Over the last four decades, America’s colleges and universities have become some of the most ideologically lopsided institutions in public life.

Consider the growing ideological homogeneity of university faculty (Figure 1). In 1995, the liberal-conservative ratio among faculty members was roughly 2 to 1. By 2010, the ratio was 5 to 1. The most recent data suggests even less political diversity within the professoriate. According to a 2018 survey of liberal arts faculty, for example, the Democrat to Republican party registration ratio was nearly 13 to 1. Data from 2019, suggested that more than 60% of faculty were left-leaning while only 10% were right-leaning. In 11 surveys conducted between July 2021 and August 2022, those calling themselves “far left” or “very liberal” (11.4%) outnumbered the entire right-of-center faculty combined (10.9%).

Figure 1 – Faculty Political Ideology, 1969-2022

Notably, one of those 11 surveys found that more than 40% said labels like “radical,” “activist,” “Marxist,” or “socialist” described them at least moderately well (Table 1).

Table 1 - Faculty Identifying as “Radical,” “Activist,” “Marxist” or “Socialist”

Campus-level snapshots at high-profile universities paint the same basic picture of the professoriate’s increasingly homogenous political dispositions. Surveys of Harvard faculty since 2022 have placed the share identifying as liberal somewhere between 63% and 82% (Figure 2). A survey of Duke’s faculty found only 14% of faculty identify as conservative. Most recently, the Buckley Institute’s 2025 report on Yale concluded that 83% of faculty are Democrats and less than 3% are Republicans.

Figure 2 - Faculty Political Ideology at Harvard University

Nor is the political homogeneity confined to faculty. More than 70% of college administrators now identify as liberal (Figure 3).

Figure 3 - Political Identification among US College Administrators

Further, approximately 60% of students have consistently identified as “liberal” or “Democratic Socialist” in the last four iterations of Foundation for Individual Rights and Expression’s (FIRE) annual College Free Speech Rankings (CFSR) survey. As Figure 4 shows, while there are very few campuses where conservatives are a majority (14 of the 257 schools FIRE surveyed in 2025), there are a large number of campuses (118 of the 257) where liberals outnumber conservatives by at least 3 to 1.

Figure 4 – Student Political Identification Across Campuses

Who Cares about Campus Monocultures?

Defenders of the academic status quo are either unbothered by this data or see it as a good thing. This dismissive attitude, however, represents a growing danger to higher education’s ability to carry out its primary responsibility to pursue and transmit knowledge. A high-quality education requires a campus environment where all students feel they can speak freely, argue with faculty and peers honestly, and dissent from prevailing orthodoxies without fear. When students feel uncomfortable expressing themselves or routinely self-censor, the university is failing at its core mission. The question that everyone involved in higher education needs to take seriously, then, is whether a politically homogeneous faculty and student body expands or contracts the number of students who are keeping quiet?

Recent evidence from a small number of campuses suggests that faculty ideological homogeneity is leading more and more students to self-silence. In a political monoculture, the classroom stops functioning as an arena for genuine “truth-seeking” and, instead, starts functioning as a conformity test that students cannot afford to fail. For example, in a study published in August, 88 percent of students at Northwestern University and the University of Michigan said they had “pretended to hold more progressive views” so that they could “succeed socially or academically.” Similarly, Harvard president Alan Garber said the university “went wrong” by tolerating faculty activism in the classroom and argued that professors’ political views have chilled student debate on campus.

While suggestive, anecdotal evidence from this small handful of elite schools can’t shed light on how ideological homogeneity might affect students more broadly. Fortunately, data collected for FIRE’s 2023 CFSR survey can. This survey, which sampled more than 43,000 students at 203 colleges and universities, asked respondents to place “the political views of the average faculty member on campus” and “the political views of the average student” on a 7-point scale ranging from very liberal to very conservative. More importantly, FIRE’s 2023 CFSR survey also included a series of questions that allow us to assess student experience in a way that maps directly onto educational quality:

• Self-censorship: “On your campus, how often have you felt that you could not express your opinion on a subject because of how students, a professor, or the administration would respond? (Never, Rarely, Occasionally, Fairly often, Very often)”

• Comfort speaking: How comfortable would you feel doing the following on your campus?: Publicly disagreeing with a professor about a controversial political topic, Expressing an unpopular political opinion to your fellow students on a social media account, Expressing your views on a controversial political topic during an in-class discussion, Expressing disagreement with one of your professors about a controversial political topic in a written assignment, Expressing your views on a controversial political topic to other students during a discussion in a common campus space (Very uncomfortable, Somewhat uncomfortable, Somewhat comfortable, Very comfortable).

