Public opinion polls often show large majorities in favor of hypothetical changes in public policy, such as universal background checks for gun purchases. But when voters have the opportunity to enact those changes in ballot measure initiative campaigns, the results are usually much closer. Jonathan Robinson finds that state public opinion is related to initiative voting results, but large majorities are substantially reduced. Part of the reason is status quo bias: the electoral context makes it clear that voters are being asked to change current law. Robinson is a political practitioner engaged in scholarship, who merges the perspectives of both worlds. As co-author of a major report on voter turnout and vote swings in 2020, he also looks ahead to 2022.
Guest: Jonathan Robinson, Catalist
Study: “When Mass Opinion Goes to the Ballot Box”
Transcript
Matt Grossmann: When public opinion goes to the ballot box. This week on The Science of Politics. For The Niskanen Center, I’m Matt Grossmann. Public opinion polls show large majorities in favor of hypothetical changes in public policy, but are those polls representative of what happens in the context of a real world political campaign where voters are being asked to make legal changes? Many states have initiative and referendum processes for putting issues directly before the voters. So there’s an opportunity to find out how well they match. This week I talk with Jonathan Robinson of Catalist about his new paper with John Sides and Christopher Warshaw, When Mass Opinion Goes to the Ballot Box. They find that state public opinion is related to voting results, but large majorities are substantially reduced at the ballot box. Don’t expect the overwhelming public support for gun background checks for example, to materialize when voters have a chance to enact them.
Part of the reason is that the electoral context makes it clear voters are being asked to change current law. Robinson is an exemplar of a political practitioner engaged in scholarship. So we also take the opportunity to discuss how practitioners and academics see the world differently and how they work together. Robinson is also co-author of a major report on voter turnout and vote swings in 2020 and he’s gearing out for 2022. Being in the midst of the action not only allows him to have access to unique data, but also to see social science from a distance and as a user. Our conversation shows that we can learn from applying our ideas and results to real world elections. It turns out he’s been working on understanding ballot measure results since he was a student.
Jonathan Robinson: As a person who is pretty steeped in the research on public opinion and who’s also following along in practitioner politics circles, I pay attention to the news a lot. And an area of interest with the groups and in organizations we work with is very often ballot initiatives. And it was interesting to me because I had sort of a sense from analyzing survey data about what the public thinks on a variety of issues, your standard American national election study, or Gallup press release, or The Washington Post poll, or what have you. And I thought I had a really good sense of how the public thought about these sort of things. And it was very interesting to me when I started to follow ballot initiatives more closely that sometimes it seemed that the results didn’t necessarily jive with what I thought I understood about public opinion in those areas.
It turns out to answer that kind of question systematically with data is really hard. Right. So, you sort of want to be able to have surveys in the same state talking about issues. And you want to have initiatives on a similar-ish topic and doing that at scale is really hard. But it turns out an academic, Chris Warshaw has a project that does sort of exactly this, aggregating lots of publicly available survey data over time, using complex statistical modeling to then project that data down to the states, as well as over time where that data doesn’t exist or when the question wasn’t asked and pull together this really amazing data set on how public opinion has changed over time over the last 40, 50, 60 years. And my thought was that we could use this data to try to answer those questions. And so we went about this process of trying to identify questions in a similar topic or specific domain as to initiatives going back in time, quite a ways.
And we ended up finding about 200 or so pairs of initiatives and public opinion estimates in the states. And at the same time, that’s a lot of data, at the same time, right, it’s only 200 plus observations, so it’s not a lot, but it was something to be able to try to answer this kind of question of were these things that I had seen in passing in my work, or in talking with colleagues, was that sort of the exception or was that a rule? Were there these really big gaps? And so that was the Genesis of the project. The idea was that it might be possible to use ballot initiative results on specific domains, similar to questions that have been asked in surveys to see what the comparisons were. And what we found I thought was really pretty interesting. Number one is that these things are related.
I think the correlation between a ballot initiative result in a state and the survey estimate is between 0.6 and 0.7. So we’re not talking about something where there’s no systematic relationship or because a common critique of surveys is that there is some underlying sort of dimension that structures how people respond. But there’s also some research that suggests that actually among the public, there are other kinds of dimensions at work that are much more kind of random, or idiosyncratic or difficult to explain, or might be even consistent with random guessing. And so there might be a concern actually that the relationship between these things might be not as strong as you would imagine, but we found this relationship. So that was a good finding. I think that was a win for public opinion. Because I think one concern that you had was what motivated these projects were examples where the relationship was not very strongly related, but that was great.
