Although the 2020 presidential candidates are investing huge shares of their time and resources in Iowa and New Hampshire, new research suggests early-state momentum may not matter much in our nationalized presidential race. John Sides finds that Donald Trump dominated media coverage well before election results in 2016, crowding out his opponents. Marc Trussler finds that state election victories didn’t seem to cause bigger-than-normal shifts in polls in 2016, with any campaign day just as likely to see an influential media event. Momentum may be dying with the growth of pre-primary media coverage and an earlier cementing of candidate coalitions.

Studies: Identity Crisis; “Knockout Blows or the Status Quo?” 

Interviews: John Sides, Vanderbilt University; Marc Trussler, Washington University


Matt Grossmann: This week on the science of politics, do early primary States still pick presidents? For the Niskanen Center, I’m Matt Grossmann. The 2020 presidential candidates are investing huge shares of their time and resources in Iowa and New Hampshire based on historical examples where they’ve seemingly changed the race. But new research suggests early state momentum may not matter much in our nationalized presidential race. If voters already know the candidates and demographic and ideological coalitions have developed, they may not learn as much from the early votes.

Today I talk with John Sides of Vanderbilt University about the nomination sections of his Princeton book with Michael Tesler and Lynn Vavreck, Identity Crisis. He finds that Donald Trump dominated media coverage well before election results in 2016, crowding out his opponents. Bernie Sanders did benefit from some early wins, but not enough to keep it close. I also talked to Marc Trussler of Washington University about his new Journal of Politics article with Joshua Clinton and Andrew Englehardt, “Knockout Blows or the Status Quo?”

They show that state election victories didn’t seem to cause bigger than normal shifts in polls in 2016 with any campaign day just as likely to see an influential media event. Momentum may be dying. Both focused on 2016. Sides, Tesler, and Vavreck found that Trump media coverage drove Republicans and elite consensus drove Democrats.

John Sides: On the Republican side, one of the biggest takeaways was just the lack of elite consensus. The fact that party leaders within the GOP largely didn’t endorse any of the presidential candidates despite there being so many, and ultimately never could achieve consensus either before or even really during the primaries on any particular front runner. That really opened the door for a candidate like Trump to get further than they typically would. It also opened the door for the volume of news coverage that Trump got to be important in helping him get the nomination, because there was just no alternative source of information coming to voters that was communicating to them who the candidates were that were worth paying attention to and who was the proper alternative to Trump.

On the Democratic side, I think the elite consensus was clear. I mean, Clinton had locked up so much support from elected Democratic leaders. She was really an outlier. She was an outlier in terms of how much support she had. And I think that support is what helped to get candidates out of the race, unlike on the Republican side when so many got in.

I think the thing that was similar between the two races is that you had a candidate, Sanders in the democratic race, who was able to mount a more significant challenge to Clinton than many people thought he could. Again, he did so on the basis of being able to generate a great deal of positive news coverage, which contrasted very much with the less positive coverage that she was receiving. And that pushed him not to a position where I think he was never likely to win, but it certainly helped him overachieve.

Matt Grossmann: Trussler, Clinton, and Engelhardt found that there was not much evidence of momentum at all from elections in 2016.

Marc Trussler: You know, I think the first thing here is that the word momentum is used pretty much for any pole movement in the primary. In a way that’s kind of rendered it useless as a term. So in our paper here, when we’re talking about momentum, we’re really focusing narrowly on one definition and that’s whether the act of winning and losing primaries is crucial for building candidate support. So we can imagine, for example, a candidate with a surprising win in Iowa and telling a story about voters witnessing this result and using that information to update their support for candidates. So that’s really the question we’re trying to answer here. And in short, when looking at 2016, we really find no evidence of this. We find no evidence that winning and losing primary elections was a unique and powerful driver of support for candidates. So how we determine this is by using pretty much a huge amount of data that was collected by SurveyMonkey.

So SurveyMonkey, drawing from their pool of people answering lunch surveys, averaged around 1600 respondents per day during the primaries, which is more data on the primary elections than anyone’s ever collected before. So with this, we were kind of able to ask two questions.

  • First, given existing trends and candidate support, does a primary win or loss change levels of support?
  • Second, whether these changes are noticeably larger than changes from non-primary days.

And that was kind of the important second question that says, maybe there are significant results from these elections, but are they bigger than what we’re seeing on just other random days in the campaign? The answer to the first question is yes, election results do change people’s attitudes. But crucially, the answer to the second question is no, that these changes are not larger than what we’re seeing on just random other days during the campaign.

So voters are shifting their opinions and learning from the campaign but are not uniquely doing so from wins and losses in primaries. I think this matters, or we think this matters for two reasons. First it kind of speaks to how voters learn in election campaigns. So specifically there’s this existing neat, rational explanation for momentum as a very rational heuristic. So instead of something like identity driving voting decisions, we have people rationally responding to wins and losses and making decisions about electability. So like so many other things in political science, we were finding scant evidence for this kind of ideal, independent, dispassionate voter. And the second reason why we think this matters is if this sort of momentum exists, it would make the order of the primary elections matter. So if wins and losses in early states dictate the future success of candidates, then it matters that demographically distinct states such as Iowa and New Hampshire go first and help shape outcomes. Our results suggest that because people aren’t uniquely learning from these elections, that this order doesn’t matter all that much.

Matt Grossmann: Sides was building on an earlier book with Vavreck in 2012 where they found similar patterns in the effects of media attention and elite consensus.

John Sides: One of the big differences between the 2012 Republican primary and the 2016 primary, I’d say there was just more consensus among elites. Not an overwhelming consensus, but it was clear that even though lots of leaders didn’t endorse any candidate and there wasn’t a front runner as clear as, for example, George W. Bush in 2000, that nevertheless, Romney had the majority of those endorsements and really didn’t have a significant challenger that could seemingly provide a consensus choice across different factions of the party. One of the things that was similar in both races is that you could see moments in time where candidates were able to get some episodic attention from the news media, but then ultimately lose that and see their candidacy kind of fade from view.

In 2012 there was several cycles of this where candidates like Rick Perry and Herman Cain and Rick Santorum and Newt Gingrich would experience what we call a cycle of discovery, where you’ve been in the race for a while, but now the media suddenly has a reason to pay attention to you, gives you a bunch of attention, your poll numbers go up, but then you end up experiencing the second stage of scrutiny, where the media examines your record and your biography more closely and usually finds a number of things that are potentially objectionable, and so the coverage becomes more critical. And then you end up ultimately in the third stage which is decline, which is driven either by that scrutiny, if it’s ultimately fatal to your candidacy, or by the discovery of a new candidate. That happened several times in 2012. It also happened in 2016 for candidates like Carly Fiorina and for Ben Carson. You know, Trump never experienced exactly that cycle. Neither did Romney, for that matter, in 2012, but you can still see some of the same media dynamics play out across those two elections.

Matt Grossmann: Trussler was taking advantage of a wealth of new day to day data on candidate support during the primaries.

Marc Trussler: But the genesis of this project was really a bit backwards. We had access to these data from the primaries. Again, just a really unprecedented number of respondents, especially during the primary campaign. And really our our, our question was, what sort of question can we ask of these data that can help us make an impact to our understanding of primaries?

And it became clear to us that with this level of data and this granularity, day by day granularity, that our comparative advantage was being able to look at the impact of certain events at the individual level and in a way that hadn’t been done before. And so we add that to just kind of our frustration in the coverage that everything in anything was called momentum. And that really led directly to our research question of whether election wins and losses uniquely matter. In some ways, though, this data-first approach, though, it was beneficial because we really didn’t have a strong prior going in to whether we were going to find momentum or not. And I think on balance we probably all thought that we would have found effects and we didn’t.

Matt Grossmann: Sides places the last race in the context of a long conversation about whether the media would replace elites in choosing nominees.

John Sides: When the primary system first shifted after the reforms in the early 1970s to make delegates at the conventions selected on the basis of outcomes in primaries and caucuses, not about this proverbial smoke filled rooms and the convention of earlier years, people thought, okay, this is going to make the process completely a media driven process, because now it’s just candidates going off and appealing to voters. And the way that you appeal to voters is through different kinds of media. And so a lot’s going to ride on who the news covers and how the candidates are able to use public relations to their advantage. And I think initially, I mean, we would certainly say that that those fears were somewhat overstated, in part because there were a variety of ways in which party organizations saw fit to affect the actual outcome by throwing their weight behind certain candidates, marginalizing others, encouraging some candidates not to run in the first place.

So maybe the initial fears weren’t born out. But even though the parties were working to effect the outcome of these primaries and caucuses, it was still always the case that information mattered to voters. And there were many cases even early on, relatively speaking, let’s say in the 1980s, where candidates like Gary Hart benefited from what looks very much like a form of media driven momentum, where outcomes in primaries and caucuses would lead them to get additional news coverage, which would then help them rise in the polls. Now, not necessarily to win the nomination, but it would certainly change the dynamics of the nomination. And so to some extent, I think you can look at the results in 2012 and 2016, for example, in our books and see something of a continuation of the same theme.

It’s not that party organizations or party leaders are irrelevant to the process, although there are certainly years in which they are more successful at achieving consensus than other years, but throughout all that, you still see the potential for voters’ views to be significantly shaped by the amount and the tenor of the coverage that the media provides these candidates.

Matt Grossmann: Trussler says they were addressing the conventional wisdom that voters learn about the candidates from their early wins.

Marc Trussler: The conventional wisdom is that early wins, and especially wins in Iowa and New Hampshire, matter a great deal for affecting who people support. And I really see this conventional wisdom as an outgrowth of this kind of mid-century rational voter model or ideal, where citizens are independent and dispassionate voters who act rationally. And I think we want to believe that these early wins signal something about electability and voters respond. This really doesn’t seem to be the case, however, that people aren’t really responding to this.

And we even broke these results down, for example, by education level. So we might think, okay, people with lower levels of education, they might use this heuristic more. And we don’t find that. And we break it down by people who live in states who have not yet voted. And we say, okay, maybe those people who haven’t had the chance to vote yet, they’re still learning. They will be affected by this. But they’re not affected any more than that. So I think the conventional wisdom is that these early wins and losses in a world where voters are not paying a great deal of attention, serve as unique signals of electability or viability that people use to learn and update and change their support levels for candidates. But we’re not really finding that.

Matt Grossmann: They found only a few potential impacts from late surprising wins.

Marc Trussler: There are very few of our days were both correlated to election success and failure and had effect sizes that were larger than what we saw on non-primary election days. And so few of them that it might just be random chance. But let me talk about just who that did seem to matter in particular, and those were both the Oregon and California primaries on the Democratic side. Both of these were late primaries. In Oregon, sanders convincingly beat Clinton in a somewhat surprising way. And in California, Clinton beat Sanders by a greater than expected margin. So again, what do I mean when I say that these two matters? Well, these are contests where Clinton support changed significantly in the expected direction. So after Oregon, we estimated her support shifted compared to what it would have been in the absence of the primary by about eight or nine points.

So she lost about eight or nine points. After California, we saw her support rise compared to what it would have been with the absence of the primary by both the same amount. So first voters are shifting in the expected direction given the results of the primaries. Second, these shocks to her support are substantively larger than the range of shocks that we estimate for all other days in the primary season. So these days seem to stand out as significant shifts in candidate support, but it’s hard to justify these shocks, Oregon and California, two of the latest Democratic primaries with a story of momentum. They were both relatively late. And if we think momentum is a thing, it should be in the earliest primaries, when people are still forming their opinions, that’s when we expect that these election results should really shape beliefs about electability or viability.

Matt Grossmann: It wasn’t that nothing changes during the campaign, but that lots of campaign events matter. So elections don’t stand out.

Marc Trussler: Elections are really just one piece of information that voters receive, and in a different media environment, it’s possible that voters would use election results as an important signal of quality. But when we compare elections to other events, debates, or speeches or ads, or even just the ongoing daily hum of the campaign, it’s not clear that they matter in a special and unique way. It’s not the case that we find that each day has non-significant impacts on candidate support. We find that a large number of days have significant shifts in candidate support, significant as in statistically significant movements. So I think it’s closer to the truth that there’s just all sorts of events that changed the way that voters act. So voters-

All sorts of events that changed the way that voters act. So voters are learning and changing through the campaign and reacting to events. I think it’s more the case just that elections don’t stand out against these other events in a way that maybe we thought they did before. Rather than a world where nobody’s changing and everybody’s locked into their preferences.

Matt Grossmann: We’re now focused on the role of Iowa and New Hampshire in 2020. Sides is not ready to throw away the conventional bet to focus there, especially for those behind.

John Sides: I think Iowa and New Hampshire are … The outcomes there can have a measurable impact on the race, but they’re neither necessary nor sufficient for a candidate to win. You don’t have to win both those states to win the nomination. We saw that in both years. I do think however that for candidates who are lagging back somewhere in the pack, the fact that Iowa and New Hampshire … I mean Iowa, New Hampshire might be more necessary for them because without having some kind of early success when you’re not a front runner to begin with, it’s very difficult to, I think, get the kind of news, attention, and voter attention that you need to succeed in the later primaries and caucuses.

For someone like Rubio, doing better than expected in Iowa in 2016 was absolutely helpful to him. Unfortunately, couldn’t capitalize on that. Trump won relatively handily in New Hampshire, and ultimately then Rubio’s candidacy never caught on despite some late-breaking endorsements from party leaders. But you would still say that even if Iowa and New Hampshire victories aren’t necessary for the ultimate nominee, I think those races are pretty important for helping candidates break out somewhere in the middle of the pack. In a year in which the field is somewhat less conclusive, I think that could have a larger impact.

One of the interesting questions then about 2020 for the Democrats is if you go into the Iowa caucus and you’ve got two or three candidates polling somewhere between 15 and 25% and no one really clearly has a significant lead. Then obviously we know that the outcome could be somewhat different than what poll numbers are suggesting and that could have some interesting implications for what might happen after New Hampshire.

Matt Grossmann: Tresler says the candidates might be overestimating Iowa and New Hampshire or there could be other reasons beyond their effects on voters.

Marc Trussler: It’s certainly a possibility that there is just stickiness to how these campaigns are run. There could be a belief amongst the consulting class that momentum is real and is important. I don’t know. I don’t work in that field and I don’t know how they feel. But it’s possible that while these results and election results might not matter much for change in candidate support, they might matter for influence of elites and a certain class of donor, for example. It might also matter for the candidates own belief about whether they can win going forward. It seems unlikely that success or failure in a primary will necessarily lead to many voters changing their opinions. What seems more likely is that a candidate may use the primary result to decide whether to proceed and if they drop out. Then another candidate might appear to gain momentum from that, but only because they are choosing the next closest remaining candidate given voters’ preferences. Not because voters are using the results to change who they supported.

So it might not matter for changing support, but it might matter for candidates attracting more dollars and donations and might change their own belief about whether they have a viable campaign going forward.

Matt Grossmann: Sides didn’t see much early state movement in 2016 either.

John Sides: Well certainly in 2016 with the exception of Rubio getting a boost coming out of Iowa, you didn’t see a clear evidence that your success in the earliest primaries and caucuses created a durable shift in the polls in your direction. Do I want to say that that’s never going to happen again? I don’t think so. Again, it’s going to depend upon the specific configuration of the field in any particular year and whether Iowa and New Hampshire are able to generate surprising outcomes. I don’t think we’ve completely lost the element of surprise.

To the extent though that we might expect to see less of that kind of momentum shifts coming out of Iowa and New Hampshire, I mean one reason why that might be true is if there’s been so much coverage of the so-called invisible primary of the period leading up to Iowa and New Hampshire, then it’s possible the voters have gotten enough information that the outcomes in Iowa and New Hampshire are not going to be as likely to provide them with new information that they should update that would then create the kind of surge, let’s say that Barack Obama got after he won in Iowa in 2008. Maybe voters have already learned enough, and they’re more settled.

John Sides: But again, I’m always conscious that we have a relatively small sample size of these primaries every four years. I do think there are some secular trends like the growing amount of news coverage of the invisible primary, but I don’t know that that means that we should rule out the potential for Iowa and New Hampshire to matter. I mean, a related thing might be, well, if they matter a little bit less, are the candidates over-investing in those states? I mean, there’s always the possibility that candidates are being overly risk averse and spending more money than they need to spend. That’s an affliction of lots of candidates for lots of different levels of office. But I would say that with regard to Iowa and New Hampshire, for some candidates really that’s their only bet. If they don’t succeed there, they’re not going to really have much of a chance beyond Iowa and New Hampshire. Certainly for some of them it makes sense, particularly the underdog type candidates.

Matt Grossmann: They’re both cautious about drawing too many lessons from 2016. Tresler says Trump dominated 2016 and might’ve changed the role of early states. But it’s also possible that politics is now shifting into a news heavy, not so invisible pre-election primary period.

Marc Trussler: We can’t ignore Trump here. Having Trump in the election is obviously enormously different, and there’s likely a good amount of merit to the argument that Trump caused people to be constantly paying attention to politics every day of the campaign in a way that would dull the effects of elections in particular in driving the way that people make decisions. But maybe let’s think about what would be the ideal case for momentum. So imagine a world where politics was extremely low salience. The only thing that people learned was who won or lost an election. That is people were only checking in on this primary occasionally. So these wins and losses would play a large role in how they come to evaluate the candidates. I think this is a world where we would see a lot of momentum, but that’s very different from the 2016 world.

In 2016, politics was extremely high salience. People were being constantly bombarded with messaging. In this world, it’s harder, of course, for election results to stand out as uniquely important. I think the big question going forward is what world are we going to live in? Are we going to live in one where politics is low salience, so the result of the Iowa caucus might be the first time that people check in and are therefore very highly affected by it? Or are we going to live in a world in which people are constantly bombarded with political news? If I had to put money on one, I think it’s going to be the latter. I don’t think that politics is going to drop from its level of salience. I don’t think we’re going to live in a less concentrated or a less salient media environment. If that’s the case, if politics will continue to be salient going forward, if people are going to be even more so constantly bombarded with political information, I think these results from 2016 are pretty informative going forward.

Matt Grossmann: Sides says Trump media coverage mattered a lot, but because it directed people to him who already agreed with him.

John Sides: I think the media coverage was absolutely essential and it helped put him in a very strong position. He certainly was not the candidate that was supported by even a majority of the party at the time when the primaries and caucuses began. But he had a clear plurality of support and moreover, there was no candidate that was really a clear second favorite or a second choice. What we would argue is that in a field like 2016 with so many candidates and no clear front runner, voters need news coverage to tell them where to focus their attention. So that’s in part what the news coverage was doing when it gave Trump so much air time and made it harder for any of the other candidates to get the oxygen that they needed to break through the pack.

In addition, what that information is doing is it’s helping voters understand more about who Trump is and it’s helping him build support among the elements of the party that had political attitudes that were in line with his. So if you are a voter that had particularly strong concerns about immigration, then you got to learn by dent of this news coverage that Trump was the candidate you should support. Or if you are a Republican voter who had more liberal views on economics than many of the elected leaders of the party, now you learn that Trump maybe was the candidate for that too because he was willing to talk about protecting social security and Medicare. Part of what the media coverage is doing is just getting your name in the headlines, but part of what it’s also doing is helping voters realize which candidate is the one that’s best aligned with how I think about politics.

Matt Grossmann: Trump was helped by taking outlandish positions that stood out in his party.

John Sides: I think one of the things that helped Trump was that he was willing to take stances on immigration and related issues from not just immigration from Mexico but also Muslims both traveling to this country, Muslim immigrants living in this country. That whole set of issues. He was willing to take positions that were far more conservative than anybody in the party who was a plausible presidential candidate was willing to take. I mean, Cruz was thought of as a fairly hawkish figure on immigration, but among the people that had the most heightened concern about immigration, the most opposition to immigration, we’ve found that those voters broke away from Cruz and toward Trump. Had there been consensus among the Republican candidates, had they all agreed with Trump, then it would have been harder I think for voters to choose among those candidates on the basis of those issues.

The same thing happened on economic issues too where Trump was willing to go against party orthodoxy and his views about government spending and taxation. Now, it didn’t happen. It’s not the case that’s how he’s governed, but at the time he was certainly making noises that a Republican candidate typically would never make suggesting they could raise taxes on the wealthy and protect entitlement programs.

The difference I think on the Democratic side was, although it was fashionable to try to find daylight between Sanders and Clinton, particularly on questions about the government’s role in providing social welfare, whether it’s healthcare or childcare or something. Clinton and Sanders weren’t that far apart. The differences were really differences of degree rather than kind. So we’ve found that even though voters who call themselves liberal or very liberal were more likely to support Sanders than they were Clinton. Your positions on issues, the issues that candidates were extensively debating didn’t really predict your choice because actually there wasn’t as much difference between the candidates as you would have thought. Similarly, there wasn’t much difference between Clinton and Sanders and their view of issues tied to race and immigration. Your attitudes on those issues didn’t predict your choice between Clinton and Sanders the way they predicted your choice between Trump and the other Republicans.

Matt Grossmann: Tresler says Trump might have affected both parties last time, but he might also be a sign of things to come.

Marc Trussler: The ideal world where momentum exists is one where there’s a low salience political environment, and these elections are very present and salient to people and maybe it’s the first time that they checked into the election. Obviously with Trump sucking up so much oxygen, it was hard for that to be the case. That people were constantly checking into politics and it was really dulling the effect of anything else that wasn’t Trump. Even on the democratic side, the presence of Trump in the news would make the election results in Iowa less salient in people’s information environment. Again, though, we’re not going to have Trump forever sucking up air in these things.

But looking forward, do I expect to see a media environment where politics has low salience and we have a large number of people checking in for the first time at the Iowa caucus? I would say no, that we’re not. Even without Trump, we’re living in a political world where people are going to have thought about and engaged with information about the primaries far before Iowa, which really dulls the ability for something like the Iowa caucus or New Hampshire to affect the results or to change people’s opinions about these candidates.

Matt Grossmann: Sides agrees that the invisible primary now has more media and money that weaken the old gatekeepers.

John Sides: Well, I mean the party decides evidence which goes through 2004 suggests that that party endorsements were not just bandwagoning. They weren’t just following the candidates that were already being talked about in news coverage. But I think even in those data before the more recent primaries, you can see that, again, there are years in which the field of candidates doesn’t lend itself to a clear front liner. Has it gotten worse? I think there’s certainly a sense now that the invisible primaries, the amount of coverage in the invisible primary is increased. So there’s broader set of information that’s reaching voters that in some sense makes the conversations among party leaders public in a way they didn’t use to be. There’s also a potential to earn money and to raise money in your campaign in a way that didn’t used to be the case. Sanders is a good example of that in the amount of money he was able to raise, largely through individual small donors and stuff like that outside of traditional party networks and wealthy donors and other kinds of gatekeepers.

If those trends continue, then you would expect there to be these continued centripetal forces within primaries, which put more power outside of party leaders and put it in the hands of candidates and voters. To be honest, I’d be more confident in making assertions or predictions if we had at least the chance to see a few more primaries.

Matt Grossmann: But Tresler says that elite driven consensus could still limit campaign dynamics.

Marc Trussler: Unified elite information behind a candidate might provide the sort of information that could protect against a potential poor primary result. We can ask, for example, in what ways are elites providing these in between contest cues that are filling the information environment and kind of dulling the effects of these elections? When people are paying attention to the daily grind of the campaign, it might be those party elites shaping the message. So that, for example, might help us rethink some classic cases of momentum. Every time that I talk about this paper, someone’s going to bring up what happened in 2008 after Obama won Iowa.

That was really the primary case for momentum. Is it the case that voters use the signal of his win to change their opinions? That’s possible. Another possibility is maybe the win signaled his viability not to voters in general, but to party elites who were then able to fill the airwaves, not just right after Iowa but throughout, the campaign, for the rest of the campaign, with pro-Obama messaging. Our results, I think, show that election drivers are not unique supporters of opinion change and suggests maybe that elite driven story about wins in Iowa changing elite support for candidates or elite beliefs about viability and then having kind of a two step flow from there to the information environment to voters, that might be closer to the truth than voters directly making decisions about viability and electability based on the results of an election.

Matt Grossmann: Their findings on Bernie Sanders in 2016 are a bit different. Trussler says, “Sanders didn’t really gain much from his wins just from the race gaining attention over time.”

Marc Trussler: If the first election was on super Tuesday and there was the same level of scrutiny, of advertisement, of discussion of Sanders, would that be remarkably different from having witnessed the actual election events? As far as our data goes, we would say that there’s no difference between those two things. That yes there was learning, yes Sanders had media attention, but those weren’t being primarily driven by the elections, but just the constant intention of the campaign. Comparing those two things, whether there was elections or not, from our data, generalizing from our data, we don’t see a large difference in what would have happened to Sanders with and without those elections.

Matt Grossmann: Sides says, “Sanders really never had much of a chance but was running to influence the party.”

John Sides: If you go back and you look at the prediction markets, they were predicting the democratic nomination. He really didn’t move very much. I don’t think prediction markets are perfect, but you really have to ask the question why weren’t the markets essentially lowering Clinton’s chances of winning and raising Sanders chances of winning? I think it’s because you could see fairly quickly that despite these occasional successes the nature of the delegate selection process and the fact that the democratic party allocates delegates proportionally means that after the first few primaries and caucuses, it was clear he was not going to catch up. He was going to have to win, not just a victory, but an overwhelming victory in a number of states to make up the difference and this is even before we think about what the super delegates in the democratic side might’ve done or not though.

For Sanders it was clear that he wasn’t really in it just to win it, he was in it to have an influence on the party or that’s how I interpret his decision to stay in the race as long as he did, even after it was more than clear that he was not going to have a majority of the delegates that were awarded as a consequence of the primaries and caucuses. I think he was making a play for influence in the party and probably was somewhat successful in the way that the convention went and the way in which some of Sander’s ideas have become somewhat more widely accepted in the party.

Matt Grossmann: Trussler found that demographics predicted the vote from the beginning and didn’t weaken in influence.

Marc Trussler: Political identity is not just the sum of your demographic characteristics, but as a rough proxy, the way that we use demographics is to try to get identity. We do find that when we use people’s demographics to predict their vote, we don’t really lose prediction power over the course of the election. The same type of people were supporting candidates throughout the entire election, which really puts a hard cap on the ability for momentum to shape things. It really points to identity as a heuristic, as shaping candidate support rather than kind of people rationally, “Rationally going into the election and using election results as a way to make their decisions.”

Matt Grossmann: Sides found that the big demographic differences matched the normal outsider insider dynamics for democrats.

John Sides: The main differences between Clinton and Sanders supporters were fairly straight forward. Clinton had the support of Democrats, self-identified Democrats, African-Americans, a very loyal and highly democratic identified group. Sanders had the support of younger people, the support of the people who didn’t identify as strongly with the democratic party or identified as politically independent. Some of those patterns are fairly predictable. Anytime a candidate is sort of an alleged outsider challenging the so called party establishment, they often do better with voters who aren’t themselves very aligned with that party. Similarly candidates who come along and advocate this sort of revolutionary type of change, typically will do better among younger voters than candidates who advocate for more modest or more incremental kinds of policy goals. Those patterns were fairly predictable.

You see a bit of that same kind of difference this year, a candidate like Biden who’s highly identified with a party and of course identified with Barack Obama, is doing better among the traditional constituencies within the democratic party, particularly with African Americans, at least today. He’s not doing as well with young people who gravitate more toward a candidate who likes Sanders who retains some of the appeal that he had in 2016, although he faces more competition. At the moment I think the field is still more unsettled, right? There’s just more candidates and it’s harder to get the same kinds of strong alignments between demographic characteristics and your vote when there are multiple candidates competing relative to just having two. Two to simplifies the choice.

Matt Grossmann: What about 2020? Trussler sees signs that candidate appeals might already be settling in.

Marc Trussler: I think what I would say about 2020 is the things that we should care about are happening right now. The way that people are forming their opinions of candidates, the way that candidates are shaping their message and appealing to different identity groups and how those differences in identity are getting locked in to support for different candidates or groups of candidates. That’s happening right now. That’s going to have a large amount of weight on what happens in the election. In general I would say, stop worrying so much about what’s going to happen in Iowa and how that’s going to shape things and start worrying more about how candidate appeals to identity and their positioning is shaping the future of the campaign now in this kind of pre-primary stage.

Matt Grossmann:          Sides says the prediction markets are mostly pricing in the remaining uncertainty.

John Sides: Right now the prediction markets have Warren and Biden priced, essentially the same, Warren slightly ahead. That’s pretty much in line with where the polls are nationally. The most recent polls have had them pretty much neck and neck. There’s some variability here and there, but I wouldn’t say it’s unreasonable to think of them as moving closer to a tie. In the experience of 2016, if the markets are pricing in the experience of 2016, they’re trying to avoid going all in on the more establishment candidate and that would be Biden, I suppose and they’re trying to be sensitive to the fact that the challengers maybe come more from the ideological wings of the party can be successful.

The other thing I think that maybe valid and way the market’s pricing this in is that we’re still at a point in terms of where party leaders themselves come down that we don’t have a large number of endorsements. Biden leads, overall in the number and the prominence of the people that have endorsed him, but by no means is there any kind of consensus like that, that existed for Hillary Clinton. If you were to look at the landscape and you’re saying there’s lots of candidates, there’s not a clear front-runner in terms of endorsements or support within the party establishment, the polls are tied for the most part. It seems reasonable to me to think that the prediction market should have Biden sort of tied for the lead rather than have him as a clear favorite.

Matt Grossmann: The media will play a big role. It does follow real news, but it also makes interpretations that matter for how voters see the candidates.

John Sides: When Warren announced her plan for financing her version of Medicare For All, that got news coverage and I think that makes sense. I think there are other cases in which you would say that the news media is judgment is important in terms of elevating which kinds of events get coverage, what really is judged to be newsworthy or not newsworthy. Then moreover, to provide some kinds of interpretation for what those events mean and which candidates those events benefit. I think debates are a good example of that because to some extent who won and who lost and who had a good night or a bad night or a good moment or a bad moment, that involves some kinds of subjective decision making and there’s evidence from political science that voters are shaped by the interpretations that the news outlets provide about debates. We outsource some of that interpretive work.

I don’t think that there’s any way in which the success or failure of candidates is completely a construction of news coverage. I don’t think that’s true. At the same time, I do think it’s true that the choices that media outlets make are important for whether candidates get news coverage and for what kinds of news coverage they get and that reflects the values and the incentives that news outlets have. I usually find it more helpful to point out the ways in which the choices of news outlets matter, because I don’t think that that’s something you’re going to see in news coverage itself. No reporter’s going to write a story telling you that the story is going to change your mind or it’s going to influence the race. The reporter’s going to think they’re describing something that has already happened. I do feel that when you add up the collective decisions of a set of news outlets, it does have a meaningful impact on the way that voters see the race.

Matt Grossmann: Elections still help inform voters, Trussler finds, driving up their candidate searches on election days but don’t necessarily give voters new information.

Marc Trussler: I think it’s entirely likely that part of the reason that we’re seeing this is attention is likely variable based on attachment to candidates. People who are more attached to candidates might be the people who are driving search results after campaigns and those people are not going to be the ones that are changing. More generally though I would simply say that while elections drive attention, they’re happening in an information context that is so rich with information and has been rich with information far before the campaign for so long that it’s very unlikely that whether these campaign events, whether people are searching or not are providing voters with new and important information, given the huge stock of information that they already have gained from a very, very long campaign, even before the Iowa caucus.

Matt Grossmann: They both see growing role of social identities as the road forward for research. Sides says inter-party coalitions aren’t necessarily hardening but candidate specific coalitions may be building earlier in the process.

John Sides: I think the coalitions that exist within the parties are highly dependent on who the candidates are. Hillary Clinton’s coalition in 2016 was so different than her coalition in 2008, within the party, which had to do with just whether… Because her opponent was an African American in 2008 and not in 2016 was a big part of that story. Across election cycles, I don’t think we’re seeing hardening coalitions. In my mind the question, is the timing of when those coalitions form changing and are those coalitions forming earlier in the election cycle rather than later? Are we going to see more movement in people’s preferences before we get to Iowa and New Hampshire and less afterward, which connects to the question about the role that Iowa and New Hampshire play in terms of generating change or momentum for individual candidates.

To the extent that voters are getting a lot of information about the candidates and paying a little bit more attention a year before the general election than you might expect by about this time, November of the year before the election, that some of these coalitions Biden, Warren, Sanders et cetera have started to form and it’ll be a little bit harder to move them than it would have been an earlier election cycles with less information coming to voters. That’s a plausible hypothesis. Whether that’s true, I guess really would require some serious empirical testing and I think it would also require some additional election cycles to make sure that’s really the case. I’m also conscious that given there could be yet a few unexpected twists and turns even in 2020 that might lead to some shifts in these coalitions, that I don’t want to suggest that the cake is entirely baked.

Matt Grossmann: Trussler will also be looking at the role of identities in primaries where partisanship can’t guide voters to a candidate.

Marc Trussler: The big thing is going to be identity and this is work that Sides and Fabric and Tesla have done and I hope they have plans to do more. We really skirt around the idea of identity in the paper by looking at demographics to predict candidate success. This measure, using demographics, is crude and we know that mere demographic description doesn’t capture identity. I’d love to know more about how people’s social identities shape their attachment to candidates, especially in the absence of party cues. How that’s happening in the early stage and how those identities create strong allegiances to candidates or group of candidates. An identity and an attachment that, of our data at least, seems much stronger than the signal received from election wins and losses.

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. Thanks to John Sides and Mark Trussler for joining me. Please check out Identity Crisis and Knockout Blows or the Status Quo and then listen in next time.

Photo credit: Excel23 under CC Attribution-Share Alike 4.0 International