• Difficulty having open and honest conversations: Some students say it can be difficult to have conversations about certain issues on campus. Which of the following issues, if any, would you say are difficult to have an open and honest conversation about on your campus? (select all that apply):

  • Abortion

  • Affirmative action

  • China

  • Climate change

  • Economic inequality

  • Free speech

  • Gender inequality

  • Gun control

  • Immigration

  • Israel-Palestine conflict

  • Police misconduct

  • Racial inequality

  • Sexual assault

  • Transgender rights

  • None of the above

Student Perceptions of Faculty Political Ideology

A useful starting point for assessing the educational effects of campus-level ideological homogeneity is to explore student perceptions of where their professors and their peers are politically. Unsurprisingly, the vast majority of students in the 2023 CFSR said that their campus’s faculty members are on the political left. While 11% of students placed faculty on the right-side of the political spectrum (i.e. slightly, somewhat or very conservative) and 20% of students placed faculty in the middle of the scale, 69% of students placed faculty on the left-side of the political spectrum (i.e. slightly, somewhat or very liberal). In fact, there were only 4 schools out of the 203 surveyed in the 2023 CFSR (University of Notre Dame, Claremont McKenna College, Texas Tech University and Texas A&M University) where students, on average, placed faculty members on the “conservative” side of the ideological scale (i.e. average above 4).

Figure 5 – Student Perceptions of Faculty Political Ideology and Actual Faculty Political Ideology

How accurate are these perceptions? While we don’t have data on self-reported political ideology for faculty members across every campus included in FIRE’s 2023 CFSR survey, we do have it for some. Specifically, FIRE conducted a large-scale survey of faculty members in 2024 that received a sufficient number of responses from professors working at 55 of the 203 campuses covered by the 2023 CFSR survey. As Figure 5 shows, there is a strong correlation (r=.57) between where students placed the average faculty member on a 7-point ideological continuum and the percentage of faculty members who identified as conservative in FIRE’s 2024 faculty survey. In other words, when students say their faculty feel politically one-sided, they’re usually describing something real.

Student Perception of Student Body Political Ideology

Most students also believe that their peers are on the political left. There were only 18 schools in the 2023 CFSR survey where students placed their average classmate to the right of center (i.e. an average above 4).

Figure 6 – Student Perceptions of Student Body Political Ideology and Actual Student Body Political Ideology

Much like student perceptions of faculty political ideology, student perceptions of their peer groups’ politics are also accurate. As Figure 6 shows, where students placed the typical student at their school along the left-right ideological continuum and the actual percentage identifying as conservative in FIRE’s 2023 CFSR survey was highly correlated (r=.79). Once again, therefore, students are well-aware of where their campus sits politically.

Student Perceptions of Faculty and Student Body Political Ideology

For most moderate and conservative students, the campus environment is not just liberal faculty or liberal peers. It’s both at once. Combining responses to the measures of faculty and student body ideology, Figure 7 shows that the vast majority of colleges and universities are clustered in the “liberal faculty AND liberal student body” quadrant, placing moderate and conservative students as consistent minorities across institutions of higher education. Specifically, 184 of the 203 campuses FIRE surveyed for the 2023 CFSR were “left-dominated” (i.e. perceived to have both faculty and students to the left of center). In stark contrast, there were only three schools that were perceived to be “right-dominated” (i.e. where both faculty and students were, on average, placed to the right of center). Overall, then, moderate and conservative students are typically “out of step” politically with both their professors and their peers.