That was sort of like a cool validation. Survey estimates in these states are related in a positive direction to ballot initiative results. The thing we were interested in though, and what motivated the project was, why would they be different? And we had this view, at least from some of the initial initiative categories where we saw the biggest differences. That it actually might be that surveys were biased in the favor of liberal, the liberal position in public opinion. So the place we kind of started on this project on was on issues related to guns. And most often the initiatives had been run were run by liberal groups, putting up a ballot initiative to make policies around gun control, purchases, et cetera, more in the progressive directions. So I wouldn’t say gun control, but you’re talking about background checks measures and things like that, where we saw these big discrepancies. And that was the motivation. That was initially what brought this up to look at this more systematically.
And the really fascinating thing that we found was that actually wasn’t really true. It was literally true that for guns, there was a relatively big gap between where the initiative, which was fairly competitive or maybe more in line with the state’s partisanship than what you see in public surveys, where support is between the 70s and 80s in terms of how popular it is. But that was also true of things that were very popular that conservatives were putting on the ballot. So an example of that was there are initiatives that have been passed on abortion policy related to parental notification or parental consent for abortion where the public policy choice at hand and in that period of time is if a child is under the age of 18, if they need to get parental consent to get an abortion.
And in surveys, that public policy is actually relatively high approval from the public, relatively high support. But when that initiative gets put up on the ballot, it also underperforms. So we saw this sort of interesting balance here where basically when you saw something that was very popular, whether either the conservative position and initiative was popular or in another area, the conservative position was popular, that’s where we saw the biggest differences. And one of the big reasons that we thought it might be the case was that just, and this is maybe a more academic way of putting it, but practitioners, when they talk about ballot initiatives, they talk about the no bias, right, that just like no is advantage in a ballot initiative. And it’s interesting. My dad’s an economist and he was like, oh, this is sort of a risk aversion thing where there’s sort of again, people are really unwilling to make these sort of changes to the status quo. Whereas in political science, the term art is status quo bias, right, but things are much more likely to stay the same than they are to change.
And so we actually ran survey experiments to sort of try to validate this idea. And it does seem to be the case like when you emphasize in a survey question about gun control or about parental consent, it does seem to be the case that when you prime survey respondents in an experimental setting, that not just are you pro or against this particular public policy, but that this policy option that is the yes option is going to change things from what it was, which is like, this is how you purchase guns. And under this regime you would purchase them differently, right, as opposed to asking a survey that’s question, are you in favor of the current regime. Right. That distinction.
When you told people that they were in favor of making a change to the status quo, that seems to reduce their likelihood of wanting to support changes. And it most often came from people who were sort of not really sure or maybe weekly in favor of an initiative or the public policy change. And so that’s sort of the sort of five, 10 minute version of what this project is in sort of soup to nuts. And it’s been a really interesting piece of work and been eyeopening for me and my understanding of politics.
Matt Grossmann: So just so people understand the details, let’s go through an example. So we have multiple states that have tried to enact background checks. And so what does our kind of data look like for those states and what does the outcome look like?
Jonathan Robinson: Absolutely. So one example that we have that’s just very instructive is that in 2016, there was something on the ballot in Nevada called question one. And the title of it was just Nevada background checks for gun purchases. And so when we start with the survey data, as I mentioned, my colleague, Chris Warshaw, he has this big data set that uses a statistical method called MRP multi-level regression post stratification, where you’re using survey data and auxiliary census data and as I mentioned, historical polling data as well and using that data, there’s an estimate of what Nevadans think about background checks. And it’s important to know. Like the survey question that was asked of the people, a lot of that data comes from the congressional election study, sort of a publicly available government funded, large scale survey of the country of where there’s a large number of Nevada voters in that survey.
And the question is relatively straightforward. It’s something like on the issue of gun regulation, do you support or oppose background checks for all sales, including at gun shows and over the internet? So that’s the survey that’s asked of respondents and the estimate that Chris Warshaw, one of my co-authors on this paper had is that 86% of Nevadans support that public policy. Now question one in Nevada, and that’s 4-20-16. So question one in Nevada, the vote share that that received at the ballot box was 50.4%. And now this question I granted is different. Right. It’s a little bit longer of a question. It’s a bit more in legalese, but it says shall chapter 202 of the Nevada revised statutes be amended to prohibit except in certain circumstances, a person from selling or transferring a firearm to another person unless a federally licensed dealer first conducts a federal background check on the potential buyer or transferee?
So they’re not exactly the same thing, but of course it never really is. Our goal here was does our understanding of what supports for background checks is, which is largely informed by survey data. How does that line up with granted a slightly different rendering of the public policy, but in spirit is very much on the exact same concept. And so that’s a really big gap. So that’s about a 35 point difference from something that’s seen as broadly popular, common sense reform. When Nevadans voted on that, it was basically 50, 50.
