What happens when an entire profession can’t see what’s hiding in plain sight in its own data? That puzzle animated Stony Brook University sociologist Musa al-Gharbi’s keynote at the Heterodox Academy 2026 West Coast Regional Conference, held recently at UC Berkeley. Al-Gharbi walks through study after peer-reviewed study on polarization, symbolic racism, trust in science, and trust in the press, arguing that researchers and journalists have systematically misread the last decade of American politics by scrutinizing only the “red line” while leaving the much larger shift among highly educated knowledge-and-culture professionals unexamined. The deeper problem, he contends, is not bad-faith activism but a structural one: peer review, editing, and committee deliberation only correct for bias when the people doing the correcting actually differ from one another, and the academy and the press increasingly do not. His full speech is transcribed below.
What I want to talk about a little bit is the issue that really got me involved with Heterodox Academy in the first place. So one of the things that I became known for a little bit in my research is that during the 2016 election, I was kind of disturbed by the fact that pretty much all scholars and analysts and pundits were uniformly convinced basically that Trump would never win the Republican nomination. And then after he won the nomination, there was like two seconds where people were like, huh, I feel like we should learn something from this. But then the second that Hillary Clinton won the nomination, it was right back to, oh, it’s going to be a landslide, the Republicans are going to be destroyed for a generation, it’s a bloodbath, and so on. And then once again, we had one of the largest collective predictive failures in modern political history.
But rather than learning anything from that, what scholars often tried to do, the kind of vast preponderance of academic research that I encountered was basically scholars and journalists trying to pathologize the voters, to try to explain the electoral result by some kind of deficit or pathology that held among people who voted for Donald Trump. So almost nothing was learned from this. Rather than asking, well, what did we get wrong, what are we missing, what did we fail to understand, what’s wrong with some of our institutions, the kind of dominant question that researchers were asking, especially in my field and a lot of other adjacent fields, is basically, what’s wrong with normie Americans?
Okay, so before I get into the meat of the lecture, I want to start with a quick pop quiz. So here on the screen behind me, you see two lines. And you can see at this point in the series, they’re close together. And at this point in the time series, they’re further apart. If you had to explain to me which of these two lines shifted more to the left or the right, like which one is driving the growing gap between A and B, which one would you say?
Okay, great, great, great. So same thing again. So you have two lines that are further apart. They grow further apart over time. Which one’s responsible?
See, so the thing is, when you have the lines like this, in black and white with generic labels like A and B, no one has any problem at all understanding what’s going on in these graphs. But then this funny thing happens. When you start to color the lines red and blue, and you attach labels to them like Democrat and Republican, then all of a sudden, the thing that’s super obvious to anyone who just looks at the lines becomes almost impossible for a lot of people to see.
So that first chart that I showed you was from a political scientist, from an article called “What Happened to America’s Political Center of Gravity?” Now, what the chart visualizes is the Democrat and Republican Party platforms from 2000 to 2016. And so you can see that around the time Barack Obama was elected, the Democrat and Republican Party platforms were not actually very far apart. By 2016, they were very far apart. Well, what you can see when you look at the lines is, in the red line, there was kind of a mild polarization in 2012 before the party moderated under Donald Trump in 2016. And the reason that the lines are further apart was because starting in 2012, the Democratic Party shifted radically to the left, and they continued on that same trajectory through 2016. And so the lines are further apart than ever.
But of course, if you read this article called “What Happened to America’s Political Center of Gravity,” that is not the impression you’ll get. The impression you’ll get is that what happened to America’s political center of gravity is that Republicans went crazy.
One thing that’s critical to note is that this kind of wild shift that you see in the Democratic Party platform, in the words of Kamala Harris, “it didn’t fall out of a coconut tree.” Instead, what you see when you look at public opinion is that the Democratic Party was responding to changes in the Democratic base. So the Democratic base, starting in 2012, started shifting radically in a way that put them not just out of step with Republicans but also with independents. So the Democrats grew further apart from the median voter. So this is not a case where you have kind of regular Republicans and regular Democrats going radically apart from each other in kind of roughly equal ways. This is a case where you see a kind of radical shift on the blue line, a uniquely asymmetrical shift. And in fact, when you interrogate what’s going on with this blue line, it turns out that this isn’t just Democrats in general who shifted, but a very specific subset of Democrats who shifted even more extreme than that looks, to the point where they’re shifting the whole curve. But we’ll talk about that more soon.