Figure 7 – Student Perceptions of Faculty and Student Political Ideology Across Campuses

The “Outsider Penalty” in Higher Education

The data above show that many moderate and conservative students believe they are out of step with the two most powerful social reference groups on a college campus at the same time: professors and peers. But the real question, of course, is whether these perceptions color students’ educational experience? Figures 6 and 7 show the predicted scores on the self-censorship question and an index of responses to the “comfort speaking” questions generated from regression models with school fixed effects. These models compare students within the same campus, holding constant each university’s baseline characteristics, while also controlling for students’ demographics and socioeconomic status. This analysis allows us to determine whether students attending the same campus report systematically different experiences depending on how ideologically skewed they believe the faculty and student body are and whether the impact of perceived faculty and peer ideology is different for students that identify as liberal, moderate or conservative. Importantly, the models include perceived faculty ideology and perceived student ideology in the same regression (such that the impact of perceived faculty ideology reflects faculty perceptions net of peer climate and the impact of perceived student ideology reflects peer perceptions net of faculty climate).

Figure 8 – Predicted Self-Censorship and Comfort Speaking by Perceived Faculty Political Ideology

As Figure 8 shows, there is a significant “outsider penalty” being imposed on college students, with those who believe they are politically out of step with their campus’s faculty feeling far less comfortable expressing themselves and engaging in far more self-censorship. Importantly, this is not just a “penalty” imposed on campus conservatives. When students perceive faculty as very liberal, both right-leaning and moderate students report substantially higher levels self-censorship than left-leaning students. As perceived faculty ideology shifts in a conservative direction, self-censorship rises for students on the left and falls for students in the middle and on the right. Comfort expressing oneself follows a similar pattern, with comfort for right-leaning students high at the “conservative faculty” end of the scale and low at the “liberal faculty” end of the scale. In other words, the same campus can feel like an open forum or a repressive monoculture depending on how students perceive the ideological commitments of their professors.

Figure 9 – Predicted Self-Censorship and Comfort Speaking by Perceived Student Political Ideology

Student perceptions of peer ideology also show evidence of an “outsider penalty.” When students believe their fellow students are very liberal, for example, conservatives and moderates report more frequent self-censorship (Figure 9). As perceptions of the average student’s ideology shifts right, however, self-censorship declines among conservatives and moderates and rises among liberals.

Ideological skew among the faculty doesn’t just change whether students feel comfortable speaking in the abstract, it also changes whether specific substantive issues feel “off limits,” with the effect depending entirely on whether students experience themselves as aligned with (or out of step with) the campus’s ideological center of gravity.

Figure 10 – Predicted Difficulty of Discussing Controversial Issues by Perceived Faculty Ideology

Figure 11 – Predicted Difficulty of Discussing Controversial Issues by Perceived Student Ideology

Figure 12 – Predicted Number of Issues that are Difficult to Discuss by Perceived Faculty and Student Ideology

Figures10-12 show a consistent “outsider effect” across a wide range of salient political issues (and in the total number of topics students find hard to honestly discuss). When faculty and students are perceived as very liberal, conservatives and moderates are more likely to say these issues are difficult to talk about openly. When the perceived climate shifts to the right, these constraints are loosened and having frank conversations becomes easier for them. Similar to the patterns described above, this relationship flips for those on the left, with liberal students reporting more difficulty when the campus is perceived as very conservative. Taken together, these findings show that ideologically imbalanced campuses can wind up with a far narrower “Overton Window.”

Where We Do Go From Here?

These results show strong correlation and not the kind of definitive proof of causation that might come from a randomized experiment. Importantly, students who feel uncomfortable and self-censor for reasons having nothing to do with the faculty political climate (e.g. they are dispositionally conflict-avoidant) may interpret their educational environment more negatively. But three of the findings presented above are hard to ignore. First, most campuses in FIRE’s CFSR data are (correctly) perceived to have both left-leaning faculty and a left-leaning study body, meaning that many moderate and conservative students face a “double outsider” environment. Second, the perceived ideology of both professors and peers is strongly associated with educationally meaningful outcomes such as self-censorship, comfort speaking and reluctance to have open and honest conversations about salient political topics. Finally, the relationship between perceived faculty and peer ideology and these educationally meaningful outcomes hold remain strong even with full controls and even when the models include perceived faculty ideology and perceived student ideology simultaneously (suggesting that both professors and peers independently shape whether students feel free to talk).