Matt Grossmann: So you did find some differences across issue areas. You talked about there not being a broad difference across the ideological direction of the policy change, but you also had some of the conservative initiatives were kind of concentrated in a few issue areas. So talk about what you found differences and similarities across issues.
Jonathan Robinson: The area where there is sort of the least difference between what public suggests an initiative or a public policy area should have some level of support in the public on and where it gets in the ballot initiative realm is on issues that are more heavily contested. Part of the interesting sort of quirk of the ballot initiative data set is that these are initiatives that are either put forth by a public petition. Right. So you need a certain number of signatures in a state or are legislatively referred. So the legislature, sometimes citizens ask the legislature to weigh in on a question and they defer to the voters. Sometimes legislators just vote on putting a initiative on the ballot about a particular public policy.
And so there is sort of a interesting sort of strategic case of what initiatives get put up on the ballot and which ones don’t. And that’s consistent when we actually sort look at the underlying surveys and 90% of the time in the aggregate, the initiative that’s being put forward, whether conservatives put it up or liberals put it up in their direction, it’s over 50% in the polling that we have. So there’s certainly some strategic aspect of that, but generally speaking issues that are more 50, 50. So at the time some of this has changed over time because we’re looking at a data set that goes much further back in time. But generally speaking, what you end up seeing is that things that are more popular in either direction have this issue. So I mentioned guns is one.
Popular in either direction have this issue. So I mentioned guns is one where you see a big difference on the left. Similarly, we see a relatively large difference on minimum wage initiatives, for example, even though they still pass. And then on the right, an example of these are, as I mentioned, abortion initiatives at large, not just on parental consent, those tend to be in that category and similarly education initiatives. So stuff around charter schools or changes in the status quo to education policy that are much more often happening on the right wing, the conservative view of things.
Matt Grossmann: So one interpretation of this could be, or has been that public opinion polls are thus not very good, that the real opinion is the one expressed at the ballot box. What do you think?
Jonathan Robinson: It’s very funny. One of my other co-authors on this paper, who I know has written stuff with you is John Sides. And I actually first met John Sides in a classroom as an undergrad. And when we were writing this paper, I was actually thinking and ruminating on the same question. It’s of one of the reasons that I was attracted to this topic, not just because ballot initiatives are of substantive or practical importance to people who work in politics, but it brought up these big picture questions for me about really what is public opinion. And I was looking around for classic, stuff you read in your intro seminar political theory around what public opinion is. I found this great quote from Vio Ki, where he said something to the effect of public opinion is opinions held by private people about the government, which the government or elected officials should find it prudent to heed.
And I had this quote and it turns out that John had talked about that. It’s in one of the slides from the first class in Public Opinion 101 that I was in. So it’s this foundational thing that maybe gets forgotten that public opinion used to be at least conceived of. And I know there are great books on this. Like protests was considered public opinion until public opinion was more formalized with surveys in the government realm and your Gallup and Roper and the sort of greats of the initial survey world. So there’s sort of a standard way people have around asking survey questions developed by the Michigan School in the sixties that give us this great time series that we understand how public opinion works going back in time and how it changes. But I don’t think it’s necessarily the end all be all.
It’s just sort of… There’s a standard way of doing things, a best practice. Something you learn that if you want to ask about guns, or abortion, or climate change, or the minimum wage that you ask a question this certain way, there’s always been this concept of latent opinion. There’s always been this concept of multifaceted opinion. And so I think two things. One is, I do know that polls sometimes struggle. It’s not always true to match up how an initiative will perform. So that might suggest if there really was a problem with initiative specific polling that maybe there’s a representational issue. I really do think that there’s just something qualitatively different when a voter has to go to the ballot box, has to really think about this in a way that maybe a survey is more abstract about.
When you go to vote, maybe that weeds out a number of people who maybe have a positive view. Maybe it actually is really true that background checks is really popular. It’s just that when people are forced to think about it and actually say, it’s actually on you to make the change, not on lawmakers to make the change, you become a little bit less certain of your vote, or you’re more susceptible to counter arguments that may be very persuasive. And so I really do think that public opinion is this murky mushy thing. So maybe not so easily defined in this one way. And so I actually think ballot initiatives really could be a useful tool for people who are studying what the public really thinks. I think it’s debatable to me about which is more valuable. I think the answer is probably somewhere in between, some sort of smart averaging of the two probably is telling a fair story.
Matt Grossmann: I was just going to say your experiments suggest that this is about a general orientation to change, not the kind of campaign environment, but the usual status quo bias story is that the arguments made by the negative side are more easily persuasive or that money on the negative side is more valuable or something along those lines. So do you think this is just an inherent at the point that you are asked to change policy and you know you’re being asked all these things kind of click in, or is this that it creates a political environment that allows peoples views to actually be changed?
Jonathan Robinson: That’s a great question. We’re actually in the process of trying to answer that kind of question. You can sort run survey experiments that try to do this kind of pro, con framing. And so we are trying to answer these kind of questions in a more systematic way. My understanding is that from other work in the public domain, that people really do think that this is a factor. In the survey experiments, we do see the status quo framing does reduce support, but it doesn’t explain all the variation that we see. So we actually have two treatments that we give to survey respondents. One is, again, all this is compared to that baseline standard classic way of asking a public opinion question. One is this one that really makes it obvious to the respondent that you have to make a change. And then another one that’s actually just mirroring the initiative language, where we don’t really have a strong sense of what that treatment’s actually doing, just that it’s sort of saying, hey, maybe something about how these things are written is sort of doing it.
And in some cases that initiative framing is just as impactful as the status quo framing. So it’s either that they’re operating on a similar thing, maybe an official explanation reminds people or concerns people or causes people to be more, again, concerned about making a change, or it could just be that official language and that official context or whatever is just another feature of what’s operating on voters. Because in these survey experiments, the effects that we observe are they exist and they’re real, but it doesn’t seem to be the only thing that’s going on. So I’m very open to there being other things that are going on in these things and in this kind of work, just the interesting thing to me is that from a logical, just purely theoretical perspective, in my view, you have organizations who are putting up initiatives, right?
Sometimes you could imagine that this is not the case, but people are putting up initiatives because they hope they’re going to win. And more often I imagine the pro side is relatively well resourced. It’s actually very hard to find systematic data on this, but my sense is, if you are putting up an initiative on the ballot, you will potentially have foes but sometimes you don’t and sometimes, and oftentimes you will know the kind of support or opposition you’ll get. And so I’m a little bit less favorable towards something about the campaign spending environment being different or how it might be very different across issues. The case of the gun control example in Nevada, the organizations on the left who are very in favor of changes to the gun policy more on these safety issues. These are relatively well funded campaigns.
QSo I could imagine there is sort of a partisan Q piece happening. So for example, Nate Cohn in the New York Times wrote about these gun issues in particular, after the shooting in Uvalde, Texas. And in some of these cases where he looked at precinct returns or county returns, the relationship between voting on a gun initiative that happened in Maine had a stronger correlation with presidential vote than the congressional vote did. So, you can tell a story where there is this more partisan anchoring, maybe in more partisan electoral context or where partisan strength or identity is stronger. You might see these kinds of things operate in that kind of way, where the partisanship stuff really matters. But I think the status quo bias is important and there are other theories to litigate as to the rest of the difference.
Matt Grossmann: So you have to go and translate this incomplete research to people who are actually trying to pass or oppose ballot initiative. So let’s say you’re invited to Michigan to try to pass an abortion rights initiative or a voting rights initiative, what do people need to know?
Jonathan Robinson: Yeah. So this is actually a fascinating question. So I think I probably spend most of the time talking about the abortion one, because I mean, both of these are topical. These are initiatives that seem likely to be on the ballot in November in Michigan, though at the time of this recording, I know that’s a little up in the air, but seems likely. So that’s a really interesting question. So I think when I’ve talked to friends and colleagues and other practitioners about this particular initiative, it scrambles a clean way of understanding how this might operate. So in advance of the Dobbs decision where the Supreme Court overturns Roe V. Wade, you could really make a case that codifying Roe in some way in the state constitution, which is what this Michigan ballot initiative that’s being proposed would do among other things, you could really imagine that, that’s actually changing the status quo.
In Michigan public policy is a certain way, and you’re trying to change it in this material way. You could imagine that would be an uphill battle for all the reasons that we mentioned. It’s a contentious public policy issue. I would say it’s the pro-choice side on this particular issue is maybe relatively popular, but not strongly popular in a state like Michigan that’s maybe a little bit more closer to the national average when it comes to public opinion. And that was just the context in which it was seen. But now that the Dobbs decision has come down, it scrambles things a bit. So now the status quo sort of was Roe and now it’s gone. So the argument that you could conceivably make is similar to the one that was made in Kansas, an initiative that happened recently, where the argument that was made is that in order for us to keep things the way they are, where abortion is enshrined in the law, in particular, in this case that the state constitution is a vehicle for that.
In Kansas, it was different. They wanted to take that away. But in Michigan, it still, I think will be difficult, but it’s moderated a bit by this interesting change of the status quo from when the initiative was proposed to when it’s being put up. So it represents the challenges of translating academic or historical research to modern times. Sometimes things look relatively straightforward, but sometimes they’re overcome by events and it’s unclear. And so I think the main thing I think that’ll be really interesting and I’m sure people are feverishly studying this right now is just how voters see the initiative pre the Supreme Court, putting down its ruling in the Dobbs decision and post, and if that’s really changed. Because if that’s changed, I think that’ll be an advocate’s favor, right? Because if it really is about preserving the status quo, that actually the people who’ve made the change have been conservatives, the Supreme Court, national conservative groups, and that we need this initiative to pass, to protect abortion rights.
If that’s changed. I think and the public sees things that way I think that’s to the favor of advocates. If it really is seen in a more traditional light, then it will face maybe this more uphill battle, but it really depends. And obviously it depends on what latent opinion on abortion is in Michigan, which is a question that we can answer with other data points. But it really does show the complexities of when you think about these concepts, that do seem a bit more concrete or straightforward, but then what is seen as the status quo changes underneath you in current events, muddies exactly what the takeaway might be.
Matt Grossmann: So you’re at this crossover point between social science, research and practice. So what kinds of perspective do you bring when you’re the practitioner in a political science room and what kind of perspective do you bring when you’re the most political sciencey of the practitioners in a practitioner room?
Jonathan Robinson: When I’m with an academic audience, I’m encouraging them to be more practical. I’m encouraging that them to focus on particular practical component or… So here’s an example I was thinking about this the other day. We work with a lot of organizations who care a lot about election administration and research on voting rights. And very common the way that researchers will assess the effects of election administration is in a regression context. You might look at, in the case of a voter ID law, maybe you’ll look at who has an ID and who doesn’t has an ID, or maybe in the case of a rejected mail ballot. Maybe that’s the example we’ll focus on the case of a rejected mail ballot. We’ll look at among people who vote by mail, who gets rejected and who doesn’t. And we’ll do comparisons between those two groups and we’ll use a regression model or a difference in means, comparing the different groups to see whose ballots are more likely to be rejected based on demographics. And what you’ll see in an example of these kinds of research is that younger voters, voters of color, people who are less experienced in voting by mail, they’re more likely to have their ballots rejected. And that is true in the data, however as a practitioner, you also want to know just who gets their ballots rejected? Just descriptively. You do care about the differences, but you also care just about the descriptive facts of these populations. And it turns out, in most contexts, the differences between people who are rejected and not are not very large and actually people whose ballots are rejected, even though they’re younger, even though they’re more likely to be people who are less experienced with voting by mail, even though they’re more diverse, they’re actually older, whiter, more experienced voters because that’s who votes by mail.
It’s just that if their rejections were of mail ballots were standard or similar, and there was no bias across demographic groups, there would be fewer ballots rejected among younger voters, experienced voters and more diverse voters. And so when I talk to academics, I encourage them to try to translate the practical findings of their work to an audience that both does care about differences, but also cares about the underlying population of voters. So, that’s an example. When I talk to practitioners, I kind of do the opposite. Sometimes they’re very narrowly focused. A very common thing I hear when I speak to practitioner audiences is that every scenario they’re working in is sort of unique. Like, “Oh, I know you guys have conducted this analysis in the Ohio, but in Tennessee, it’s really different.” But the Ohio people say the same thing, “Oh, I know that you’ve done this analysis in Wisconsin, but in Ohio it’s really different.”
And so sometimes the practitioner focus is rightfully so, extremely narrowly focused, targeted on the area that they’re in, in a way that’s just incredible. You’d have people who are just deep experts in the area that they work in, that I could never be because I work all over the country. These are people who maybe you’re the expert in Michigan politics in a specific jurisdiction. I’ll never have the depth and breadth of knowledge, no matter how much time I spend trying to focus on this compared to someone who’s just focusing on it their whole time. But sometimes then, because they’re not looking at outside of their area or they don’t have a chance to operate in different contexts or in the case of people who are paying attention to politics and they’re political junkies, maybe you’re looking at international elections, comparative politics. You’re seeing what’s happening in German elections or Israeli elections or Brazilian elections.
Sometimes that lack of context, big picture stepping away from your area, that’s something I often say to practitioners, “Hey, we actually did this research somewhere else. I think it has some bearing on the work that you’re doing.” And what’s interesting is that I think part of that is the nationalization of elections. There’s a lot more of a transferability of what you might learn in one area to another, with the proper demographic caveats and changes in the environment and just sort of knowable things that yes, actually Tennessee is very different from Ohio, but it’s not something that’s totally not overcome-able with just a proper understanding of politics. But it is sort of interesting that you do say to the academics, “Hey, should care a little bit more about nitty gritty politics stuff in terms of how you translate this.” And then to practitioners, it’s sort of like, “Hey, a broader perspective actually can be very helpful.” As with most of these things, the answer is almost always in the middle. And whenever I do research, I’m trying to find a happy medium, a middle ground on those two dimensions.
