In July 2018, the most widely respected analysts were decidedly uncertain whether the Democrats could retake the House. On July 6, Cook Political Report, for example, listed 180 seats as “solid” for Democrats, with 12 likely/lean and 3 “toss-up or worse.” If the Democrats won all of those and the 22 GOP-held seats described as “toss-ups” — they’d still be one seat short of a majority, at 217.
But on July 1, newcomer Rachel Bitecofer, assistant director of the Judy Ford Wason Center for Public Policy at Christopher Newport University in Newport News, Virginia, released her prediction of a 42-seat “blue wave,” while also citing the Arizona and Texas U.S. Senate races as “toss-ups.” Her startling prediction was numerically close to perfect; Democrats will end up with a gain of 40 or 41 seats, depending how the re-run in North Carolina’s 9th district turns out. (Democrat Kyrsten Sinema won the Arizona Senate race, in a major historical shift, and Beto O’Rourke came close in Texas.) Furthermore, she even strutted a little, writing on Nov. 2 that she hadn’t adjusted her seat count, but that “the last few months have been about filling in the blanks on which specific seats will flip.” Her resulting list of those was also close to perfect.
With a record like that, you’d think that Bitecofer’s explanation of what happened would have drawn universal attention and become common sense — but you’d be sadly mistaken. She’s barely beginning to get the recognition she deserves, and more troubling for the country, the outdated assumptions her model dispensed with continue to cloud the thinking of pundits and Democratic Party leadership alike. (Follow her on Twitter here.)
This hampers efforts to counter Donald Trump’s destructive impact on a daily basis, and spreads confusion about both Democratic prospects and strategy in the 2020 election prospects. Above all, the mistaken belief that Democrats won in 2018 by gaining Republican support (aka winning back “Trump voters”) fuels an illusory search for an ill-defined middle ground that could actually demobilize the Democratic leaners and voters who actually drove last year’s blue wave.
“I am concerned with the longer-term trends that provide the context for these short-term effects,” Abramowitz said of Bitecofer’s model. “She’s very interested in the impact of demographic change and growing racial and ethnic diversity on the party system, especially in contributing to growing partisan polarization and negative partisanship,” which are key themes of his work.
That’s not to say Democrats don’t have anything to worry about — especially in 2022, when the tailwinds of negative partisanship may well be blowing in the other direction. But worrying about the right things, rather than phantoms, is the first step toward deal with them. So I sat down to talk with Bitecofer, in hopes of bringing some sanity and perspective to what is already a bewildering election cycle. This interview has been edited for clarity and length.
When I first saw your 2018 midterm projection, I felt two things simultaneously: First, this really makes sense, and then a sense of dread. “Oh my God, is this just motivated reasoning on my part, that makes me think this is so smart?” I no longer have divided feelings, but I understand the doubts.
Well, I can tell you as a person creating the analysis, I have that same problem. Imagine being the person doing the analysis and having that problem: Is this actually what’s going on, or is this just wishful thinking?
I had an advantage since I had lived through the Virginia cycle. [That is, the state legislative races of 2017 that resulted in big wins for the Democrats.] In the Virginia cycle, I kept telling everybody else, “This can be a massive landslide. You guys should be investing in delegate races because you guys can probably pick up 15-20 of them.” And everyone was saying, “Oh no, that’s not how it works in Virginia politics, you’re just too new to Virginia.” So I already had the advantage of knowing that. But that doesn’t mean I wasn’t absolutely sick to my stomach on Election Day.
When made your initial prediction of a 42-seat wave, other analysts weren’t even sure there would be any blue wave at all, and everybody had toss-ups where you are saying these will flip or are likely to flip. You were proven right, but the common-sense explanation that Democrats won over moderate Republicans by campaigning on health care was very much at odds with your explanation. You had this very prescient insight, and then everyone else catches up, but they sort of drop your insight. So how did you know, and how is that explanation mistaken?
I’m really glad to hear you frame it that way. I haven’t heard it framed that way, even in my own brain, but you’re exactly right. I am way ahead of everybody, they finally catch up as we move into the final two months before Election Day — certainly that last month — and then the election happens and it happens exactly that way, and then they abandoned my explanation. Now I’m out there trying to fight to get the explanation accepted.