The implication of these findings is not that colleges and universities need ideological quotas for faculty hires or student admissions. Yet this evidence should serve as a wake-up call for those who claim to want to improve education but also reflexively oppose any interventions to increase ideological diversity on the faculty on the grounds that it is little more than “DEI for conservatives.” Student self-censorship is serious educational problem that demands a strong response. If we want campuses where students test ideas openly, we cannot treat campus ideological climate as irrelevant. The evidence suggests students certainly do not.

Figure 1 – National Pride, 2001-2025 (Gallup)

Focusing only on the overall decline in national pride obscures the extent to which weakening attachment to American national identity is a generational and political problem. As Gallup’s data show (Figure 2), young Democrats (particularly those born after 1996) are far less likely to express pride in being American than Republicans, Independents, or even older Democrats. But low levels of patriotism are not just a youth problem. Even Democrats born long before 1996 have become markedly less proud of their country over the past decade. What happened?

Figure 2 –National Pride by Party and Generation (Gallup 2025)

Unfortunately, Gallup’s reporting offers few answers. It is not even clear from the figures Gallup publishes whether differences across political identification and age are the largest or most meaningful gaps to focus on. Without a breakdown by race, religion, education, and other political attitudes, it’s impossible to diagnose the causes behind the country’s collapsing national pride.

In this post, I draw on the results of a large (N=4,188) nationally representative survey from 2024 to explore the ideological, cultural, and attitudinal roots of declining American patriotism.1 In addition to revealing the substantial impact that religiosity has on feelings of national pride, the analysis presented here provides the first empirical confirmation of what many have long suspected; namely, that the strongest predictor of national pride today is whether someone is “woke” or “anti-woke” on questions of racial inequality. Even after controlling for race, education, religiosity, and political engagement, the consistently “anti-woke” are now twice as likely as the consistently “woke” to say they are “extremely” or “very proud” to be an American.

This wasn’t always the case. As I describe below, in surveys from both 1996 and 2014, views on racial inequality were only weakly correlated with feelings about the country. Views on race have recently become intertwined with conceptions of national identity in a way they were not a decade ago.

With this data in mind, it is possible to spell out a more nuanced account of American patriotism’s decline than is typically offered in mainstream outlets. Beginning in 2012, the so-called “Great Awokening” produced a profound shift in the country’s views on race, with young people on the political left endorsing increasingly “woke” answers to long-standing questions of racial inequality. As the analyses I will describe show, if we had the same amount of “wokeness” on racial inequality issues as we did 20 years ago, the percentage of Americans saying they are “extremely” or “very proud” to be an American would be 15-points higher.

Subscribe now

But it’s not just that “woke” views on racial inequality have become more popular. It’s also that “wokeness” now plays a much bigger role in shaping how people feel about the country than ever before. A decade ago, many Americans could hold progressive views on racial inequality and still manage to feel proud of their national identity. Not anymore. Today, when Americans think about whether they are proud of the country, they are likely to base that answer on how they feel about the country’s performance on racial inequality. If current levels of “wokeness” were held constant but their relationship to patriotism looked like it did in 2014, we would expect strong national pride (i.e. the percentage of people saying they are “extremely” or “very” proud of their American identity) to be about 14 points higher. In short, both the rise in “woke” attitudes and the way they’ve come to dominate people’s sense of national identity explain why patriotism has unraveled over the last two decades.

Contrary to popular explanations that attribute the left’s growing disaffection with the country entirely to Donald Trump, then, the root causes of eroding national attachment precede the MAGA era and run deeper than simple partisan politics. In fact, Trump may be as much a symptom of the cultural transformation driving the decline of American patriotism as a cause of it.

Political Identification and National Pride

The most obvious place to start any examination of American patriotism is by looking at political identification. Liberals and conservatives and Democrats and Republicans have long diverged on questions related to national identity, but, as the Gallup data show, the gap has increased sharply in the last decade. Yet, Gallup’s reporting does not tell the full story because it fails to distinguish those holding intense political commitments from those who are less stridently partisan or ideological. In Gallup’s presentation, for example, “strong Democrats” and those calling themselves “independent” but who “lean Democratic” are lumped together to provide a sense of how “Democrats” (defined in the broadest possible terms) feel about “being American.”