So the second chart was from an article by another political scientist, Tom Wood, called “Racism Motivated Trump Voters More Than Authoritarianism.” Now before I get into the substance of the chart, I’ll just note that that’s a really interesting way of designing the study. So there’s this whole genre of studies that are structured this way where it’s like, basically, which negative trait best explains why someone voted for Donald Trump? Is it that they’re more racist or sexist or authoritarian? More on that in a moment.
So what the chart visualizes is it looks at Democrats’ and Republicans’ embrace of symbolically racist attitudes from 1988 to 2016. Symbolic racism, quick nutshell, is the idea, it used to be the case in polls and surveys, social scientists just asked Americans, do you like black people? Do you want black people to live in your neighborhood? Would you feel comfortable with someone in your family marrying a black person? And they would just say no. And then increasingly, people stopped saying no. And so the question was, is this a progress story, Americans becoming less racist? Or is it the case that people have just become less comfortable expressing racist attitudes, but the baseline levels of racism remain roughly the same, they’ve just become better at concealing their racist attitudes?
Most social scientists just largely assume that the latter is true, that in America there has been very little progress in terms of actual racism, that the main thing that’s happening is that people are just less likely to express their racism in a direct way. And so scales like this one are designed to try to measure this kind of latent hidden racism. Now there’s a bunch of problems with these scales, but let’s just set that to the side. Let’s just pretend like this chart measures racism.
Okay, so what you can see in this chart is that in 2016, when Trump was running, the gap between Democrats and Republicans on these racial attitude measures was bigger than any previous year. And on this basis, Wood argues, race must have played a bigger role in the 2016 election than in other cycles. So far, very reasonable inference. The problem is the next assumption he made. He said, so if race played a bigger role in this cycle than in previous elections, it must be that Trump voters are driven primarily by racism.
The problem with that narrative is that his own chart — he created this chart, I did not create this chart, this is his chart — what you can see is that on each and every one of these measures of symbolic racism, the red line is going down from 2012 to 2016. Which is to say his own chart shows that Romney voters were more racist than Trump voters. Trump voters are less racist than previous cohorts of Republicans. His own chart shows that. The reason why the gap is bigger is because starting between 2012 and 2016, there was this massive shift in the blue line. Democrats shifted radically. The Republican trend was towards convergence. So again, moderation — the reason the gap grew was because of these radical shifts in the blue line.
Democrats became less likely to endorse these symbolically racist attitudes in 2016 than in any previous year on record, less than when there was a black candidate from their own party, America’s first black president, Barack Obama. In fact, one thing I’ll just note parenthetically, for the years that Obama was on the ballot, Democrat endorsement of symbolic racism actually went up by a lot of those measures, which is really interesting, completely unanalyzed in the literature though, because we don’t actually analyze the blue line, we’re focused on the red line.
So one of the things that’s really interesting about the structure of this, like “which negative trait,” is that we don’t really analyze why people vote for Democrats at all, actually, much at all. It’s not even an interesting research question for most social scientists. Why would someone vote? It just seems to us obvious why someone would vote for Barack Obama or Hillary Clinton, they’re the better candidate. But if someone did design a study asking, well, why would someone vote for Hillary Clinton, and the structure of the design was competing between a bunch of negatively valenced traits, like, why would someone vote for Hillary Clinton? Is it that they’re more communist, that they’re atheist, that they hate America? If you designed a study like that, by the way, you would absolutely 100% find, for instance, that if you reverse coded patriotism, it would be associated with votes for Democrats. You could absolutely design a study like this and find these kinds of things. But if anyone tried to do that, which negative trait explains why someone votes for Democrats, and tried to get that published in a peer-reviewed journal, it would be universally rejected, and for good reason. It’s a prejudicial study design. When you do things like this studying Trump, it’s just taken for granted as normal. There’s literally a whole genre, as I’ve shown in some of my other works.
Now, in fact, one of the exceptions, Julie Wronski and colleagues, looked at how authoritarianism, they studied authoritarian impulses looking at Democratic primary voters. So did supporters of Hillary Clinton or Bernie Sanders, were they more likely to be authoritarian? She found really interesting results. But this is an exception. As lots of other research shows, we just don’t really ask about Democrats very much. We don’t design a whole bunch of studies asking why people vote for Democrats. This is not a research question we ask.