The explanation, of course, is that it was this giant turnout of core constituencies, that either are Democrats or favor Democrats — they’re independents who favor Democrats — and they have a huge turnout explosion. So it’s not the same pool of voters changing their minds and voting Democrat after voting Republican because of the issue of health care. It’s a whole different pool of voters.
They might have many reasons that they cite, and probably this is not the reason they would cite. But what made them enraged and show up is Trump Inc., the negative partisanship. I don’t know why Nancy Pelosi, the DCCC or many of these moderate members are convinced that moderate Republicans crossed over and voted for them. I have the data for some of these districts and the data tells a very different, very clear story: If Republicans voted in huge numbers, they voted for Republicans.
In your look back, you actually said that there was a Republican surge, but it wasn’t a match for the Democratic surge.
The truth of the matter is Democratic turnout, particularly in midterms, is so bad that with the giant surge the Democrats managed to put together, what they were able to do is come close to matching Republican turnout. Which is good, that’s a major victory. But in many districts, especially where the candidates were focusing on being moderate, the Democratic turnout still underperformed its potential, and still underperformed turnout among Republicans, according to this analysis that I’ll be releasing after Labor Day.
But the turnout surge among independents — new independents, not ones who voted in 2014 — was so large combined with that turnout surge of Democrats to flip the district. If Republicans had also voted for Democrats, then these margins by which Democrats won would be much larger.
What factors predicted the turnout that allowed your model to be so accurate?
This model was built in the wake of the 2016 election. It was built as a response, answering why the models were off in 2016, and it was based on this belief that polarization has changed the behavior of voters. It has decreased the efficacy of things like the economy and increased the predictive power of things like partisanship. So the main predictor in my model is partisanship, the party competition of the state or the district, if were talking about my 2018 model, and that really sets the frame of the debate.
It has to be within a realm of possibility. To illustrate that, in 2018 when we look at the Senate map, Arizona was an R+5 state — that meant it had a partisan advantage for Republicans of about five points. Tennessee was an R+15 state, and of course, as I predicted, Arizona flipped to the Democrats but Tennessee didn’t even come close, despite multiple millions of dollars spent by Democrats in that effort.
The second thing that’s most significant and influential — and this is totally unique and new — is the percentage of college-educated residents residing in the state or in the district. This is what I said would be unique and new in the Trump era. This was why I was able to look at maps, at these races that in 2018 people saw as toss-ups, and say, “No, no, no, no. These are going to flip.”
In this new era that we’re moving into, we have college-educated voters moving towards the Democrats, and white working-class voters moving away from them. That allowed me to look months and months ahead at polling and say, these are the races where Democrats are going to do really well.
Circling back to something you already touched on, you argue that the “midterm effect” is misunderstood and that the old model — the movement of independents’ support from party to party — was a mistaken way of thinking about it. What sort of evidence was there for this prior to 2016, and then afterward?
We have thought of the midterm effect in this way: You have the president, you’ve got two years of performance. For a long time we’ve had a chunk of the electorate that’s partisan. Those conditions have just gotten worse, especially the last 20 years in what we call hyper-partisanship. But then we have these independents. The independents must be the referees, basically.
In this theory, they are free of partisanship, they’re able to look at a person and see them as they are without these blinders of partisanship and judge their actual abilities and make a referendum upon their performance in office. So there’s this idea that in the midterms they look at the incumbent party. And if they’re not doing great, like, say, Obama overreached with Obamacare, there’s this giant backlash of independents and they have this revolution because, God help us all, Obama modestly tweaked health care!
What I’m saying is no, the Democratic base collapsed, basically, from what it looked like in 2006, which was the last midterm when they had been out of power [and won]. It just literally collapsed among Democrats and because of that, it looks like the electorate has this giant recoil effect, driven primarily by these independent voters who are rejecting the party in power. But what we’re really looking at in much of these elections are the surge and decline of certain voters entering and leaving the electorate.
Could you going into a little more detail about how recent election results have advanced or refined your model?
The way it helped refine the model is — refer back to what happened in 2010, when Democratic participation just collapsed once the anti-Bush sentiment disappeared. Before that, the 2002 midterm was completely different because of 9/11, so we only had 2006 to look at, and we had this huge wave. Democrats flipped 35 seats in the House, Nancy Pelosi became the speaker, and then they doubled down that with Obama’s victory in 2008, and then they had this turnout collapse.
So I expected, even though Trump and Republicans are just better at voting, and especially Trump with his messaging — people sneer at it, but it touches people’s base emotions. So I expected it would be pretty good at getting people to vote, though I expected some kind of decline in turnout among Republican voters.
If that happened, then we would’ve seen participation rates in the 40s, and we probably would have seen a better night in the Senate for Democrats, probably very similar to what we saw in the House. But the reason that didn’t happen is that Republican turnout surged just as much as Democratic turnout surged, and that was something I did not expect to happen.
It certainly did not happen in Virginia in 2017, and that taught me that negative partisanship can be artificially created with very good strategy, and the Republicans have that. Some people sneered at it. You’ll even hear really smart GOP analysts, political pundits, say “Oh, Trump’s caravan strategy was so stupid.” Well, the 40-seat loss was locked in, and it didn’t matter, but what he probably did was save a couple Senate seats. So my 2020 model has been refined to account for my belief that Republican turnout is going to be extremely high too.
As a corollary to that, would you say concerns that “We shouldn’t do X because Trump will exploit it to rile up his supporters” are off-base, because that’s going to happen anyway?
Absolutely. In fact, in the district-level analysis of the voter file in California and Virginia that I’ll be releasing after Labor Day, I have competitive districts in those states. The data shows the turnout surge and how much different the composition of the electorate was between 2014 and 2018. I’m also able to show that even in these districts where Democrats ran Blue Dog candidates who were as unobtrusive as possible — with, exactly as you stated, the goal of not riling up Trump voters — the turnout for Republican voters in those districts was huge.
In fact, not only did Democrats not get the benefit of not stirring up the Trump base — the Trump base was stirred up and showed up in huge numbers — but by not tapping into anti-Trump sentiment in their own campaign strategy, by not intentionally activating that Trump angst, they paid a price in terms of their own base turnout. It was depressed, compared to other districts.
This year you released your presidential prediction on July 1, showing the Democrats winning a bare majority of 278 electoral votes, just above the 270 needed to win. Trump is at 197, with four toss-up states. But Pennsylvania, Michigan and Wisconsin are not toss-ups in your model. You write that Trump is in trouble in the Midwest and that there’s a profound misunderstanding of how we won those states in 2016. What’s at the heart of this misunderstanding, and why does it matter for 2020?
When we hear the punditry talk about what happened in the Midwest, generally speaking they’ll say the Midwest swung toward Trump, right? Well, that’s only true in two states. In Iowa and Ohio, where Trump cracked 50%, you can genuinely say he won over voters in those states. The others — Michigan, Wisconsin and Pennsylvania — Trump won those states by carrying only a plurality of the vote, and I would categorize that as a win by default. He won somewhere around 46 or 47% of the vote.
The reason that neither candidate was above 50% was that a huge numbers of voters in those states cast what we call protest ballots, and in the polarized era, the average that goes to a protest ballot is about 1.5%. The famous Ralph Nader election, the spoiler in Florida in 2000, that was about 1.5%. In Wisconsin, 6.2% of voters cast protest ballots in 2016. That’s an extraordinary amount.
So when you understand the role that third parties made in Trump’s victory there in the Midwest, and his inability to crack 50%, you realize that it’s a path to victory which was very complex. It becomes heavily tied to the fact that he was up against not just a Democrat, but that particular Democrat the GOP had managed to cause a public opinion backlash toward. That’s certainly not going to be the case in the 2020 cycle.
So what about 2020. What does your model say?
My model for 2020 starts off with Democrats at 278 Electoral College votes, and that’s a problem for Trump, because of course you need 270 to win. It does that because of my model’s prediction, based on turnout and predicted vote share, that Pennsylvania and Michigan will slip back to the Democrats. I’m uncertain about Ohio, but even if Trump wins Ohio, he can’t win the other three Midwestern states. Then as you point out I have four tossup states: Arizona, North Carolina, Florida and Iowa. Even if he wins all four of them, the Democrats have already won the election — and the idea that he would win all four is pretty unlikely.
I will have a much better sense about this once we see the participation rates in the Democratic primary. But I think what we’re going to be looking at is Arizona, Texas, North Carolina and Georgia as states where Trump is forced to play defense to hold on. I think by the time we get into September [of 2020] — I don’t think we’re going to get to the point where Democrats are comfortable in the Midwest. I think we’ll see a full-bore campaign and spending press in the Midwest all the way through to Election Day. But I think coming into September and October, they’re also going to be spending resources in the Sun Belt and other states like that.