Figure 3 – National Pride and Political Identification (2024 CFSR National Replication Survey)

The left-hand panel of Figure 3 shows the widely reported relationship between political identification and national pride from a large, nationally representative survey conducted by YouGov in the summer of 2024. Consistent with other data, the results of this survey show a vast “pride gap” between those on the left and those on the right, with conservatives 30-points more likely to express strong pride than those on the left (54.4% to 84.0%).

As the right-hand panel of Figure 3 shows, however, there are large differences in patriotism among liberals depending on the intensity of their liberalism. Specifically, those identifying as “liberal” were vastly more likely to express high levels of national pride than those identifying as “very liberal” (62.1% to 42.2%). In fact, the patriotism gap between “very liberal” and “liberal respondents is almost as large as the patriotism gap between “liberal” respondents and “very conservative” respondents (20-points to 25-points). The extent to which highly progressive Americans stand apart on questions of national pride is hidden by the tendency to lump the American public into three broad camps (left, center and right).

As the Gallup report suggests, however, even this more fine-grained assessment of ideological gaps on patriotism obscures vast generational differences within the political left.

Figure 4 – National Pride and Ideology and Age (2024 CFSR National Replication Survey)

Figure 4 shows how age and ideology interact with each other in structuring feelings on national pride. What’s notable about the data in Figure 4 is the extent to which getting older leads to only modest increases in national pride for most Americans (i.e. those who identify as “very conservative,” “conservative,” “moderate,” and “liberal”). Among “very liberal” Americans, however, age is a determinative factor. As Figure 4 illustrates, while roughly 30% of liberals under 30 express high levels of national pride, nearly 60% of “very liberal” seniors do.

The stark patriotism gap displayed in Figure 4 and confirmed in Gallup’s reporting raises as many questions as it answers. For example, none of this data tells us whether there are different levels of national attachment based on key demographic variables such as race, religiosity or education. Most importantly, it cannot begin to help us understand why young progressives are decidedly less prideful of the United States than other Americans.

In an effort to paint a richer picture of the factors shaping American patriotism, Figures 5 through 13 show how strong national pride varies across race, education, religiosity, and political engagement.

Race and National Pride

The data shown in Figure 5 defy some common assumptions about racial differences in national pride. Contrary to conventional wisdom, there are no large differences in patriotic sentiment among white, Black, and Hispanic Americans. In fact, more than two-thirds of each group say they are “extremely” or “very” proud to be American. Asian Americans, however, are significantly less likely than members of these other groups to express high levels of national pride, with only 52.8% saying they are “extremely” or “very” proud to be American.

Figure 5 – National Pride and Race (2024 CFSR National Replication Survey)

Education and National Pride

There’s been considerable attention paid over the last two election cycles to the “diploma divide” in politics (i.e. the widening gap in political attitudes and voting behavior between Americans with and without a college degree). At the same time, a parallel debate over whether higher education breeds anti-American sentiment among students has emerged, with many critics (mostly on the political right) claiming that colleges and universities now teach young people to hate their own country.

Figure 6 – National Pride, Age and Education (2024 CFSR National Replication Survey)

The data suggest a complicated relationship between national identity and education. As Figure 6 shows, while there is a modestly-sized “diploma divide” in patriotism, it does not follow the pattern that most observers might expect. Among older Americans, college graduates are, in fact, somewhat less likely than non-graduates to express high levels of patriotic attachment. This relationship, however, reverses among “Zoomers,” with young, college-educated adults being more likely to say they are “extremely” or “very” proud to be American than their non-college-educated peers. In other words, rather than eroding patriotic sentiment among young people, higher education now appears to bolster it.

Religiosity and National Pride

The connection between religiosity and national pride has received scant attention in most mainstream accounts of patriotism. This is unfortunate. In my 2024 survey, across multiple measures, self-reported religiosity was strongly related to national pride.

