Brendan Nyhan spends a lot of time researching America’s political polarization and the strength of its democracy with the organization he founded, Bright Line Watch. In part 2 of our interview with him, he tells us how questions about the state of America’s democracy really need to be put in the context that we didn’t really have an honest, full democracy until the Civil Rights revolution of the 1960s. (Listen back to part 1)
In this episode, we discuss the following work:
- “Fake News and Bots May Be Worrisome, but Their Political Power Is Overblown” by Brendan Nyhan (New York Times, 2018)
- “The Minimal Persuasive Effects of Campaign Contact in General Elections: Evidence from 49 Field Experiments” by Joshua Kalla and David E. Broockman (American Political Science Review, 2017)
- Tyranny of the Minority by Steven Levitsky and Daniel Ziblatt (Crown Publishing Group, 2023)
- “A Blueprint for the Regulation of Tech” by Drew Dimmery (self-published on Substack, 2023)
There’s a great summary paper. It’s Kalla and Broockman 2017. It’s a meta-analysis of campaign contact experiments. I actually found it through an op-ed that you wrote in the New York Times. they look across 40 plus experiments and they basically find zero political persuasion effects.
Yes, yes. The meta analytical result is a zero effect in all these different kinds of studies of persuasion. Then there’s a great follow up study by Alex Koppig and his co authors called because sometimes people say, well, that’s true, but I know the effects are different because XYZ and they ran a study titled “The Small Effects of Political Advertising are Small, Regardless of Context, Message, Sender or Receiver, Evidence from 59 Real-Time Randomized Experiments.” So they tested 59 ads that were run during the, I believe, 2020 campaign in the context in which they were relevant and just minimal effect after minimal effect. So taking this back to social media, what does this mean?
I think it means we’ve made a kind of inferential mistake, which is social media makes visible many of the worst aspects of our polarized politics. We observe polarization there very directly. And we observe some of the most politically engaged among us, a very unrepresentative subset of people as we’ve talked about. And we infer from that the social media made them that way or made us that way.
What these kinds of findings suggest, whether it’s the ad studies or the social media studies is that political polarization has deeper roots. It’s not simply a media effect story, whether it’s TV or social media. People’s attitudes really are harder to change. And in some ways that’s good because it means it lowers the stakes in some of the questions around social media.
I would argue actually we should reorient the conversation around the harms of social media towards specific kinds of harms like threats to democracy, like undermining vaccination efforts during a pandemic, like targeted forms of hate and violence that take among subsets of extremists who consume lots of like-minded content online. That’s a very different set of problems and questions than these generalized persuasion and polarization effects that so much of the conversation has focused on, especially in the last few years.
So I’m really hoping we can start to make that pivot and say, we’re not polarized because of social media. There are demonstrable harms that are associated with the kinds of content that circulates there, but it is not the cause of our problems in some generic sense, even though it’s such the most visible manifestation of them in some ways.
In some ways, it sort of feels like this suspicion of social media is almost a cognitive shortcut. So if we’re feeling very uncomfortable about the state of American democracy—and I think a lot of us are—the class I teach later today is called “Defending Democracy in a Digital Age.” It’s really, you know, front and center on a lot of people’s minds. It feels like it should be as simple as, well, let’s unplug this thing. Or if we can unplug this thing, let’s try these simple fixes. We’ll get rid of the algorithm. We’ll force people to pay more attention to people across the aisle. And you are lead author on one, co-author on other studies that are making, I find, a very persuasive case that it’s just not that simple. And my guess is that if you could do it for six months, if you could do it for 12 months, you probably see similar effects.
I think this theory that what’s happening is that social media is actually surfacing that polarization is a terrific hypothesis. I think there’s also, as some people have pointed out, there’s lots and lots of other partisan media out there. People are choosing Fox News or MSNBC. People are choosing the politicians that they’re listening to who are their own deeply partisan voices.
Brendan, what’s actually shaping our political beliefs? If we actually wanted to figure out why partisanship is at such an apparent high point in America at the moment, if it’s not media effects, what is it? What has brought us to this particular moment in time?
Yeah. Well, we could have a whole, and we do have political, whole political science class is just about that question. I will not be able to fully do justice to it.
Let me make one small point first, which is you’re absolutely right to talk about other partisan media. And there’s, I would say provocative evidence around the rollout of Fox News suggesting that it can create vote shifts on the margin at the aggregate level. But I would definitely encourage your listeners to check out David Broockman and Josh Kalla’s experiment where they paid Fox News, Fox News heavy, primetime Fox News viewers to turn off Fox News for a month. So it’s actually quite parallel to our study in the sense of it’s taking the kind of content we’re thinking might have problematic effects, it’s reducing its prevalence, and then it’s examining what effect that has on attitudinal measures.
Like our study, it is not consistent with a maximalist view of Fox News effects either. In some ways, it was interesting that their participants updated about Fox News and how much they could trust what they saw on it, but it did not make them reevaluate their politics. The effects are much more modest and nuanced than that. So I think I would definitely direct people to that as a kind of compliment the conversation we’re having here.
But I wanna get to your question about the sources of polarization. The story I would tell, and this is a multi-decade story. So this is the frustratingly long-term and structural political science explanation. That means there are no simple answers. Again, part of what’s appealing about social media is to suggest there are dials we could turn, right? Or we could unplug the machine, right? We could unplug the servers the way Elon Musk did in Sacramento, but just never plugged them back in, right? If that were, if only it were that simple. If you actually look at the history of American politics, What you see is we have typically been highly partisan, our politics have typically been highly partisan and conflictual.
The exception is the post World War II period around which most of our civic and political norms are based. That era was the aberration, not the norm. We are thinking of ourselves as having lost what was normal, right? But in fact, we are returning to normalcy in the sense that our politics are becoming highly polarized and contentious the same way they were, for instance, in the late 19th century before the changes that I’ll describe that made our politics less polarized in the mid-20th century.
Okay, so then you might be saying, well, that sounds great. How do we get back to what happened in the mid-20th century? And I’m here to tell you, it is both not possible to return to that world, nor would we want to, because it was foundationally built on the terrible history of race in this country, which the parties essentially conspired to keep off the national agenda. As a result, they were internally divided in ways that reduced partisan polarization artificially.
After the Civil Rights Revolution—the United States finally becoming a full-fledged democracy—that cross-cutting issue is unwound. And in the process, we go from essentially what was almost a four-party system, because of, for instance, the Southern Democrats in particular, to we returned to two highly polarized parties. And that process has been working its way through our politics now for decades.
And if you look at the data, it’s not consistent with a story where our politics were relatively non-conflictual and then all of a sudden social media shows up and whoop, things take off, right? The data are not consistent with that kind of story. I’m certainly not saying social media aren’t consequential for politics in many ways, but the story of polarization is a much longer term trend.
So I guess what I would say to you is, we could go on about the nuances of that story, but what I would say to you is, I am less sanguine about reducing polarization and more optimistic about the prospects for institutional reforms in the way our political and electoral system works that would better handle the levels of polarization we should expect in a multi-racial democracy like our own.
We have brittle and outdated institutions. The political scientists, Steven Levitsky and Daniel Ziblatt have a book coming out discussing this issue. The political scientist Lee Drutman has written a lot about how our institutions simply don’t function well under polarization. That’s a bit far afield for today, but I’ll just say, I think in many ways, we’re looking to the wrong, we’re looking in the wrong places for solutions.
So in other words, to sort of read that back to you. Yes, polarization is high. Polarization has mostly been high in American history. It is perhaps inherent in a country that has largely enshrined a two-party system, not built into the Constitution, but certainly a very strong historical tendency.
The main exception to this is this sort of high point of American democracy that tended to be idealized by really our parents’ generation, which is this post-war moment, which has some really attractive things: has greatly increased equality, has a lot of infrastructural investment. But buried beneath it is an enormous amount of raw, unreconstructed racism, failure to deal with race issues, essentially Northern and Southern versions of both the Republican and Democratic parties. And as you say, enough internal party division between Dixiecrats and Democrats, for instance, that you’re essentially dealing with something closer to a multi party parliamentary system, then to the system that we have now.
We’re back into a system that frankly, would have looked a lot more familiar to our 19th century ancestors. And now the question is whether we can survive that level of polarization with some very creaky institutions.
And as a multiracial democracy, which we weren’t at the time. Right? Right. Right.
Yeah. Right. So and as you pointed out, you know, the United States, by most external measures, is not viewed as a full democracy until after the Civil Rights Revolution and increasing enfranchisement of people of color, some of which is now being rolled back.
You are now working with this project called Bright Line Watch. Bright Line Watch, as I understand it, is a group of political scientists looking at how the United States is weathering as a democracy. How did this project start out and what are you finding?
Yeah, so this is a collaboration between myself and three other political scientists, Gretchen Helmke at Rochester, Sue Stokes at Chicago, and my colleague here at Dartmouth, John Carey. The three of them study democracy and democratic erosion and breakdown around the world, particularly in Latin America. And they were, like me, quite alarmed at what they saw in 2015 and 2016 as Donald Trump’s candidacy took off. I was quite outspoken in raising the alarm about what I was seeing. They were similarly outspoken and they organized a petition of political scientists, raising concerns as political scientists about the signs we were seeing during the campaign.
After the election, when Trump won, they were able to make connections with funders and essentially, and they asked me to join the effort and we formed this organization. And ever since we’ve been been conducting expert and public surveys tracking both the state of American democracy and the level of polarization in particular over it. Because the concern is that public opinion is one of the constraints on anti-democratic behavior by politicians.
And what we saw with Trump was that that bulwark seemed to be failing, that Trump was not paying the expected political price for violating various small-d democratic norms. And so we thought it was essential not just to assess the state of American democracy via these regular expert surveys, but to check in with the public and to see how they understood those same events. And so we’ve done a large number of surveys now since 2017. We’ve been tracking that issue ever since.
So the survey that’s featured on your site right now feels particularly timely. We’re recording this at a moment where some portion of House Republicans seem interested in impeaching President Biden. It seems to a certain extent that they’re still trying to figure out why they’re impeaching President Biden. And there is not in fact a House vote to proceed with impeachment. However, impeachment investigations appear to be moving forward. We could have a whole conversation on just how crazy this is.
But the study that’s up on the site looks at how differently the left, the right, and experts look at charges that have been brought against Donald Trump and also concerns about President Biden’s handling of classified documents.
You find that the left and the right obviously view these situations very differently. Experts tend to align with the left and a feeling that in some of these cases President Trump likely has committed crimes. You seem to find a very high state of polarization.
At the same time, later in the same study, you actually talk about feeling fairly sanguine at this moment about the state of American democracy. Is that only vis-a-vis comparisons to Latin America or West Africa, or is that sort of more of this sort of manifestation of your feeling that we’re kind of returning to a highly polarized normal?
Yeah, that’s a good question. So let me just briefly summarize the finding for the listeners, and then I’ll talk about that point.
We conducted what I think still is the most detailed survey asking about how people understand the allegations in the various cases against Donald Trump, both their factual beliefs as well as the perceived severity of the stated actions. In other words, do you think this is a crime? Do you think he committed it? And so we have very detailed questions about his efforts to overturn the 2020 election, his actions around and on January 6th, the so-called, you know, the Stormy Daniels case and the business record falsification allegations associated with that, and then finally the classified documents case.
And what we find is very few Republicans think Trump committed crimes across all of these cases. And even on the classified documents case where there’s literally photographic evidence, the gulf between the parties is vast. On the other hand, we do find some evidence that at least a minority of Republicans are acknowledging the classified documents case in particular is is suggesting that Trump did commit crimes. It’s still a minority of Republicans, but there was some movement on that case relative to a prior survey.
So there’s some evidence of responsiveness on the margin, but there’s still a vast gulf between those. And one could certainly imagine the conversation shifting towards the cases related to Trump’s actions to overturn the 2020 election, both the federal case and the case in Georgia. Those are likely to recapitulate the same partisan divisions we’ve seen ever since election day. And I’m less optimistic than on the classified documents case that we might see any convergence. So that does seem alarming.
I think there’s definitely reason for alarm, especially given what Trump is suggesting he might do if he returns to office, which is to—he perceives this as a political persecution of him, and then he’s saying he will do that to Biden and the Democrats. Whether he’ll actually fall through on that, of course, who knows, he said many things that didn’t fall through on his last term. But we should expect if he wins, he will have a more unified party behind him than he did in his prior term. And he will have a very strong incentive to follow through on exerting political control over the Department of Justice because of the federal charges pending against them. So there’s real reason to worry there. I take those points very seriously.
What our experts were thinking—so the finding you’re referring to is our experts rated the state of American democracy higher than at any point that we’ve seen since 2020. And I think this, a point I would emphasize here is it suggests that our experts aren’t simply responding as individuals. They’re responding in their professional capacity. We’ve actually seen in general, they’ve rated the state of American democracy higher than the public throughout Trump’s presidency.
And we think that reflects our professional judgment that for all of its flaws, American democracy is performing much better than the kinds of autocracies and mixed regimes we see around the world, right? We’re still a more democratic country than many of these places.
Our interpretation of that finding is that the experts are more optimistic because of the legal accountability processes that have actually started to move forward against Trump. He attempted to overturn an election. Political scientists know what it means. They know what it looks like when there’s a so-called autogolpe, a self-coup, right? To try to hold power when you should be removed from power as an elected leader. The discipline I think takes that very seriously and saw what Trump did as an extraordinary threat to American democracy. We have polling saying that they rate that as an extraordinary threat to American democracy. They rate Trump returning to office as an extraordinary or very serious threat to American democracy. They overwhelmingly do.
So I think they are optimistic about those accountability processes. The caveat to that is they don’t seem to take seriously the prospect of a cycle of political, politicized prosecution and retribution, which many people are concerned about. And I’ll include myself in that.
And even though in our survey, they didn’t know this at the time, but we actually find a substantial number of Republicans explicitly endorsing the next Republican president prosecuting Biden and other Democrats. Nearly two thirds of Republicans endorsed prosecuting Biden and Democrats in retaliation. And whether that’ll happen again is unknown, but it suggests that maybe we shouldn’t be quite as sanguine about the overall ramifications of these indictments.
Well, for one thing, healthier than any time since 2020 is maybe a low bar to clear. But I do think it’s helpful to sort of realize that the Trump autogolpe ends up failing, and the ways in which it fails may actually be a reaffirmation of this sort of multi-layering of institutions. The fact that you can’t just claim power at the federal level, but there are state levels, the fact that some of these prosecutions are happening at the state level is an interesting reminder of robustness of systems.
Watch this for a transition. So given the multi-layered robustness of political systems, given the enormous centralization of the media systems that we’re dealing with and how brittle they can be, you and I are both former Twitter addicts. We’re both people who I suspect are now going through some form of withdrawal because that platform has changed radically under the leadership of a particular individual. Taking in mind your really helpful work here suggesting that media effects don’t have this sort of massive political effects that we need.
If you were in the position of fixing social media and you found out that simple things like turning off the algorithms and putting in chronological feed, you found that simply exposing people to more cross-ideological information, these don’t magically fix democracy. Are there problems in the space of social media that you would want to take on and that you have fixes to suggest?
Yeah, that’s a great question. One simple point that is not a generalizable fix is that deplatforming clearly works, but we should also be very uncomfortable with the power that it gives to giant corporations. So I think the decision to deplatform Trump immediately after January 6 was the correct one. It clearly had a very significant effect on his reach. We’ve seen from various toxic social media figures that deplatforming had a remarkable effect on them.
I don’t want to think of that as a solution to our problems though, because we we should be very uncomfortable in democracy with the idea of companies deciding who gets to have reach in the political square. I think that’s a tremendously dangerous power and one we should exercise with real caution.
And obviously, and actually we saw, we’ve seen various examples of mistakes that were made, for instance, during COVID with masks, with, you know, we saw with the Hunter Biden’s laptop, the list goes on. So that is the last resort, but I hope we never have to use that again.
Beyond that, I would say I’m not ready to recommend a series of solutions because I don’t think we know what the problems are. And I’ll just make a pitch here for the generalized problem that the combination of a lack of transparency by social media companies and a lack of researcher access to data means we actually have a very poor understanding of what people are seeing on social media, let alone its effects.
So what I would pitch is first we should start with transparency mechanisms like the Coons-Portman bill that’s pending in Congress or the various efforts to create more transparency in Europe. I think those will generate findings that may inform our understanding of what the problems are. I’ll also say, my co-author on those papers that we talked about at the beginning, Drew Dimmery, who was formerly at Meta and has since left and is back in the academic world, wrote a very interesting post on his Substack about the experiments that we conducted with Meta as a kind of template for informing future regulation. That if experiments like that were able to be designed and conducted in partnership with independent academics, that would help inform our understanding of the kinds of regulatory questions that might come to bear. Like for instance, would mandating a reverse chronological feed, improve this or that outcome that we might think is important. So I think there’s real potential to learn.
My general take, and this is informed by my work on studying untrustworthy website exposure in 2016, by my work on the Meta project and by my work on a recent paper, my co-authors and I conducted looking at exposure to potentially harmful videos on YouTube. My general view is—and I alluded to this before—is we’re focused too much on these generalized political persuasion effects. And I think we should focus much more on where the effects of misinformation are most acute and the harms are most direct and in evidence.
So we just lived through an attack on democracy that could have overturned a presidential election. It was powered by misinformation. The misinformation helped enable the elite actions that were at the core of the conspiracy to overturn the election. The evidence seems clear that misinformation contributed to vaccine hesitancy on COVID-19. Of course, we should be very cautious about how we try to address misinformation there because of the kinds of mistakes I described.
So I guess what I would say is I’m less convinced that there’s a particular set of tools, but more convinced we should focus: A) on these areas where the harms are most acute, like threats to democracy and threats to public health. And second, we should focus on the small minority of people, I would argue, who
are, have the potential to create the greatest harm in the world who are consuming lots of potentially harmful content.
Because defining again and again, Most people don’t see very much of the worst of the worst. But the people who do may be inspired to take real world actions and/or their responses may shape the behavior of elites and other people who take actions in the world.
So first of all, for the record, I want to point out that you brought up the Platform Accountability and Transparency Act without me prompting, which makes me very, very happy because I’ve been working on aspects of the legislation and promoting it and talking about this problem for years now, that the best researchers in the field, i.e. people like you, Talia Stroud, so rarely get the sort of access that you had during the 2020 Facebook study.
Frankly, as we head into 2024, we do not have a parallel study that looks like this. In many ways, we have many fewer tools than we had for either 2016 or 2020 because we’ve lost Twitter as a tool for study.
I worry that there are some legislators out there who are taking your message of looking at select groups and misinformation. And the select group that they’ve chosen is young people and are now responding by essentially trying to cut off media access in general. As we start seeing legislation trying to prevent people from using social media without parental permission, unfortunately, it looks like COSA has some bipartisan support and unfortunately seems to be a bill that I suspect is actually going to harm a lot of children, particularly LGBTQ children who are sort of turning to these tools.
But the last thing I wanted to ask you was around this challenge of focusing on people who really are seeking out mis- and disinformation. So people who are vaccine skeptics, people who might have white nationalist beliefs, who are looking, people who might have subscribed to aspects of QAnon.
I think many researchers see these groups as unreachable. Are you looking at these sorts of questions of deplatforming and sort of thinking of that as a possible response to this, or are there other interventions to help people who really are putting themselves and society at harm with things like vaccine resistance?
Yeah, that’s a huge question to ask with like two minutes to go. I’ve conducted a number of experiments of various kinds trying to reach people who are most convinced by misinformation or people who have relatively extreme views in some of these areas. There are no silver bullets, there are no messaging fixes. I think managing social media to think about the affordances they provide enable those folks to coordinate and concentrate is something that we could do a better job of.
So I’ll give you a very simple example, since we’re short on time, the former head of the civic team at Facebook posted about how they should be managing for the 99th percentile of exposure to potentially harmful content, because we keep having conversations where People like you and me say this kind of content is worrisomely prevalent on Facebook or some other platform. People at the platform say it’s 0.1% or 0.01% of what people see on the platform.
And that’s true. And the median user is seeing barely any of it at all. But if there is a subset of people who are seeing a lot of it and are organizing to take action around it, and it’s in in one of these areas where potential harms in the real world might be caused, then I’m much more concerned. And so, and that’s a hard one to regulate, but I think it’s a kind of focus shift that we could make both in civil society and academia in terms of how we research, but also in terms of how platforms are managed.
And that actually helps me understand a piece of yours that I routinely teach, which is some of the work that you did with ADL on looking at exposure to extreme content on YouTube, where you end up finding that there’s a small number of people who are responsible for consuming most of the content. I show your results on a slide that says racist got a racist. And in that sense, it’s a little bit dismissive. It’s sort of saying, you know, there are people who are gonna find that extreme information.
You’re offering this really helpful corrective, which is that that’s not a rounding error, right? We actually have to think about people who are in fact bathing in this and really getting exposed to it.
Listeners, he’s Brendan Nyhan. He really is one of the social scientists that I turn to the most often to look at ways of really rigorously answering these questions. And unfortunately, these answers are very rarely easy. The good news is that there are answers. We’re getting closer to them over time. Brendan, what a pleasure. Thank you so much.
My pleasure. It was fun. Thank you.