You say that if Joe Biden is the nominee, he needs to consider a running mate who will motivate base voters, and that if a progressive is the nominee, there could be a self-reinforcing dynamic of misguided questioning in the media. While neither of those things should cause Democrats to lose, they are concerning, and they reflect a lack of understanding of what your model and analysis provides.
The Democratic leadership — the way they’ve chosen to navigate the Trump impeachment stuff, and certainly the way they talk about their House victories and how to maintain their House majority, it tells me that they’re still living in a understanding of the data that is outdated. If you don’t understand how you won, and what the concurrent political data environment is telling you, that is concerning. So I do see a lot of evidence that Democrats don’t get this. I’m not sure why.
I do know there is an increasing voice within Democratic politics that is leaning toward seeing the environment through my lens, and we saw that play out in 2018, in the campaign strategies of Beto O’Rourke and Stacey Abrams. What they were able to do in terms of their contests in both those deep red states was absolutely remarkable, and it speaks to the efficacy of a turnout-based approach, a strategic approach.
In Texas and Georgia, O’Rourke and Abrams both carried the votes of independents, whereas in Missouri and Indiana, where [incumbent senators] Claire McCaskill and Joe Donnelly positioned themselves much more in the Blue Dog camp in terms of issues positions, both of them lost independents.
So you might think, “Why is that? If one group of candidates took more liberal issue positions, why did they win over independents?” It seems counterfactual, and the reason is what mattered was turnout. O’Rourke and Abrams carried independents because turnout surged, with different independents showing up to vote, motivated by the targeting strategy deployed by those campaigns, which were run under my suggested model rather than the old playbook that used to work back in the ’90s and ’80s.
Democrats remain very worried about their freshman class, based largely on the notion of a static electorate, rather than one that’s still realigning. But you recently released 2020 projections citing 18 Republican-held House seats as top targets, nine of them in Texas. What are you seeing here that others are overlooking, and what will it take to realize that potential?
I see a couple of things. No. 1, I want to point out that these are possible targets. It’s not my official 2020 House forecast, because I don’t include all house races. It’s just 18 targets for pickup opportunities. And you’re right, because the realignment is still in progress, conditions are improving for Democrats. My model identified many districts Democrats did not even weaponize as competitive because they are still operating under this assumption that what makes an election competitive is a candidate like Amy McGrath [who is running against Mitch McConnell in Kentucky].
I’m arguing, no, it’s conditions like the suburban areas with a lot of college-educated voters. No matter how great a candidate, if you put them in the wrong electoral condition, no matter how much money you throw at them, they’re not to be able to overcome those conditions.
So here’s the other thing that both parties’ leaders should come to understand. In these swing areas, your power time is limited, so you should probably use the power when you have it, because the idea that you’re going to hold it indefinitely is totally wrong. Under my research assumptions, under my model, Democrats win the White House in 2020, and then in 2022 they’re going to have a very tough electoral cycle, because turnout for Democrats will go back to normal.
And because Democrats have poor electoral strategy, they’re going to compound that problem, probably by not appealing to Democrats to get them to the polls. So no matter how moderate you keep your Blue Dogs’ legislation, they’re all going to get wiped out anyway. So use your power where you have it. No. 2, there are ways to keep them in office, but the ways they’re choosing are not the ways to do it.
Turning to the Democratic primary campaign, on Twitter you been repeatedly warning that name recognition is the primary thing the polls are measuring this far out. How can folks be more realistic about thinking about the 20 candidates in the race right now?
People reading your article, people following me on Twitter — we forget how atypical we are, even in terms of those who are going to vote in the Democratic primary. Primary voters are already more active and aware and more typically engaged than average Americans, but still not like us. They’re not paying attention at all, not watching the debates, they’re not reading news stories, they’re not on political Twitter, reading political news sites, but they will vote.
So multiple blue-check reporters on my feed have said, “Oh it’s interesting that Bernie Sanders voters choose Biden for their second choice.” No, that’s not interesting at all. That’s the only other person that voters have ever heard of. So the problem is the proliferation of analysis that’s happened that’s completely analyzing shit that’s totally wrong. It’s in the data, okay? But it’s being driven by this name ID problem.
I’ve said a million times that we are living in the golden age of data. We just have to understand its limitations. And the sad truth of the matter is that the only way I’ve been able to figure out for people to do that is by following me and reading my research, because there’s no one else talking about it.