Figure 7 – National Pride and Religiosity (2024 CFSR National Replication Survey)

As Figure 7 shows, those who attend church and pray more frequently are far more likely than those who never do these things to say they are “extremely” or “very” proud to be an American. Subjective reports of the importance of religion in one’s life show a similar relationship. Collectively, the data in Figure 7 suggest that, for many Americans, religious commitment and national identity are closely intertwined.

Political Engagement and National Pride

There are good reasons to believe that political engagement will also be associated with feelings of national pride. Individuals who are more politically engaged typically feel a stronger connection to their fellow citizens and to national political institutions. Further, regularly consuming political news reflects a deeper investment in the future of the country, a stronger sense of political efficacy, and a belief that change is possible through civic participation. By contrast, political disengagement is likely to be associated with feelings of alienation, distrust, or apathy (all of which can undermine patriotic sentiment).

Figure 8 – National Pride and Political Engagement (2024 CFSR National Replication Survey)

As Figure 8 shows, those who are politically engaged are, in fact, significantly more patriotic than those who are not. The patriotism gap between the political engaged (i.e. those who follow political news “most of the time”) and the politically disengaged (i.e. those who follow political news “hardly at all”) is nearly 20-points (56.7% to 74.4%).

Racial Attitudes and National Pride

As discussed above, the role of racial attitudes has loomed larger than perhaps any other factor except political ideology in mainstream debates about patriotism. Racial attitudes, of course, are tricky to measure, however. The primary approach in the social sciences has been to construct an index from responses to the following four statements (where responses can range from “strongly agree” to “strongly disagree”):

• “Irish, Italians, Jewish and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors.”

• “It's really a matter of some people just not trying hard enough: if blacks would only try harder they could be just as well off as whites.”

• “Generations of slavery and discrimination have created conditions that make it difficult for blacks to work their way out of the lower class.”

• “Over the past few years, blacks have gotten less than they deserve.”

There’s a lively academic debate about how agreement or disagreement with these four statements should be labeled and interpreted. While agreement to the first two statements and disagreement with the last two statements have historically been described as evidence of “racial resentment,” “symbolic racism,” “modern racism” or, in less nuanced accounts, “racism,” recent work has suggested that an entirely different, non-racialized label should be used.

There are several reasons to avoid describing responses to these statements in language that implies overt anti-Black animus. First, the items intentionally conflate conservative ideological beliefs (such as individualism, personal responsibility, and meritocracy) with attitudes about African Americans as a group, making it impossible to interpret responses as straightforward expressions of racial prejudice2. Second, there is a growing body of work showing the items cannot be accurately described as a measure of “anti-Black affect” because they “measure attitudes toward any group, rather than African Americans alone.” Finally, some research has demonstrated that, “contrary to the scale’s typical title, it measures sympathy for blacks as least as much as resentment toward them.”

All of this is to say that there is no consensus in academic circles about how to describe responses to these statements. Some have proposed framing them as measures of “just-world orientation” given that they reflect beliefs about work ethic and fairness more than feelings about African Americans. Others suggest the collection could be better described as a “racial sympathy” for African Americans since it “captures favoring of Blacks substantially more than disfavoring.” Still others recommend referring to them as the “Structural vs. Individual Attributions for Black Americans’ Status” scale. The most generic approach (and least informative) approach is to simply discuss them as measures of “views on racial inequality.”

In popular discourse, these measures have become synonymous with the “Great Awokening," the sharp rise in progressive, race-conscious attitudes (especially among white liberals) on issues of identity, inequality, and social justice. As Figure 9 shows, the way liberals answer these questions changed dramatically after 2012.

Figure 9 – Racial Attitudes Over Time (VOTER Survey)

Rather than use the uninformative label “racial inequality views,” which does not indicate what high and low scores on the scale mean, I will settle here on the imperfect description of these items as “woke vs. anti-woke” views on racial inequality.3

Regardless of their label, however, it is obvious how such views might impact patriotic attachment. For those viewing the contemporary United States through the lens of historical and ongoing inequality and injustice, expressions of national pride are likely to be seen as an endorsement of a political, economic, and social system that should be drastically reformed or completely dismantled. Patriotism, from this point of view, might be incompatible with a commitment to “social justice.”

Yet, hard evidence about exactly how much “woke” racial attitudes matter for feelings of national pride is in relatively short supply. Unfortunately, there are very few surveys that ask questions about both patriotism and racial attitudes. Major academic surveys (e.g. the General Social Survey, the American National Election Study, the Cooperative Election Study) that routinely tap racial attitudes, for example, have only infrequently included items focused on national attachment. Polling organizations that frequently ask questions about national attachment (e.g. Gallup, Pew Research Center) do not often measure racial attitudes in the same surveys. The lack of datasets that simultaneously track racial attitudes and national pride means arguments linking the two are necessarily speculative.

The only good data we have to empirically identify the relationship between racial attitudes and feelings of national attachment comes from the General Social Survey (GSS). The GSS has asked the first of the “woke vs. anti-woke” views on racial inequality questions (“Irish, Italians, Jewish and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors”) since 1994. In 1996 and 2014, these items were included alongside questions about American national identity.4 Specifically, the GSS asked two questions connected to national pride: (1) a question asking respondents: “How proud are you of America’s history?”; (2) a question asking respondents for their level of agreement with the statement: “I would rather be a citizen of America than of any other country in the world.”

Figure 10 – Pride in American History and “Woke vs. Anti-Woke” Racial Attitudes (GSS)

Figure 11 – Pride in American Citizenship and “Woke vs. Anti-Woke” Racial Attitudes (GSS)

Figures 10 and 11 show the predicted impact of agreement with the idea that blacks “work their way up” without “any special favors” and these two measures of national attachment (generated from a regression model controlling for age, ideology, income, race, and education). As the slopes of the lines indicate, in both 1996 and 2014, more agreement with the idea that blacks should “work their way up” without “special favors” was predicted to lead to more national pride. As the slopes also indicate, however, the impact was very small (though statistically significant).

Figure 12 – National Pride and “Woke vs. Anti-Woke” Racial Attitudes (2024 CFSR)

But how do things look since the “Great Awokening”? Fortunately, my 2024 survey included each of the four items required to measure “woke vs. anti-woke” views on racial inequality.

Figure 12 shows the relationship between views on racial inequality and feelings of national pride. As Figure 12 clearly shows, those with the most “woke” views on racial inequality are far less patriotic than those with the most “anti-woke” views on racial inequality. For example, those who “strongly agree” with the idea that blacks should “work their way up….without any special favors” were twice as likely as those who “strongly disagree” to say they are “extremely” or “very proud of being American (39.3% to 89.5%). In other words, those who see America as fundamentally unjust on questions of race are now far less likely to express pride in being part of it than they were in the past.

A Multivariate Analysis of National Pride

The relationships shown in Figures 3 through 12 make it clear that political identifcation, race, education, religiosity, political engagement, and racial attitudes are all correlated with national pride. They do not tell us, however, which of these factors matters most. The primary problem with looking at simple bivariate relationships like the ones in the figures above is that all of these factors are deeply interrelated. Whites, older people and the highly religious, for example, are vastly more likely than non-whites, young people and the secular to identify as politically conservative. As a result, any apparent relationship between a single variable and patriotism may be misleading. To truly determine the unique influence of each factor individually, we need a multivariate analysis that holds the other factors constant. Only then will we be able to assess whether all (or only just a few) of these characteristics are truly important for shaping feelings of national pride.

Figure 13 – Multivariate Analysis of National Pride (2024 CFSR National Replication Survey)

With this in mind, I ran a logistic regression model predicting strong national pride (i.e. the respondent saying they are “extremely” or “very” proud to be an American) on the basis of the factors described above; namely, racial attitudes (an index summarizing “woke” vs. “non-woke” views on racial inequality), race, education, religiosity (an index of church attendance, prayer frequency and self-reported religious significance), political news interest, age, ideology and their interaction.

The coefficients for each of these variables are presented in Figure 13.5 Negative coefficients indicate that individuals in the labeled category are less likely to say they are “extremely” or “very” proud to be an American while positive coefficients indicate that individuals in the labeled category are more likely to say they are “extremely” or “very” proud to be an American. When the confidence intervals (i.e. the shaded areas around each coefficient estimate) touch the “0” line, they are not statistically significant. When the confidence intervals do not touch the “0” line, they are significant.

With few exceptions, the results in Figure 13 tell the same basic story about the determinants of patriotism that Figures 3 through 12 tell. Even with a full range of controls, religiosity, political engagement, education, age, ideology and the interaction terms exert a significant influence on how Americans feel about their national identity.

The most striking finding to emerge from this set of analyses is the fact that “woke vs. anti-woke” views on racial inequality hav the most consistent, direct, and powerful impact on national pride. To illustrate this point, Figure 14 shows the predicted probability (derived from the fully specified logistic regression model) of being “extremely” or “very” proud to be American based on racial attitudes. Even after controlling for all other factors, “wokeness” has a profound impact on patriotism, with shifts from the most to least “woke” predicted to double an individual’s chances of expressing strong national pride.

Figure 14 – National Pride and Woke Racial Attitudes (2024 CFSR National Replication Survey)

The steep slope of the line in Figure 14 is a stark contrast to the slope of the lines in Figures 10 and 11. This suggests that the relationship has likely become much stronger over the last decade. In other words, two things have occurred simultaneously to drive down patriotism: the number of Americans holding “woke” views has increased and these views are increasingly determining how one feels about the country.

A Post-Patriotism America?

Patriotism means love of country (critically rather than blindly) and is bolstered by a shared consciousness of what it means to be an American. Twenty five years ago, Americans were nearly unanimous in saying they were “extremely” or “very” of their national identity. But that national unity has eroded due to a combination of political and generational changes. Liberals today are much less likely than in the past to express pride in their country. Additionally, Millennials and Zoomers are much less proud of their country than their elders are.

There’s a temptation in an analysis like this to focus exclusively on Trump. It’s true that pride among liberals has risen and fallen along with his time in office over the last eight years. His symbolic role in American politics cannot be ignored and there’s little doubt that his dominance in the discourse has not helped young progressives to feel better about their country.

But it is also crucial to realize that the decline in patriotism began well before Trump took office and it appears to have its origins in longer-standing questions about American society. To be more precise, one of the central causes of disaffection with the country is the growing “wokeness” of young progressives. As shown above, not only has the number of Americans endorsing a “woke” view of racial equality increased but these views are increasingly tied to how these Americans feel about their country.

The main question going forward is whether “very liberal” and “woke” Zoomers and members of Gen Alpha (those born after 2010) can reconcile their progressive values with a sense of pride in their nation. It is possible, of course, that liberal Zoomers’ disaffection with the country is temporary. Perhaps time will foster a greater sense of national attachment in the same way that Boomers seem to have come to terms with the bitterness of the Vietnam era. Commentators suggest that the disaffection of Gen-Z stems from spending their formative years in a period defined by political polarization, economic uncertainty, COVID, and Black Lives Matter protests of 2020. If these experiences fade from memory, Gen Z may age out of their alienation.

Nevertheless, globalization and multiculturalism continue to gnaw away at the prioritization of a common national identity and the value of citizenship. The zeitgeist dominating the elites in academia, the media, and multinational corporations denigrates rather than celebrates American history and current social progress. What’s more, loyalties in place at age 30 rarely change much.

It’s easy to be dismissive about the decline of patriotism. For some, the trends in Figures 1 and 2 will feel like progress, a kind of validation that fewer Americans are embracing what they see as a version of national pride that excuses inequality and enables injustice. But patriotism – when properly conceived as critical love of one’s country and a commitment to ideals rather than myths – has long served as a vital form of “civic glue.” It links people across lines of race, class, religion and ideology by offering a shared sense of purpose, obligation, and destiny. It motivates civic participation, national service, and self-sacrifice. Perhaps most importantly, it fosters the kind of institutional trust and commitment to democratic norms that pluralistic societies need to endure.

Ultimately, this is why the current decline in patriotism is significant. Not because pride in America needs to remain high for its own sake, but because a country with millions of citizens who view it as shameful and morally illegitimate is a country that will be increasingly difficult to hold together. