Democrats did far better than expected in the 2022 election, especially in swing states and districts. Do the early results provide new data on the effects of campaign messaging and careful candidates? Can Democrats overcome structural disadvantages by campaigning on popular issues? David Shor argues that campaign effects, especially from moderate popular candidates, help explain the outcome, whereas appeals to turnout and changes in electoral composition do not. The conversation also covers what can and cannot be learned with immediate data and how to use both political science and practitioner data to understand the effects of campaign strategy.
Guest: David Shor, Blue Rose Research
Transcript
Matt Grossmann: Does the 2022 election show how democratic campaigns win, this week On the Science of Politics. For the Niskanen Center, I’m Matt Grossmann. Democrats did far better than normal for the party of the presidency in the 2022 midterm election, especially in swing states and districts. What do the early results say about the impact of turnout, electoral coalitions and campaign messaging, and what can be learned about how Democrats might overcome their structural disadvantages? On this special election edition, I talk with David Shor of Blue Rose Research about his initial analyses of election results, his studies of campaign effects and his tracking of democratic strategy. Shor argues that campaign effects, especially from moderate popular candidates and policy positioning, are underestimated, while changes in the composition of the electorate are overestimated. He works at the intersection of political science and practice, and I think you’ll enjoy our conversation. So in 2020 it was pretty immediately clear that Democrats had gained votes with white college educated voters and lost votes with Hispanics. It seems a little more stable this year. What’s your take and what can we learn from that?
David Shor: Yeah, I think it’s still early to say. One of the annoying things about looking at… The most complete data that we have is county level data, and doing ecological regression off of county level data is always terrible just because the most diverse counties are just these large urban counties. We’ve managed to piece together precinct data in a decent chunk of the country. It’s still probably something like 10% of the population. And so far, it really seems like things have been pretty stable. There’s been pretty uniform swing by race as far as I can tell. Really the only evidence, now to be clear, that’s uniform swing from 2020. And in 2020 there was this massive realignment on racial lines, and it does seem like that realignment is here to stay. The only place where there really was much reversion at all was a couple of border counties in South Texas, which swung dramatically more than anywhere else in the country.
There were parts of South Texas that swung 30 percentage points from 2016 to 2020, and it looks like those areas had something like a 10% bounce back in the gubernatorial race. Though to be honest, you would expect that. Generally speaking, in the industry, the thing with off year elections or gubernatorial elections is they usually are kind of best predicted by a weighted average of the previous two presidential elections. And so from that perspective, the fact that we haven’t seen bounce back in most of the country does kind of imply that in a sense non-white voters continue to trend against us.
Matt Grossmann: Just it on the education polarization side, it seems like if anything it remained at least stable, if not further increased, and yet democratic votes were sort of more well distributed.
David Shor: Yeah, on the previous point, I do want to say that it does seem like there were large gaps in turnout, at least defined in terms of relative turnout from 2020. It’s a little too early to say relative to 2018, which is probably a better comparison. But yeah, just to answer your question, I think that the most interesting part of this election, the thing that would’ve been most surprising to me a year ago, is that if you look at democratic incumbents who were in swing races, swing house races, swing senate races, swing gubernatorial races, they really outperformed quite a bit more than the rest of the country. Just to put some numbers on that, if you look at Democratic house incumbents in swing districts just defined by Cook, a toss up or lean, they saw about a half a point drop off under performance from Biden’s 2020 democratic two way. Incumbents outside of those swing districts saw something like a three and a half percentage point decline.
And so that’s what I find very interesting about this is that we might end up with a popular vote of something like 48.5%, but we’re going to really outperform. Political scientists talk a lot about the vote C curve. And from what I can tell, Democrats are going to get something like eight or nine more seats than you would’ve expected just from uniform swing. Because it really seems like there was a red wave everywhere in the country except for the places that mattered. And I feel like I follow elections a lot. I see C curves a lot both here in other countries, and I’ve never seen anything like this. And so I think that that’s definitely the most interesting story of this cycle.
Matt Grossmann: So how did that happen, given that the education polarization didn’t seem to decline? So democratic votes are still mostly coming from the same kinds of places. But as you say, we could get a fairly normal vote shift from the presidential year to the midterm year. Democrats won the national House popular vote by three points or so last time. They could lose it by that amount this time, and yet they were so much better distributed. Why is that?
David Shor: Yeah, I think what’s interesting is this is really about the races. If you look, there are clear media market discontinuities. I think one of the interesting things is if you look in Pennsylvania, there are parts of Pennsylvania that are covered by New York media markets instead of Pennsylvania ones. And Fetterman really did much worse, as did Shapiro. And it’s hard to distangle exactly how much of that is advertising versus how much of that is local media environment. But I think that there really is this story that Democrats in swing districts and in swing races really acted very differently than Democrats in the rest of the country. They had a different media environment, they had very large spending advantages. I think that that’s one of the big underrated stories about this. If you look at the four closest Senate races, Democrats outspent Republicans massively over the course of the election cycle in a way that wasn’t true nationwide or true in non-swing races.