Matt Grossmann: So you try to do that with your co-author report on what happened in 2020 and have done that for some other elections. So talk about trying to use the lots of data that you have with some understanding from academics, but also some practical concerns and trying to get at this question of the relative impact of turnout and voter share change and what the big changes were.
Jonathan Robinson: Yeah, it’s funny. I think the question we were trying to answer, the reports that we’ve done on the elections have of been tongue in cheek, what happened in the 2020 election? What’s going on? What happened? But it has a forbearer in traditional political science. So who votes? A classic book trying to literally just describe, what are the difference between voters and non-voters? And these kinds of questions are foundational questions that we know much less about than we would like in any certain terms. So that book was written largely with census data, from the voting and registration supplement of the CPS. The fact that we’ve returned to it, I think suggests that it’s not necessarily an open and shut question, but it’s foundational. Who votes in American elections is really important. And even beyond that, this question of just then conditional on who votes, who supports which type of candidate and at what rate and how that’s changed over time is actually both a thing of academic interest.
You’re trying to estimate these quantities. You’re trying to understand things in broad terms, but it makes a lot of sense, it’s also something that practitioners really care about. The difference I think is that we obviously bring a ton more data to bear on this question. So at Catalyst, our major product, what we focus around is the voter file, which is a conglomeration of 51 state voter registration databases that we’ve harmonized and pulled together. Appending demographic modeling to try to estimate the demographic composition of the electorate on a number of different axes and dimensions. And then supplementing that data with large scale survey data and the most granular, widespread available data on election results from precincts all over the country, that we then have to just have an entire team associating the individual level election results, and those precincts with precincts on the voter file, which actually are different quantities.
So it’s actually a very complicated process where we’re combining all this stuff together, where we’re trying to answer this really simple, but really hard question, which is just who votes in American elections and for which candidates? And so I think it’s a cool project, because it’s continuing this tradition in political science of asking this basic question. That allows us to answer this very simple question, which is of extreme importance, which is in the 2012 election, what percent of voters who are white and who don’t have a four year college degree voted for Barack Obama? Then in the 2016 election, that same demographic group, though it’s changed somewhat, it’s a slightly smaller share of the electorate, as the electorate gets more diverse, as college education attainment increases, how that group then changes their sentiment towards voting for Hillary Clinton. And then again in 2020, how that group then decides how they’re going to vote for Joe Biden. And telling that trajectory is just of extreme importance to understanding politics, and quantifying that and how that’s different in the Rust Belt versus in the Sunbelt.
Those kinds of questions are really, really important to thinking about the electorate. And I think there’s actually a case that misunderstanding the electorate can have pretty important impact on campaign strategy, or just the idea of where the electorate is. So there’s this classic book called the Emerging Democratic Majority by Ruy Teixeira and John Judis. When you read the book, as a few have, I guess, it of seems to be the political version of Capital by Thomas Pikkety. They are more nuanced, but the driving force around the book was just that the electorate was going to get a lot more diverse, and that was going to lead to this potential for Democrats to have a permanent majority that was a growing, diverse, supportive Democratic base, that was alongside a white working class in this case, operationalized by voters without a four year college degree levels of support that would provide Democrats a governing majority for decades, or in the next decade or something like that.
And part of the issue there is that some of that intelligence was driven by exit polls, which suggested that the people who vote in presidential and midterm elections were younger, more diverse and better educated than our data or census data had suggested. And I wouldn’t say that we had a new take on this, Mike McDonald had written a paper about this in Public Opinion Quarterly in 2004. So wasn’t the first case on this that had been brought up, though obviously we’re bringing a ton more data, a ton more modeling across a lot of different areas to this kind of question. But you can imagine that if you see an electorate or think about a coalition of voters, and this actually is sort of true. Popular commentary about President Obama’s coalition was that it was a young, cosmopolitan, well-educated, diverse, a coalition that needed to be reassembled in 2016 by Hillary Clinton.
And this was something that was said on election night in 2012 and was something that even I had misperceptions about heading into the 2016 election before we were researching this very basic question of what percent of the voters in Wisconsin are white or in Ohio or Florida are white, and found this big discrepancy, at least in political practical terms with what the exit polls had to say. And so it actually is really important to try to adjudicate and get specific answers in some of these questions because people interpret elections and use their popular understanding of why elections went the way they did to then change how they act in politics. It’s an equilibrium. Something happens and then you learn something new or something changes, and it changes the way that people act.
