a filibusterproof Senate majority in 2025. That’s something I’m incredibly happy about.

Nevertheless, in 2024, we have eight senators facing reelection in states that are more Republican than the nation as a whole. And we’ll also need to defend states like Montana, West Virginia, and Ohio, where, even in this cycle, no Democrats came close to winning. So we’re at a high risk of losing six or seven seats. The fact that ticket splitting didn’t decline gives me some hope that we can avert that outcome. But a lot can change between now and then.

One of the big open questions is what the Republican Party decides to do in response to this. Even though Trump has created a coalition that gives Republicans an enormous structural advantage, it’s still true that he is one of the most unpopular politicians in America. So a lot depends on who the GOP nominates in 2024.

Trump and DeSantis have both staked out more moderate positions on abortion than almost every other major politician in their party. And it’ll be really interesting to see the extent to which Republicans decide to react to this by moderating on abortion.

It’s important to remember that the parties have agency. The electorate is substantially more educated, less white, and much more secular today than it was ten years ago. And yet, despite that, the Democrats got 52 percent of the vote in 2012 and 52.3 percent of the vote in 2020. And the reason for that is basically that the parties shifted their positioning. As the electorate moved to the left, Democrats did too. And the net effect ended up being the same.

Going back to Florida: Since 2000, the nonwhite share of that state’s electorate has almost doubled. I’m sure that the college-educated share has increased quite a bit, too. Yet Florida has gone from being a literally tied state to now being quite Republican. Demographics didn’t do that. Decisions made by each party did. So, moving forward, a lot depends on what Republicans and Democrats decide to do.

We were gifted an electoral opportunity by the Dobbs decision. But the reason why we capitalized on it was because candidates in swing races showed enormous message discipline, both in their external comms and in paid advertisements. If 2024 turns out to be less apocalyptic than statistical models suggest, it will be because Republicans did not respond to the information that voters just gave them (by moderating on abortion), while Democrats continued doing what they just did.

David Shor’s (Premature) Autopsy of the 2022 Midterms