The volume of ads, I think was pretty unprecedented for a midterm. And that was absolutely true in governor’s races and also secretaries of state races. I think this gets to some of the positive sides of education polarization, which is something I talk about a lot, is that democracy doesn’t resonate very much with voters. And we see this very clearly in ad tests. And that’s one of the reasons why less than 1% of ads that were rare aired in Senate races this cycle were about January 6th, they’re about democracy. But donors, donors really care quite a bit about democracy. There was an unprecedented effort by the donor community to pump money into governor’s races and secretary of state races. And I think the results show for themselves. That’s what I find very interesting about this story is ticket splitting by some measures went up from 2018 to 2022. We need to until all the data’s in to know that for sure.
And it really seems like we have kind of a return a little bit to an older politics of campaign effects and ticket splitting. And I think that, just to go back to the media market discontinuities, I think it does highlight, maybe not just ads because I’m sure that there were digital ads in those places or not, but also really different com strategies. People talk about Fetterman for example. And one of the things that both Fetterman and Shapiro really were very methodical about is that they made an enormous amount of local TV appearances. They spent a lot of time trying to do external facing coms. And so I think that we kind have a two party story where there’s the party of Jared Golden, and then there’s the party of blue state Democrats who are optimizing on different objectives. And it’s interesting to see. I wouldn’t have expected it a year ago.
Matt Grossmann: So in a normal midterm, voters balance the party of the presidency usually through some combination of a relative turnout advantage, and swing voters moving against the party in power. So this year, was there less of a backlash to democratic policy direction or was there just a near equivalent backlash against the Republican policy direction under Dobbs?
David Shor: Yeah, again, nationally, this looks like a very normal midterm. Again, they’re all rough numbers and they may change as we get more data, but I think from what I can tell, it looks like the turnout environment for Republicans turned out at higher rates than Democrats did. And it looks like there was something like a four percentage point shift from 2020 to 2022. Obviously, it all depends on how your baselines. And I think probably about half of that was out and about half of that was people changing their minds.
And so again, that’s a very normal thing. In 2018, it was something like a 75/25 split for persuasion. In 2010, it was really 25/75 turnout in the other direction. But this is all pretty expected. And so I think that this is really a story of democratic incumbents being able to preserve their incumbency advantage against a bad environment. I think if you look at the turnout electorate this cycle, I think it’s going to really resemble the Virginia turnout electorate in a lot of ways with large drops and non-white turnout with Republican mobilization. And Democrats were still able to win anyway, which again is incredible.
Matt Grossmann: But democratic ads were hyper focused on abortion, as was as democratic messaging. And so it does seem like there was a bit an effort to take advantage of a backlash to the abortion decision. That also seems like it might be a one off, sort of hard to repeat thing for future elections. So what should we take from that focus?
David Shor: Well, obviously there’s a thing, the effects of… It’s hard to know the long run effects of policy positioning. On one hand, you have one set of literature focused on ads that tell you that effects decay and that there’s mean reversion in politics. And you have, I think another set of literature, if you look at spatial voting models and all of that, it suggests that this kind of thing could be very durable. I think less quantitatively, something I think about a lot is that Democrats were able to run against Republicans for a very long time on the idea that they want to cut social security and Medicare. And I think that those attacks were effective really right up until the Republican party loudly repudiated Medicare and social security cuts in 2016. I think the big question in my mind, I mean, I say this a lot, but if you look at the electorate and the median voter since 2012, in the past 10 years, the electorate has gotten much more educated and much more secular and much less white.
And yet, despite that, Barack Obama got 52% of the two party vote and Joe Biden got 52.3% of the two party vote. And the reason for that is that the parties changed their positioning due to internal demands from stakeholders and within the party. And I think at this point, the saliency of Republican demands on abortion have really increased a lot. And some of that is from the Dobbs decision, but I think some of that is just because they actually have to defend these bans. I think that if Republicans don’t change their positioning and continue to make support for these bans necessary, I think that will continue realignment on these lines. The Republican position on abortion right now is extremely unpopular. I think it’s as unpopular or more unpopular than defunding the police. And if they continue to make that a big part of their agenda and don’t distance themselves from it in the way that Democrats have, I think that will continue to cause damage.
I think one data point on this that I find super interesting is the single most effective ad category that we tested this cycle was, on the Democratic side, where all of the ads, and there were a bunch of them by democratic incumbents where they went and they said, I don’t support defunding the police. And that was actually very, or not, even though the defund debacle was two years ago, those ads still did a lot. And I think they were something like, I think 1.7 or 1.8 times more effective than the average ad. And so I do think that if a lot of this just kind depends on what the parties do, which is one of the most boring answers.
Matt Grossmann: So there was a story that there was sort of a long term shift in relative emphasis to cultural issues relative to economic issues. And we’ve seen an international pattern where that’s associated with educational polarization that might be bad for, especially geographically, for the Democrats in the US. Some would say this was an example of an election where the Dems were able to win on cultural issues, so maybe that story is overplayed. But you now seem to be saying that you think there might be, if Republicans continue moving rightward on maybe this issue or transgender issues or other things, schools, that there might be an opening to win on cultural issues going forward.
David Shor: Yeah, I mean, I think that you have to disentangle. There’s the structural bias of the American electoral system, which overweights the views of people who live in rural areas and underweights the views of people who live in urban areas. And that really does meaningfully put a damper on a lot of things that otherwise would be theoretically popular, like I don’t know, various kinds of gun control.
I think, to be clear, a lot of gun control is less popular than people think. But it’s true that there are some versions of it that are popular with the median voter, but unfortunately not popular with the median voter in the median Senate state.
But the Republican position on abortion isn’t like that. Abortion bans have electorally failed at the ballot in places like Kansas and Montana and Kentucky. This is actually a big losing issue for them. I think an underrated story of 2016 and 2020 was that the Republican Party really moderated on a whole host of issues. They passed these large multi-trillion dollar spending bills. The one time that they did try to do a healthcare bill, they did pay an electoral price for it, but they ruled out the Ryan Care Plan. Even the ACH plan that they tried to pass, I think was much more moderate than I personally would have guessed going into that administration, though, Matt, I think you’re a lot more calibrated on these things than most people, so maybe you weren’t surprised.
And I think in the aftermath of 2020, I think Republicans… And I think this might be common. The time to go to the ideological extremes is when you are the out party, since you already have the wind at your back. And it seems like Republicans were a little overconfident. They loudly embraced a very high salience on popular issue. And I think you could really see.
I mean, I know this sounds like a Just So story, but you can really see the effects. The polling immediately changed. And polls suck these days, so there’s some other indicators that also changed. If you look at the Special Elections Index, which is basically in-
Matt Grossmann: You’ve got pre and post ops. They changed.
David Shor: Yeah, there’s pre and post ops. There was an enormous shift on every electoral indicator, whether it was polling, whether it was special election performance, whether it was primary vote share. There were really these very large changes. And I think that the most straightforward explanation for that is that voters punished the Republican Party for doing something extreme.
Now, voters have really short memories. If the Republican Party stops talking about this stuff and makes this a less salient part of their platform, then they can recover and do well. And I think, to be clear, that’s not an academic thing. Both Ron DeSantis and Donald Trump both have reasonable… I mean, not reasonable, but by the standards of their party, fairly moderate stances on abortion, 15 weeks and whatever. Florida does not have a total abortion ban. So it is possible that they can continue doing that, and they’ll be able to sustain activist pressure.
I wish I had a good intra-GOP model of politics. I don’t know. I know on the Democratic side, it’ll be interesting to see if they have the same pressure to move to the extremes that Democrats did in the run-up to 2020. I don’t know. But I think that’s the big question. But the flip-side of that is that the national environment will probably improve for Democrats between now and 2024, as it has for almost every party between the midterm and the general.
Matt Grossmann: So Republicans heavily focus their advertising campaign on inflation and the budget and crime. Democrats mostly responded on crime more than keeping up with Republican ads on the economy and inflation. Evaluate the Republican ad strategy and the Democratic response to it.
David Shor:
I think that the Democrats in swing races really showed an enormous amount of discipline when it came to the ads that they put on. I do think that Dobbs created a giant opportunity for them, but they did actually have to go and exploit it. And I think on the Democratic side, when we look at our testing, Democrats put most of their ad spend toward issues that were much more effective on average: the economy, abortion, healthcare, social security. I’d have to look at the comparative data, but I think that there was a boomlet in social-security-related advertisements this cycle. It used to be a mainstay for Democrats running, but I think it’s come back. I like to joke that sometimes [inaudible 00:19:45] play hits.
On the other hand, I think that Republicans did run a lot of ads on crime and the economy and inflation, and those were good issues for them. But I think that probably the biggest low-hanging fruit that they didn’t pick is that these incumbents, had they tried to distance themselves from the party on issues like abortion, which successful Republican incumbents like Susan Collins did in 2020, I think they would have done better. We’ve done some preliminary ad-tagging, and we saw really very little of that from Republican incumbents this cycle versus last cycle.
Matt Grossmann: So as you said, there was more split-ticket voting, at least between governor and Senate races this year. And curiously, in those last New York Times polls, Democratic Senate candidates were outperforming, by an important margin, people who wanted to see Democratic control of the Senate. In other words, there were some voters who would rather have had Republican Senate control, but were still voting for Democratic Senate candidates. So how did they distinguish themselves from the National Party?
David Shor: Yeah, this is the most important part, I think, of understanding this election, is that there were more Republicans. In this electorate, more people identified as Republican than identified as Democrats. But Democrats won anyway, because they convinced those Republicans to vote for them. And I think this is a common… When I talk about how Democrats have this big structural advantage, the literal implication of that is that the only way that Democrats can win close elections is to win a large share of people who want Republicans to control the Senate. And this was the struggle in 2020, when we had to win a bunch of fights in our plus-two seats.