More disturbing, even when we have. Like for instance, there are a whole bunch of studies that show a strong empirical association that goes back, that you see in almost all data sets, as far back as the empirical record goes, in every country that you study, that subjective well-being, feeling happy, fulfilled, content in life, is associated with conservative political leanings and religiosity. And there’s this whole genre of studies whose literal point is to pathologize wellness basically because conservatives come up higher on subjective well-being. We design whole studies that basically argue, well sure, conservatives are happier, but that’s just because they suck. Like if they were decent people who actually knew stuff about the world and cared about other people the way we care about other people, then they’d be miserable too. The idea that you would design a study pathologizing wellness is not a thing you would see if the data went the other direction. If it turned out that liberals rated higher on subjective well-being than conservatives, you can 100% bet that there would not be a vast literature trying to attribute this finding to some kind of pathology. Instead, the narrative would be, it’s just further proof that the left is correct. Not only is it correct, it’s good for you. And you see this in measure after measure.
One thing that you see a lot, as Christine Reyna demonstrated in this study, is that if you look at things like social dominance orientation, or almost any political scale that you want to imagine, what often ends up happening is that the gap between Democrats and Republicans is actually pretty narrow. The objective gap between them is actually pretty narrow. So you could create a 10-point scale on social dominance, and you can see that Democrats come in at 3.42, while Republicans come in at 3.75. Since both of them come in well below the midpoint of the scale, probably the most appropriate way of interpreting that finding is to say that both Democrats and Republicans are broadly egalitarian. And you can even argue that Democrats are marginally more egalitarian than Republicans. Is that the way it’s described in the literature? No. The way it’s usually described in the literature is that Republicans are motivated by dominance orientation, that they’re dominance-oriented, and that this explains why they vote for Donald Trump. And that’s wild, because to take the average 3.75 on a 10-point dominance scale and describe them as dominance-oriented because they’re just very marginally different from Democrats is a wild thing to do, but it’s also a very common thing to do, as Christine Reyna and others have shown.
There are other literatures. There’s a vast literature in sociology and political science on racial dog whistles. Racial dog whistles are formally race-neutral statements that are supposed to, in a wink-and-a-nod way, suggest some kind of affiliation with white supremacy or hatred of non-whites or immigrants. So it’s stuff that’s not explicitly about race, but kind of winks and nods towards your racist leanings. Now the problem with this field is twofold. First, the way that a lot of people have come up with examples of racial dog whistles is it’s a bunch of white liberal scholars who sit around thinking, if I was a racist, what kind of message would appeal to me? And then they come up with these and then they test them. And if Republicans say that they like these messages, then they go, aha, it’s proof that these statements must be racist. What they don’t do, implicit in the racial dog whistle narrative, is that these are narratives that appeal uniquely to whites, because again, they’re signaling, winking and nodding towards white supremacy. But they don’t actually usually test non-whites’ receptivity. They just only test white people.
One exception to this rule is one of the pioneers of this framework, Ian Haney López, who, to his credit, decided to actually investigate non-whites’ receptivity to racial dog whistles. So he took canonical examples of racial dog whistles from Donald Trump, but didn’t say they were from Trump, and presented them to black, white, and Hispanic voters on law and order, immigration, things like this. What he found was that these messages actually resonated the most with Hispanics, and then African Americans, and they resonated the least with whites. Now this is a major problem for a framework that’s predicated on the idea that these are messages that resonate uniquely with whites, and they actually resonate the least with whites. That’s a big issue.
But scholars have largely insulated themselves from having to deal with that problem by largely ignoring the views of non-white voters, including and especially in ostensibly anti-racist literature. They just ignore the views of non-whites because they’re inconvenient. I suspect one of the major reasons why they just focus on whites is because of an intuitive understanding, not tested. Because again, if they do test it, you have to deal with this kind of problem. So they just don’t even test non-white public opinion much. And there’s all sorts of research looking at how this is a really vast problem in public attitudes about race and racism, that there’s just all these big claims that are made about unique views of whites and racism and so on, and they just ignore non-white voters because they have inconvenient views. And when you do test them, you get inconvenient views for the narrative that they’re trying to spin.
In fact, one of the things I’ve shown in my own work as it relates to Trump is that a lot of the narrative about why Trump won is that he won because of racism. But actually what you see, as I’ve shown over and over again for like a decade now, whose vote shifted the most between 2012 and 2016?
White turnout was stagnant between 2012 and 2016. Trump got a smaller share of the white vote than Mitt Romney. The reason he won in 2016 was because he got a larger vote share with black people, Hispanic people, Asian people. And in every single midterm and general election since 2016, Democrats have continued to see attrition with non-white voters.
But again, this has been very tough. And in 2020, the reason Joe Biden won was because of shifts among whites towards the Democratic Party, specifically white men. Joe Biden won in 2020 because of white men. Now, this is the thing that, again, within the dominant frameworks of a lot of these disciplines, it’s just really hard for them to even see, let alone process and talk about these findings, even though they’re kind of blindingly obvious.
Last example I’ll give, you see the same kind of thing in gender. So if you’re doing political analysis for why an election went the way it did, here’s a pro tip. Look at women. Women are a larger share of the baseline adult population. They’re also registered to vote at higher levels than men, and this has been the case for decades now. And among registered voters, they turn out to vote at higher levels than men, which is what this chart visualizes. So the total adult population is already skewed towards women. The electorate is even more skewed towards women. What this means is, if you want to understand any electoral outcome, women’s votes matter more than men’s.
But when you look at what you would think, for instance, in this last election, when you had a female candidate who stood to be America’s first female president on the ballot, you might think especially in this kind of a race, like in 2016 or 2024, it would be important to look at how women exercise their agency. That is not, for the most part, what people who study politics decided to do. Instead, the narratives are all about men and misogyny and so on, even though, again, Trump didn’t perform exceptionally well with men, but Kamala did perform really poorly with women. The only Democrat who got a worse vote share than Kamala Harris with women, you’d have to go back decades to go to John Kerry in 2004. Incidentally, the first person I ever voted for, ugh. A really ill-fated campaign.
But the reason Kamala lost was because she did poorly with women. Her vote share with men was actually comparable to a lot of other Democratic nominees. She didn’t do poorly with men. The reason she lost was because she did do poorly with women. But even a lot of ostensibly feminist scholarship declines to look at how women exercise their agency. And this is especially true if they’re trying to explain something they think of as bad. So most scholars think that Trump’s election was bad. We think that women are good, so we just don’t design studies that look at how good people created bad things, how women could have been responsible for Trump’s election.
We can see all of these trends simultaneously in this chart. So the study is restricted explicitly to whites. So when he’s looking at differences between Democrats and Republicans and this endorsement of symbolic racism, it’s just white voters he’s analyzing. He just conveniently ignores non-white voters. Even though both partisans, both Democrats and Republicans, it’s a five-point scale of symbolic racism, you would be hard pressed to know that because he truncated the Y-axis, which makes the lines look bigger than they are. In reality, for the whole time series, both Democrats and Republicans are both above the midpoint of this five-point scale. Which is to say, a way of understanding this data would be that both Democrats and Republicans are motivated by racism. If you wanted to have this kind of unfavorable, uncharitable analysis of the trend lines here, since both of them are above the midpoint of the racism scale, you could argue, hey, Democrats and Republicans are both motivated by racism, but Democrats in 2016 were moderately less motivated by racism and certainly less than they have been in the past, or something like that.
In any event, so you have this truncated Y-axis which creates the illusion of more difference between the parties than there actually is. And the data is analyzed as if the red line and the blue line people just have radically different motives. The red line people are motivated by racism, the blue line people are presumably, well, we don’t actually get into their motives, but the assumption is they’re motivated by good things, they know about the issues, they care about America or whatever. And this is just kind of a broad pattern you see.
One thing I’ll just note, if you continue this time series through 2024, it gets pretty interesting. So Republicans continue, you continue to see these declines in racist attitudes among Republicans, which is to say the whole time Trump’s been in office, in fact, the whole time Trump’s been part of public life, Republicans have been growing less and less racist. It’s not the news that you would get. But since 2020, the gaps between the Democrats and Republicans have begun to shrink. So the distance between them has grown smaller. This is for two reasons. It’s both because the Republicans continue to grow less racist, but the blue line is also converging again, back to this moderating. So if you were going to analyze this chart in the uncharitable way that the initial chart was analyzed, how you would analyze this is that Democrats have become more racist in recent years. So Democrats grew more racist between 2020 and 2024 with Kamala on the ballot. But that is absolutely not the way that they would interpret that coefficient.
You can see this kind of thing in chart after chart after chart about polarization. So Leonardo Soares and Baekkwan Park had this great study in the Journal of Politics where they looked at issues where Democrats and Republicans have grown further apart over the years. What they find is that in most issues where Democrats and Republicans have grown further apart, Democrats have been the one driving the change. They’ve shifted more. Pew, similarly, looked at all the issues where Democrats and Republicans have grown further apart from 2003 to 2023 and who shifted more. What you can see is that on the vast majority of issues upon which Democrats and Republicans have grown further apart, the reason they’ve grown further apart is because Democrats shifted significantly, with a small number of exceptions. But again, the vast majority of discourse that you’ll see in journalism or academia about growing polarization is focused on Republicans and starts from the premise that Republicans have lost their ever-loving minds.
More to the point of what we’re doing here in higher ed, if you look at public trust in the scientific community. So this is a famous study in one of the top sociology journals by Gordon Gauchat. What he’s trying to answer in this study is, if you look at the beginning of this time series, this line, conservatives, this kind of square line, conservatives are technically the most trusting of the scientific community at the beginning of the time series. It’s actually a statistically insignificant difference between the progressives and conservatives, but whatever. Let’s just say they’re the most supportive, and by the end of the time series, they’re the least supportive. And so he’s trying to understand why that is. Fine thing to study. But one thing that’s interesting is that if you look at the whole time series, at the beginning, conservatives and liberals are kind of pretty close together. Conservatives continue on this same trajectory of consistent declines that they were on over the course of the whole time series. But starting in the mid-90s, Democrats go from trending with conservatives to just abruptly shifting the other way and becoming more and more trusting of the scientific community, to the point where at the end of the time series, they’re just floating off in space.
It’s not the case that conservatives are here and the rest of the independents are somewhere in the middle and progressives are over here. The objectively weird line when you look at this chart is actually the liberal line. It’s just floating off in space. Conservatives and moderates are actually very close together and trending in much the same direction since the mid-90s. But you have this really unusual divergence, both a divergence from their previous trend and a divergence relative to everyone else in America, not just conservatives. But we don’t design studies going, “Why are liberals so credulous about the scientific community?” In fact, when we design studies, we just grant it that Americans should trust the scientific community, that if anything, the levels of trust should be somewhere way above the chart, and that anything below 100% acceptance is some kind of pathology that needs to be explained. And so we don’t even really design studies asking why are liberals so extremely credulous about the scientific community? But if we want to understand polarization around science, who it is that’s diverging the most from the rest of America, again, not just from the out party, but even from moderates or middle people, it’s actually this line. This is the weird line on the chart.
And you see the same thing, liberals, moderates, conservatives, if you switch it to analyzing Democrats, Republicans, independents, you see the exact same trend. In fact, this time series continues it. That one ends in 2010. This continues it to 2020. You can see that Democrats and Republicans are even further apart in their trust in the scientific community. But again, who’s driving the polarization? What you can see is that the blue line started polarizing first and way more, such that the gap between the Democrats and Republicans is certainly bigger than it’s ever been before, but it’s being driven largely by this shift in the blue line. And to the point where the Democrats are very far apart from independents, much further from independents than Republicans are. And you see the same trend, Republicans consistent line, Democrats starting in the mid-90s, this abrupt positive shift that then goes like into the stratosphere.
If you look at what is this chart trying to explain, what is the article that this is from trying to explain: why being anti-science is the new part of many rural Americans, read Republicans’, identity. So you have this kind of extreme hockey stick on the blue line, and it’s just unanalyzed. We just don’t go, well, why are Democrats so extremely trusting of the scientific community? It’s just not a thing that we really analyze that much. And the thing about that that’s so weird to me is that mistrust of experts is actually kind of easy to explain. As my colleague Gil Eyal in his book The Crisis of Expertise, it’s actually not a really tough puzzle about why people might be skeptical of experts. Experts make claims that you can’t verify. They’re wrong about stuff a lot, and when they’re wrong, they rarely pay the cost. Other people pay the cost for their errors. Experts are often very sociologically distant from the people that they’re making demands of. So they’re telling you to close your schools and keep your kids home, but they’re people who are not like you. They don’t live in your communities. They don’t share your demographic characteristics. They don’t share your class background. They don’t share your worldviews and life experiences. So there’s these distant people that are making demands of you. They often behave in ways that other people view as untrustworthy or weird, and they often have a different understanding of what your best interests are than you do. So when you have that kind of a situation, it’s really not a puzzle to figure out, well, why would people be skeptical of experts? It’s actually not that big of a puzzle.
What’s really interesting is that despite all of this, people often do follow expert advice. Most people vaccinate their children, for instance. And this is actually a more interesting sociological puzzle, that’s why despite all of this, people actually do defer largely to expert advice. But again, we don’t actually study that side of the equation as much, because a lot of us start from the premise that people should trust experts, and then try to explain why they don’t, usually by appeal to some kind of deficit or pathology. They’re brainwashed, they’re ignorant, and so on.
One thing that’s critical to note about those charts, by the way, is that what they’re measuring, both of those, we’re looking at trust in the scientific community, which is very different from public perceptions about science per se. In terms of science per se, you actually see almost no change in public attitudes about science per se from decades ago to the present. So this chart visualizes whether or not Americans think that the benefits of science outweigh the risks and costs. You can see a pretty static line from 1979 through 2018, basically no change in the green line. It’s a pretty overwhelming majority that believes that the benefits outweigh the costs. If you look at trust in American institutions, trust in science per se, basically a flat line, even as a lot of other institutions have seen a decline. So where you see a decline in these previous charts isn’t people who don’t trust science, but people who don’t trust the scientific community, which is the people speaking on behalf of what the science says, the experts. And that’s an important distinction to draw. A lot of us don’t attend to that distinction very much. And so we look at charts that show declining trust in scientists and interpret that as declining trust in science, which is a convenient conflation for us to make, because then it makes it easier for us to portray people who are skeptical of things we’re claiming as being rooted in some kind of irrationality, ignorance, and other pathologies. But that’s actually not what you see in public opinion trends. What you don’t see is growing distrust about science per se. It’s trust in scientists.
And part of the reason why trust in scientists might have declined is because scientists were engaging in a lot of behaviors that people found strange. So one of the things I’ve shown in my own research, analyzing tens of millions of academic research papers, starting after 2010, there was a rapid change in how scientists went about their business, like the themes that they wanted to study, how they talked about social issues. Institutions that host scientists, universities are the main place where scientists are hosted today, were riven by all sorts of wild conflicts. So if you look at cancel culture incidents after 2020, you see this kind of rapid spike and then eventually a peak. And as I’ve argued in other places, we seem to have passed kind of a peak woke, the kind of peak period of this contestation. A lot of scientific organizations explicitly oriented themselves towards politics and morality. Quick example: shortly after Trump was elected, there was a March for Science, and the literal structure of it was that a bunch of scientists, often wearing white coats and stuff, basically declaring that we have science over here, Trump over here, and you have to choose which one do you support. And so if you’re a conservative or a Republican, the implicit choice given to you is, you want to basically abandon all of your moral and political commitments in order to embrace science, or you’re going to be more skeptical of scientists and keep your moral and political commitments. If that’s the choice that people are faced with, how that resolves itself should be pretty clear.
In fact, if it’s not clear, a political scientist, Matt Motta, did a great study looking at the effects of the March for Science. Given that people broadly trust science, and before a lot of these kinds of attitudes and behaviors in universities really took off, people also trusted scientists broadly. What he was curious about is, since scientists and science are so popular, and Trump, on the other hand, was a polarizing and unpopular figure, maybe the March for Science would make people like Trump less, like scientists more. What was the effect? What he found is that the main effect of the March for Science is that it caused people to trust scientists less. And this was not without consequence, because scientists engaged in this highly polarizing set of activities right before the onset of a major global pandemic, where trust in science was actually important and highly consequential. Another way of saying this is that there might be people who are dead today who would have been alive if scientists hadn’t engaged in this kind of highly polarizing activity. If we think the work that we do is important, if we think the work that we do matters, then how we conduct ourselves matters.
The press, last thing I’ll say about this, you see a similar kind of trend. So I’m a journalism professor, so I got to talk about the press. But you see the same kind of thing you see for trust in science. If you look at the red line, a very consistent kind of path, nothing really changes in the overall trajectory. Independents, which broadly trend with Republicans for most of the time series. Starting in the mid-90s, Democrats are trending with everyone else, and then starting in the mid-90s, they shift in the other direction in a way that puts them increasingly out of step with everyone else. And then again, after 2016, such that the blue line is just floating off in space, much further from independents than Republicans are. But again, we don’t ask, well, why are Democrats so extremely credulous about mainstream media? Why is it that they’re so unusually, wildly trusting of the press? Because again, we just start from the assumption that actually it should be 100% of the people who trust the press, and any divergence from what we think people should be doing, we explain in terms of deficits and pathologies. And we just focus on the red line and the declines and appeal to things like misinformation, the Koch brothers, Trump and whatever. But if we actually want to understand polarization in the media, it’s actually important to attend to both sides of the line, especially the line that’s shifting more.
And again, if we want to understand why it is that people have started to gain suspicion about the media, this is actually not that hard to explain. So as I’ve shown in my research, analyzing tens of millions of articles in the New York Times and looking at TV coverage of different presidents, for instance, what you can see is that the way that the media covered Trump even before he was president was radically unlike any other political figure in modern history. They talked about him way more than America’s first black president, than any other president who’s ever been considered. Almost all of this coverage was extremely negative. And even after Trump was out of office, into 2022 when Joe Biden was president, they still talked about Trump more than the sitting president. And then TV news, same thing. For most presidents, what you see is when they run for president, they get a spike in coverage, and then when they’re the sitting president, it goes down, but not as much as before they were the candidate. For Trump, not only was his objective level of coverage much higher than any other president on record, but it just stayed high. We basically have four straight years of campaign-level intensity television coverage of Trump to a degree that I think would surprise a lot of people. So in the New York Times, for instance, in that year, 2018, Trump was the number three most used word in the whole New York Times, if you exclude words like “but,” if you just look at substantive terms.
On average, every single — I had a corpus that included everything published by the New York Times: sports, foreign affairs, weather, arts, culture, anything that the New York Times published was in this corpus. And on average, anything published by the New York Times in 2018 directly mentioned Trump about three times, and then indirectly, Commander in Chief, White House, POTUS, another couple of times. Trump became the lens through which the New York Times interpreted reality. Almost any story that they talked about was talked about through the lens of Trump. Now this kind of behavior would again maybe cause a lot of people to go, huh, seems like something is weird with the New York Times.
And it’s not just with respect to Trump. So my colleague and I analyzed 27 million news articles from 47 media outlets over the last 50 years. And what we found is that starting after 2010, there was this rapid spike in discussion about prejudice and discrimination, all forms of prejudice and discrimination: racism, sexism, homophobia, transphobia, Islamophobia, antisemitism, just hockey sticks all at once in a way that didn’t seem to correspond with anything that was happening in the world with respect to those issues. It predated Black Lives Matter, it predated Me Too, and so on, and in fact might help explain why those movements were able to take off. All to say, the media was behaving in a genuinely unusual way, and that’s maybe reflected in public interpretations of journalists. And you see this kind of pattern, as I show in my book, this kind of pattern that you see in academia and media of this kind of rapid shift in how they operate. You see this in most other knowledge and culture industries: finance, human resources, and so on. You see similar kinds of trends.
In fact, I looked at identities, voting behavior, protests, outputs by companies, a whole bunch of other metrics. And you see over and over and over again this kind of hockey stick that’s driven by knowledge and culture professionals. That shift that we saw on the blue line, it wasn’t like most, like black Democrats, Democrats without degrees, they haven’t shifted much at all. The reason the hockey stick looks like that is because highly educated, relatively affluent urban and suburban white people, especially people tied to the knowledge and culture industries, just changed really, really, really rapidly. If you want to know why they did this, you should read my book.