Matt Grossmann: The report was also written with the support of Democratic or left leaning interest groups, and it kind of came in the wake of a lot of discussion of Democrats losing Hispanic support and gaining white educated support was a dominant narrative of what happened in 2020. So presumably, you would get the same answers to the same questions, regardless of who was it supported the report, but the framing seemed to be quite different and to reflect some of the concerns of the Democratic reaction to that dominant narrative. So I guess talk about how you deal with that? How do you deal with the framing to make it useful to people who need to hear the message, but also are concerned with how that message sounds?
Jonathan Robinson: Yeah, I think there are two things. I generally think that this process for the support they put together, I would say is very similar to peer review. I haven’t gone through the peer review process, but I have to imagine that when you do, sometimes the paper or the analysis changes totally. Maybe they found some totally damning coding error or analytical error, but more often than not, I have to imagine they confirm that the research is on solid ground, but they’re like, “You talk about this other thing that you kind of pushed aside that this is way more nuanced about.” Or, “You talk about in this way, but when you talk about it, you kind of make seem that X is the case when really it’s more like Y or Z.” And again, I don’t know if this people trying to be kind to their reviewers, but I have heard from people that the papers do get better after review, though I assume people are mad that it takes a while.
We’re not under those same time constraints, but I think it’s really important to put your ideas and thoughts and reactions in front of people who are experts in this area. And in fact, some people who we did speak to are prac academics or practitioner academics. And so in the case of the Latino vote, we worked very closely with Matt Barreto, who’s a professor of political science at UCLA and is an extremely well regarded scholar on Latino public opinion. And we worked with Carlos Odio at Keith’s Research who runs an organization where their entire focus is on the Latino electorate. And we’re generalists at Catalyst. We’re studying-
Jonathan Robinson: We’re generalists at Catalyst. We’re studying, as I mentioned, every state, every demographic group. It’s really hard for us to get the story right, but these people are, if we don’t talk to them, untapped reserves of knowledge that we would never hope to have on these particular issues. They look at this report and ask these important questions. In the case of the Black vote, African American vote, we had two stories that you could tell with the data. One is a story where the African American electorate, the number of African American voters went up. Another story, which is confirmed in our data and somewhat in others that African American voters, particularly African American men and younger African Americans, voted more for Republicans than they had in the past. You could tell some nuanced story that just focuses on that.
But actually when you look at these things in tandem, what you realize is a really interesting nuance, which is that actually it seems like the marginal African American voter in elections, the kind of African American voter who votes in the 2020 election, but hadn’t voted in the 2016 election, people who in an election in 2020 that truly was very high turnout in historical terms, and certainly quite a bit higher in turnout overall than past presidential elections, that that voter is, in fact, more conservative. But African American voters are still a very democratic group such that you gained many more votes, even though you’re adding “voters” to the electorate who might be a little bit less supportive of Democrats. If you don’t tell that kind of story, you kind of miss the point from a practical perspective.
All these things can be true. I think working with these kind of groups is it’s really is akin to a peer review. In this case, we have relationships with close, allied organizations.
Matt Grossmann: What are the first analyses you expect to run after 2022? What are the big open questions that you think we’ll face this year?
Jonathan Robinson: We did a similar project in Virginia in 2021. One of the things that stuck out, and this was of true of our analysis of the 2018 elections, is that the political science understanding of midterm elections as being best characterized by a uniform swing among demographic groups, relatively speaking and across geographies, was something that was pretty apparent when we looked at the data there.
I mean, the interesting part obviously there is the reason why 2020 was so interesting was that there were these disparate trends going on. Maybe if you looked at just the election results in total, you would say, “Joe Biden outperformed Hillary Clinton,” but under the hood, you see all these different patterns. He did better with white, college-educated voters. He did slightly better with white voters without a degree. He did slightly worse with black voters, quite a bit in aggregate terms with Latino voters. It tells the story, but there’s a lot of things that are going on in various different directions. But in a uniform swing story, things are moving in this very relatively consistent direction that might be more of a sign of, again, a nationalized mood around elections or something like that. That’s what we saw in Virginia on top of a modest Republican turnout advantage.
I think the first thing we’re going to really, really look at is, I mean, we’ll have some, I think inkling about this from early voting returns perhaps is just is that relative turnout advantage Republicans had in the 2021 elections in Virginia, does that still hold? I think there is a theory now that, again, after the Dobbs’ decision that Democrats are more engaged, and thus, you might expect that gap in turnout where Republicans were more engaged and turned out at higher rates than Democrats, that gap might close.
That’s actually one of the first things that I’m most interested in because one of the main operational understandings of why a president’s party, why their vote share declines in the election following their one at the White House, in that midterm election, I think a good amount of it is an engaged opposition that benefits from thermostatic anger or engagement or that kind of thing. It has a turnout advantage.
Right now, we do see some evidence that Democrats are more engaged. I believe the enthusiasm gap in public surveys, how enthusiastic you are about voting or vote likelihood has narrowed to be more even, whereas previously Republicans had an advantage. Then of course, in some of these low turnout special elections that had been more favorable to Republicans in the past, we have only four, unfortunately, but Democrats have done better in those.
That is a big component. The partisan turnout environment is something that we’ll look at right away. It’s still a hard question to answer, but we’re working off of individual-level turnout data or early voting data from administrative sources. Are we seeing the trends that started in the 2016 election somewhat continued though in a different way in 2020, are we seeing those change? Are Democrats going to do about the same or slightly worse with Latino voters, for example, or does it look like it’s relatively unchanged? Or if they drop, do they drop more than other groups or do we basically see relatively uniform declines?
We’ll be able to see some of that hopefully in some of the early election results. It was certainly something you could see early on in the 2020 election results, large shifts in the Rio Grande Valley or in Miami Dade County presaged in some way the changes that we saw with Latino voters that we confirmed the extent of when we had everything together, but it certainly was able to tell the tale. You could look at precinct results or something like that from very Latino areas or very white without a four-year college degree areas and try to answer some of these questions. That’s certainly something that right away we’ll be looking at to try to tell some initial story of what’s going on and certainly compare that to what other sources are seeing the exit polls or the AP VoteCast or other people who are looking at other ways of estimating the electorate, whether it’s from standalone surveys like what Matt Baretto and his colleagues do.
Matt Grossmann: You said it’s like peer review, but you’re reviewed by all Democrats on the practitioners’ side are people who have more sympathies with Democrats. Even if you tried to go out of your way to find Republicans in academia, you would find mostly Democrats as well. There’s a pretty clear, stronger tilt and relationship between the practitioner and academic side on the Democratic side. How do you think that colors how this joint field of pracademics has developed and how different would our understandings look if we were all on the Republican side?
Jonathan Robinson: Hmm, that’s a very interesting question. I think the first part that comes to mind is just that close tie will mean that there will be some talented people who maybe might have gone into academia who are now practitioners. I think my colleague Yayir, like he’s a student of Andrew Gellman’s. He probably could do whatever he wanted, but now that there are more industry jobs, a good amount are in politics, he’s able to use his talents to leverage his knowledge and understanding and talents to improve our understanding of politics on the Democratic side.
There’s a number of other people who have had stints in democratic politics or have totally made a career change, like David Nickerson as an example, Aaron Hartmann, Aaron Strauss. These are all people who have academic backgrounds who have made contributions in Democratic politics. I think there’s that right away. Especially as the amount of data in politics has grown, I think the Democrats have been able to leverage those types of people with those types of skills to do that kind of work. I think that is a very clear component.
I don’t really know a ton about Republicans. I mean, I have tracked academic papers that have been done either in coordination with conservative groups or by Republican scholars. They do attempt to try to put together a similar methodology. There’s a field experiment done by the Cato Institute where they send pocket copies of the Constitution to students and see if their views about liberty change. I don’t know anything about this project, but I’m sure it’s something like the Cato Institute always sends pocket Constitutions to students on college campuses or in a direct mailing. We would love to know what the impact of that is. They’re trying to answer similar questions.
One area where I do think Democrats do have an advantage in this case is in survey research. I just think the changes and the underlying environment of survey response have just made it a lot more difficult to analyze surveys. I think a really straightforward, rigorous, real way without a strong background in statistics and this auxiliary data that we have in the voter file. If you don’t have that kind of talent, I think you do end up having a harder time dealing with things in this kind of environment.
There used to be a conference, I mean, it still does exist sometimes, called Brute’s Camp. There would always be a couple of Republicans who would come and listen and see what people were doing or who were Republicans, but maybe said like, “I’m a Yale student,” or something like that, but really they’re a consultant for the NRSC or something like that. They’re coming in to try to learn stuff about what Democrats are doing because there just isn’t that level of people or insight or enough people who have bought into that. In some way, the cultures are very different.
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 discussion, I recommend checking out these episodes: How Much are Polls Misrepresenting Americans, The Role of Political Science in American Life, The Past and Future of Polling, Is Demographic and Geographic Polarization Overstated, and Compromise Still Works in Congress and With Voters. Please check out When Mass Opinion Goes to the Ballot Box and then listen in next time.
Photo Credit: iStock