I think that this year, luckily, there was incumbency at their back, but I think the really big story for how they did it… I mean, there’s a couple of different pieces. They had pretty low-quality candidates. If Donald Trump hadn’t endorsed Herschel Walker or [inaudible 00:21:57], I think the Republicans probably would have taken the Senate.
And again, Democrats had massive spending advantages in basically all of these races, I think again like 2020 to be honest. And on top of that, I just think that Democratic Senate candidates ran a lot of good ads and were able to distinguish themselves. I mean, this just goes hand in hand with the ticket-splitting not declining, is that these incumbents were able to successfully distance themselves from the party. I think there is probably a little bit of a dynamic. I think obviously this has to be Dobbs-related. Dobbs wasn’t really just a uniform swing.
I think more speculatively, when you think about this from the perspective of being a campaign operative, normally the pattern for how midterms go is that you have these incumbents and swing races or in Red states, and they start off with these really big leads. But the partisanship of the electorate is actually not very good, and the reason is they’re running against these opponents who are barely well known, and people just tend to vote for folks when they don’t know who the other person is. And then what happens in the summer is that these Democratic incumbents are just pummeled with negative ads, bad-news cycles and scandals, and then things converged to partisanship and then a lot of them lose.
I think that Dobbs, just by timing, managed to disrupt a lot of that. If you just think about when it happened, it was the end of June, and it really led to this new cycle where the new cycle in Arizona was, what does Blake Masters think about abortion, and why did it change from three weeks ago, as opposed to how liberal Mark Kelly supposedly would be? And so I think it really did provide an opportunity for Democratic incumbents to define themselves, and who knows how replicable that is, unfortunately.
Matt Grossmann: So Democrats substantially underperformed in New York and Florida, which happen to be your current home state and your original home state.
David Shor: Yes.
Matt Grossmann: Anything specific about the dynamics in those states, or just more generally, why did it seem to be the case that we have state-specific effects this year? Colorado and Michigan seem to be going more in the Democratic direction, and New York and Florida less so.
David Shor: Yeah, I mean, again, this is more speculative. How many states are there? But I do think that, again, this is an example of how Dobbs really played some space for ticket-splitting in that it really did create some material stakes to government, where if you look at places like New York or Oregon, it is not actually very plausible to say if you vote for the Republican governor, that abortion will be banned. It’s not really true.
But in Michigan or Pennsylvania, it is absolutely true, because there you have these very conservative state parties. There’s a really interesting story about how the state parties in the Midwest are really every bit as conservative as the state parties in the South, despite having much more moderate electorates. And they didn’t shy away, really, in a lot of ways, from the fact that they wanted these very tough, very unpopular abortion restrictions.
And I think Florida’s interesting too, because Florida is one of the rare Republican states where they had a trifecta and a very pliant Supreme Court and decided to not push ahead with extremely expansive changes to abortion law. And when you think about what the Florida Democratic Party is supposed to do, realistically, if you have scandal-free incumbents who are not doing anything substantively unpopular in a bad year, it’s hard to know what you can do about things.
I do think another aspect of this, and I really do want to say there are strong funding and campaign effects at play: Ron DeSantis really raised so much money. I don’t have the exact number, but I think that he raised more than literally any non-presidential candidate in the history of America. And I think his number was something like 40% higher than the next highest state campaign, just because the modal rich person is someone who doesn’t like Donald Trump, but does like Ron DeSantis. That’s not a statistical claim, but I think it hints at what’s going on.
At the same time, just going back to New York and Oregon, I think another side of when you think about what’s going on with these media market in Pennsylvania… Sorry, someone on my team is obsessed with them. You can tell a story that the New York media environment was very, very focused on crime. And so all of these voters in Pennsylvania, even though they lived in Pennsylvania, we’re hearing all these stories of Kathy Hochul and her bail reform, and that changed the salience of how they voted.
Just to talk about New York specifically, obviously I don’t know what happened and I don’t spend a lot of time thinking about the politics of my home state, but something that does come through, there’s two data points that I think are interesting.
The first is, if you look at turnout. Turnout, just the number of votes in 2022 divided by number of votes in 2020, is it a pretty stable band in basically every single county, and then just drops completely off the floor in New York, Which I think hints pretty strongly at poor and non-white turnout, which we do also see basically in the rest of the country, though I think it was probably stronger there.
The other thing that we’ve done is we’ve looked at state assembly districts, because the results have been broken out there, and we looked at Schumer versus Hochul vote share by assembly district. And something that we did see is that basically, Hochul underperformed the most relative to Schumer in the assembly districts that had more working-class white people, and also, this is an even stronger effect, heavily underperformed in the assembly districts that had the most Asians.
And so I think that’s pretty consistent with her crime story and the salience of crime story. I think, again, this highlights the importance of policy: that they did bail reform and then crime went up, and there was a situation where that could mean something. And again, getting to money, the Republican Party was much better funded this cycle than they had been in previous ones. And I think that you probably had similar dynamics going on in Oregon, though I haven’t looked too deeply into it.
Matt Grossmann: So polls were off less than normal overall. You have been interested in why they were off in the Trump years, and one of the explanations was that there were these people who didn’t trust institutions, and therefore, didn’t respond to polls and might not be easily correctable by just waiting on party. So have we solved the problem? Was it Trump-specific and will it return soon?
David Shor: Yeah. What I would say is that if you look, there’s always, I think, two different… When you look at 2020, there were two different polling errors that were going on. The first thing is that there has been a pretty consistent regional pattern of polling error in 2016, in 2018, and in 2020, where Democrats would consistently get overestimated in places like Ohio and then would be consistently underestimated in places like Arizona. And I think the reason why that’s happened is basically that latent socioeconomic status, which you can measure across a whole host of whether it’s social trust, whatever, has always been very correlated with survey response, but only recently became extremely correlated with vote share. So as a result, this creates this regional pattern of polling error that has been quite consistent, and from what I can tell that probably continued into 2022.
It does seem like the public polls underestimated Democratic vote share in Arizona and overestimated Democratic vote share in Ohio. We’ll have to wait until all the data is collected, but I wouldn’t be surprised if it basically continues in some form, even if it’s a little bit attenuated. But then the other error, because this is what people care about the most, is the national number.
In 2016 and 2018, the polls were basically right nationally, but they had these very big regional patterns of error. And in 2016, that was enough to change the nature of the electoral college. In 2020 though, the polls had a giant pro-Democratic bias and that had nothing to do with social trust or shy Trump voters or whatever.
That was really specifically because, and to be clear, there’s a lot of reasons to believe this, but it was because Democrats stayed home during COVID and liberals specifically stayed home during COVID and everyone else didn’t. So liberals just started answering a lot more surveys. We can validate this in a bunch of different ways.
We’ve looked at anonymized cell phone data joined the surveys, and we could see that liberals spent much more time at home. If you look at polling error and correlate it with local COVID rates, the relationship is quite strong. Which by the way, led to a mis-specified Chris Warshaw paper in the summer of 2020 that concluded that COVID was strongly hurting Republicans when actually it was just a polling non-response mechanism. It’s funny. Nothing against Chris. I do like him.
But if you flash forward to 2022, that last piece wasn’t really an issue. It does seem like the polls on average were pretty good. Might have even had a pro-Republican bias. But I do think the regional error pattern probably is still mostly there.
Matt Grossmann: So some Republican candidates seem to face an extremism penalty, including relative to other Republican candidates running in their same places. On the other hand, at least some folks on the left think that there wasn’t that big of a moderation effect for Democratic candidates this year. So where do you think the size and character of the moderation effect looks after this election?
David Shor: We’ll be releasing stuff on this soon, but I think it’s very hard to look at this data and not see this as evidence for returns to moderation. When you hear that Democrats in swing districts outperformed Democrats who aren’t in swing districts, the big difference between Democrats in swing districts and Democrats who aren’t in swing districts is that swing district Dems are more moderate.
I think if you go line by line, there’s a graph circulating on Twitter where you compare senate versus gubernatorial vote share by state and you see that the most liberal candidates underperformed their gubernatorial while the more moderate ones didn’t. I think if you actually go through and check candidates who said that they didn’t support defund, if you just do regressions, a lot of flags associated with moderation do well.
I’ve been working with another political scientist on some very cool conjoin measures of ideology where you put Democrats and Republicans against each other and ask which ones are more liberal and which ones are more conservative. What we’re seeing so far is that there’s a roughly 3% effect going from least liberal to most liberal, which isn’t that large in the scheme of things. But one thing I’ll say is it isn’t that large in the scheme of things, but in the context of very close elections it’s obviously a big deal.
Matt Grossmann: People are being asked to support a Republican or Democratic candidate with different ideologies.
David Shor: No, no, no. It’s like you show one Democrat, another Democrat, you’re like, “Which one do you think is more liberal?” You fit IRT model and then you correlate that with election results. There we see pretty strong stuff. Obviously, it’s hard to do with stuff like nominate because nominate doesn’t work anymore because the parliamentary, because we have a parliamentary system now, whether we like it or not.
But I don’t think it’s uniform though. I think this is a statistical effect that exists. But I think some moderate candidates underperformed by a lot and some left wing candidates do well and there’s still a lot of work to disentangle. How important is being moderate versus liberal versus how important is it to be independent from your party or how important is it to appear trustworthy?
And we’re doing a lot of research on that. I think it’s a very important question, but I think it would be a mistake to not look at the world around us and conclude that moderate candidates do better on average.
Matt Grossmann: So is this more about recruiting and nominating the right candidates or running the right general election campaigns? Or another way of saying this, I guess more pointedly is once you have, say Mandela Barnes as the candidate in Wisconsin, is there a general election campaign that you can imagine working as well as for someone else and maybe not facing any penalty for previous perceived liberalism?
David Shor: These things are additive, obviously it’s better to… I do think something that’s hard is that once you… Politicians can be very popular and then something can happen that can break them and then it can take a very long time to recover. Like Hillary Clinton in 2012 seem like the most popular politician in America. And then the Benghazi scandal and the email scandals happened and then she was not.
I think if you look at the Wisconsin race, Mandela Barnes had large leads really for a large part of the election cycle. But then once Republicans started attacking him on previous statements that he had made, that it seem like he supported defunding the police. His support dropped. And so I think it’s a really interesting story. It makes me wonder if ideology of candidates becomes more important as you get through the election cycle.
I think it’s just another example of this in the Republican side is that in the very beginning, Mastriano was pulling better than Oz immediately after the primaries, just because a lot of people came into the primary disliking Oz for reasons that had nothing to do with politics and didn’t really know that much about him other than the…
I mean they knew about the January 6th stuff, that was always covered very early. But it turned out he had a lot of other extreme positions on things like abortion that voters really cared about. And once we were in a position to tell people that it, I don’t know, it ended up paying him onto a penalty.
Matt Grossmann: So I know you’re skeptical of young voter mobilization, certainly as a mechanism for this election, but I know it does depend on the baseline and young voter mobilization is going to be much higher than say 2014 and 2010, but not as high as 2018.
David Shor: I don’t know why people are so confident of that, for what it’s worth, I don’t know. It’s true. If you just look at county level 2022 over county level 2014, it’s hard to… Once you go that far back, population growth is entangled. But if you just did some naive ecological regressions, it would say that the youth share will be closer to ’14 than ’18. But I don’t know, to be clear, I don’t want to rely too much on ecological regressions for age. But anyway, you were saying?
Matt Grossmann: But obviously it’s a traditional pattern that young voters turn out a lot less in midterms than presidential elections. There’s might be some reason to believe that after you get three elections in a row of slightly higher turnout in those midterm elections, that is sort of a closer to a new normal. What do you think that this is going to be… I mean, obviously Democrats are much more reliant on young voters than they were before just because their margins are so high. So it seems like that has to be a part of your predictions for going forward.
David Shor: Well, I always like to joke, people have been saying young voters are more democratic than all voters since 2004. And actually not that much has changed, but some things have changed. I think turnout was very low in 2014, and the Trump era, 2016, 2018, 2020, really featured absolutely enormous increases in turnout. 2018 was I think in some states either matched and in some counties succeeded general election turnout, which was absolutely crazy.
In 2020 was even higher turnout than 2016. I didn’t really know what to expect when it came to how many people would vote this cycle. I think something that I found very surprising was that 2021 gubernatorial turnout was higher than 2017 gubernatorial turnout. And so I think a lot of people were expecting a very high turnout election. What we ended up seeing instead was that turnout declined somewhat from 2018.
It’s hard to study too closely because in midterms, there’s really always a lot of state by state variance based on perceived competitiveness of races. I think about this less about young versus old and really just more about how high turnout is overall. Because if turnout increases substantially from 40% of the general to 80% of the general or whatever the 2018 members were. Mechanically the way that has to happen is that youth turnout has to really go up a lot.
I think that this is mostly a story of uniform swing going up and down, and I think the good news is that turned out only fell by five to 10% and not by more. It seems like interest in politics has gone down somewhat since 2020, but it’s still is at a fairly high level. If Donald Trump is on the ballot and there’s a primary next year, I assume interest in politics continue to.
But I do think something that’s worth thinking about is that there’s a trade off here that I think a lot of people don’t realize, which is the way that turnout gets high. It’s not by student debt giveaways or whatever, it’s just by making politics really divisive. And the way you do it is usually by increasing the salience of cultural issues that really play on identity. That’s the thing that gets people excited and out there.
And I think that that is, generally speaking, not a great position for the Democratic party to be in. I like to talk about the monkey paw of the past 10 years a little bit where, in 2012, immediately after 2012, age polarization had already happened. The electorate had polarized on racial lines, and education polarization hadn’t really happened yet. And so as a result, the correlation between your chance of being a Democrat and your chance of voting was higher in magnitude, though negative, than it’d really ever been at any point. I think probably in American history, that’s hard to know, but at least any point in the previous 10 years.
And so it was really true at that point that if you could dramatically increase turnout and hold support constant, then you would completely transform American politics. There were estimates at the time that if everyone had voted, Barack Obama would’ve done 5% better than if everyone hadn’t voted, and I think this led…
It was funny, I worked in Democratic politics. I saw this in the elite circles where it really led to this idea that we need to dramatically increase voter turnout so that we can create social democracy. And it led to people funding racial justice groups, climate groups that really did a lot, I think, to raise the salience of cultural issues. And they succeeded at a dramatically increasing turnout.
The problem is that it turns out that sporadic voters, people who don’t vote in every election generally are less educated and as a result tend to be economically center left, but culturally fairly conservative. And as a result of this, non-voters went from being an incredibly Democratic group to being a, in 2020, and either neutral or even slightly Republican leading group. So it makes me think that there are actually way more trade offs around voter mobilization and turnout than people realize.
Matt Grossmann: So should 2022 change our long term forecasts of Democratic potential? It seems like there was always three options you posed. One was to institutional reform, second was getting rid of education polarization, but the third was just successfully competing in center right areas. And does 2022 show that option three is quite feasible?
David Shor: Well, it is worth pointing out, Democrats are still quite severe underdogs in the Senate in 2024. Even if they have 51 seats, I think there are three very Republican states with Democratic incumbents. And so it’s not impossible, but I do want to be clear about that. But I think that one of the best criticisms that people had of the stuff that I’ve said is that it just might not work. It might not be possible. Maybe if you try to appeal. Maybe if you do your best to not say unpopular things and to say the right things on Title 42 and put out ads that are good, maybe you’ll just lose anyway, so maybe we should just focus on loudly advocating for our values and hoping that conditions change.
And I see this election as really an indication that campaigns matter, and that issue-taking matters, and that messaging choices matter. And to be clear, the reality of our situation is that due to the structural biases of our electoral system, Democrats are at a big disadvantage, and that means that we have to do everything right and they have to do everything wrong, and then we can narrowly win. And if the Republican party decides to stop making gross malpractice mistakes, then we’ll be in a much worse position.
But then, of course, if the Republican party decides to get better, we’ll all be better off anyway. So anyway, I guess to make that briefer, the fact that ticket splitting did not continue to decline, the fact that candidates like Jared Golden were able to considerably outperform fundamentals by making better decisions, that brings me a lot of hope. And just in terms of longer-term effects, it’s very good that all of our Midwestern governors were not wiped out. It’s useful to have popular Midwestern governors with electoral records to stand behind.
Matt Grossmann: So you stand a little bit at the center of the practitioner world and the political science world in that you’re bringing some insights from political science to practitioners, and vice versa. But you also, I think, you’re okay with annoying each side now and then. So tell us a little bit about what political scientists should learn from the practitioner world and vice versa?
David Shor: I’ll say I think academic political science has just been hugely, hugely influential to how I do my job. Just to throw names out there, David Brockman’s work on representation and salience really changed a lot about how I thought about these things. There’s really a whole host of work out there that I think has been really influential.
I think the really big thing, and it’s funny, you obviously have wrote quite a bit about this, is that there’s really been a credibility revolution in political science, where there have been these big methods improvements, data has become much more available, and obviously surveys have become quite a bit cheaper, and it’s possible to do survey experiments now. And so I think a lot of the traditional conventional wisdom that kind of came out of the political science of ’90s or the 2000s, that then went through a pop science filter and kind of turned into this idea that issues don’t matter, voters don’t care about policy, there are no swing voters, et cetera, it’s very useful, I think, to actually look at the data and learn that stuff isn’t true.
I mean, it’s really incredible to me, the extent to which really basic political science facts, like the party that holds the power usually loses power in midterms, or the idea of thermostatic public opinion. I will tell you, I mean, people at the highest levels of American government have no idea about these facts. And I think that’s a shame, because I think political scientists have really learned a lot about how democracy works, and folks who are in the business of trying to win elections should take note. And to be clear, I’m not the only one. I think a lot of pollsters take the political quite seriously.
I think in the other direction, I think the speed at which academia moves really puts political science at a little bit of a disadvantage because the American political system and the dynamics of how people vote have really changed dramatically in the past eight years, in the past four years, in the past 12 years, and I think it’s very common for a political scientist to just not really be able to keep up with those facts. I’ve seen political scientists and economists make claims that incumbency advantage is 8% based on that regression discontinuity design paper, which to be clear, nothing wrong with it, but it’s aggregating over a 40-year period. I wish incumbency advantage was 8%. That would make my job way easier.
I think that really throwing yourself into the complexity of what’s out there and moving quickly, I think that a lot of political scientists were very slow to catch onto racial depolarization. I don’t know the political science community that well, but one of my personal frustrations is I think that they were very slow to catch onto the fact that education polarization exists. I think there were some quasi-ideological biases of American political scientists that just made them think that the VRA kept happening. The fact that it took Piketty to go out and actually publish a paper on this stuff. I just remember political scientists denying that education polarization was a thing, really as recently as the fall of 2016. So I think that there are definitely some blind spots, just to list some grievances.
Yeah. I’d say my two grievances are actually probably being slow to catching education polarization, and then kind of a lot of the pop political science that came out of the 2016 primary on the Republican side, where a bunch of people were really convinced that there was no way that Trump could win. And I remember being very frustrated about that, just because you if pull out the regression from the party at the sides and you do endorsement points and vote share, and you plugged it in, it was like, “Oh, Trump has a 50% chance of winning.”
And that kind of gets to kind of my final, I guess, critique, maybe not of political scientists, maybe just the academia, which is something I really like as a practitioner is that our job is to make numbers and predict things for a living, and I think that really does bring a certain level of rigor to your thinking. And to be clear, I’m obviously not perfect and we make mistakes all the time, but I think that there’s a lot of value to actually trying to make testable predictions and keeping yourself accountable and plugging things into the regressions, getting your hands dirty. I think that’s a big thing with practitioners is we just have to pump out. Our team, I think by the end of the election cycle, we were doing something like 2000 tests a week, and so it’s a very different beast. But there are advantages and disadvantages to both, obviously.
Matt Grossmann: So one of the things that you mentioned was slow speed, but obviously we’re talking in the wake of an election, and there’s been a critique of people making kind of immediate pronouncements. So on the one hand, we have limited data, on the other, at least I think it was quite clear some of the main lessons from the last election were sort of immediately clear in geographic data. But where should we be slow to jump to conclusions about what happened in the election, and where should we be more confident that we can pretty much know what went on quickly?
David Shor: I think that just to talk a little bit about the data availability schedule, what we have right now is we have county-level data in most of the country, though not everywhere, which is relevant for some things, because most parts of the country complete their counts, actually, pretty quickly. But then obviously, if you want to know what the overall popular vote total is going to be, we’re not really going to know until mid-December thanks to places like California. But I think that over the next couple days, we’re going to have completed counts on a county level in 90% of the country, we’re going to have precinct data in probably something like 20 to 30% of the country probably within the next week or two. And then probably the final bit of data, and this is what’s annoying, is we get voter files really basically in February or March. And I think it creates this kind of dysfunctional dynamic.
My big-picture thing about this is I think that people kind of have too high a bar. I think a lot of academics have too high a bar for certainty, where one example is that if you do a county-level ecological regression nationwide on age and there’s a strong negative slope, chances are you’d turn that one down, but I don’t know. On a county level, age is correlated with various things like urbanity and population growth and whatever, so I don’t know, I’d probably give it a 10% chance that maybe there was no relationship. I’d be pretty surprised if it went the other way.
But I think that in the absence of political scientists kind of trying to go out and say, “Hey, I think this is what’s happening,” there is a pretty well-funded group of people on the other side who are trying to advance claims that are wrong. And I think that it is actually very important for us to do the best that we can to get a rough sense of what happened out there. Because I mean, I don’t want to be too hyperbolic about this, but the immediate narrative after the 2012 election was that Obama won because he dramatically increased non-white turnout and youth turnout, and that Mitt Romney was too extreme on abortion. And this led the Democratic party to obviously make a bunch of poor strategic decisions, and it also pushed Republican elites to go in a direction that left themselves open for a populous takeover that ended up kind of wiping out the only pro-democracy people in the Republican leadership.
I used to tell a joke where it’s like, “I never thought that poorly-conducted exit polls was going to be the cause of the end of the world,” but I think there’s a real case that it was. And to be clear, what actually happened is that Barack Obama won cause he convinced a bunch of culturally conservative white people in the Midwest to vote for him at higher numbers than Kerry. I think it’s better to put out your tentative claims, rather than wait until March when you really know for sure, and when at that point, if everything is settled in and no one really is going to care.
Matt Grossmann: So what are your biggest unanswered questions that you expect to have better answers to soon, and what’s next for you?
David Shor: Yeah. The things that I’m most curious about, I think that it’s still kind of too early to make definitive statements about education polarization or racial swing. It seems like there are some state-specific stories. I mean, it seems like there are a lot of state-specific stories for what happened, maybe more than usual, that I think will tell us some very interesting things about responsiveness. I think that I’d say it’s still kind of hard to get a good handle on the subgroup turnout stories. And then we really have to wait until the house results come in to really be able to answer questions about responsiveness for ideology. I’d say those are probably the big ones.
And in terms of what’s next for us, I mean, I helped found an organization called Blue Rose Research. We’re a 30-person shop, mostly data scientists and software engineers. This cycle, we worked with, I think, something like 150 clients. We tested thousands of messages. I think this cycle is proof that campaigning and electioneering competently matters, and we’re going to try our best to prepare for 2024. I don’t know. Probably the first thing I’m going to do is get some sleep.
Matt Grossmann: There’s a lot more to learn. The Science of Politics is available biweekly from the Niskanen Center. I’m your host, Matt Grossmann. If you liked this episode, I think you’ll like our prior discussions of positioning, campaign practice, and policy. When Information About Candidates Persuades Voters, When Public Opinion Goes to the Ballot Box, The Past and Future of Polling, Abortion Politics Takes Center Stage, How Does the Public Move Right When Policy Moves Left? Those five recommended episodes are available on our website. Thanks to David Shor for joining me, and please listen in next time on The Science of Politics.
Photo Credit: iStock