But one thing that I’ll just note about the way that we analyze these trends is that a big problem for people figuring out what’s been going on is that the people who were undergoing these radical shifts were the same people who are producing almost all of the mainstream narratives about those shifts. As an example, for an essay I did for Heterodox Academy, I showed that if you look at the professoriate, professors are highly unrepresentative of the public at large along most dimensions, in terms of their class backgrounds, in terms of their religious leanings, in terms of their political leanings, in terms of their ethnic composition, in terms of their sexuality. They’re just wildly unrepresentative of the American public. This is a problem for a couple of reasons, but you have this kind of narrow and idiosyncratic slice of society that’s undergoing this rapid change and that notices a growing gap between them and other people. But how they interpret that gap is colored by who’s doing the narrative.
So you see the same thing with journalism. Journalists, like academics, are concentrated in very particular geographic regions. In terms of ethnicity, you see a lot more diversification in journalists, which is to say that now 18% of journalists are non-white. In terms of partisan affiliation, journalists went from being broadly representative of America as a whole to being 10 to 1 Democrat. And in fact, even a lot of people who self-identify as independent, if you actually analyze their voting behavior and their political lean, almost all of those people are also Democrat-voting liberals. In terms of class composition, in the 70s, most journalists had college degrees, but they had college degrees from a lot of different institutions and from a lot of different majors, and people entered journalism from a lot of different walks of life. Today, almost all journalists have college degrees, almost exclusively in journalism and communication. And as more and more people have had college degrees, where you get your degree from matters more and more in journalism, such that the New York Times has a higher concentration of Ivy League graduates than the US Senate and the Forbes 500 list. And this has important implications for how we tell stories. If we’re supposed to be holding the powerful to account, and the powerful people we’re supposed to be holding to account are our buddies from Columbia and our parents and our wives, we just cover the story a lot differently.
And so journalists and academics and other knowledge and culture professionals, we correctly perceive that the gap between us and huge swaths of the rest of America is growing. We mistakenly assume that the reason the gap is growing is because those people must be getting more radical, even to the point where we can produce work that clearly illustrates that we’re the ones who shifted, but we can just not see, we can literally not see our own data. So again, these very sophisticated, good social scientists produce charts that clearly, unmistakably show that the blue line shifted more and is driving polarization, and they couldn’t see the trend in their own data, and instead put forward articles arguing that the red line people are going crazy.
But it’s not just these individual scholars who missed it. That to me is not what’s most interesting. Any of us can have a brain fart. What’s interesting to me is that the peer reviewers couldn’t see obvious errors, and the editors couldn’t see this obvious problem. And the often data-sophisticated journalists who looked at these charts and circulated them and wrote articles praising them also couldn’t see the obvious problem. And the other scholars citing these studies couldn’t see the obvious problem. It’s right in front of people’s faces. And once you point it out, you can’t unsee it. I can’t show you that chart again and you not see that the blue line is driving the polarization. But until someone points it out, it’s really hard.
And this is not a problem, for the most part, of mustache-twirling villains who are willfully engaging in activist social science and misinformation. The problem is that institutions like higher ed and journalism are designed on the idea that we all have partial and situated knowledge, that we all have biases, we all have blind spots, we all are prone to error, we all have limits to our comprehension and our empathy and so on. But if you create institutions and processes that pull together people with different interests, different backgrounds, different values, then we can collectively correct each other’s errors and produce something that looks like reliable, objective, comprehensive truth. But these systems only work, systems like peer review, committee decision-making, they only work as intended when you actually do have people with diverse views and values and life experiences taking part in the scientific enterprise. When institutions are dominated by a narrow and idiosyncratic slice of society that largely shares a lot of interests, values, life experiences, risk exposures, and so on, then these same processes that are supposed to help correct our biases can reinforce them instead. It can make it harder for new ideas to break through, harder for dissent to occur, harder for innovation to occur. You can get reinforced groupthink. You can have misinformation cascades.
So this is not a problem of individual scholars who have some kind of deficit or pathology. I’m not just reversing the sign and pathologizing scholars. This is a structural issue and a collective action issue. So my next book, which is slated for publication with Princeton University Press, will explore the causes and the consequences of this growing social distance between us, professionals in the knowledge and culture industries, and huge swaths of the rest of society. What are the causes of this growing distance, what are the consequences of this growing distance? How does this distance interfere with our ability to understand other people and their attitudes and behaviors, and often lead people to misunderstand us and our institutions and communities as well?
Thank you for your time.
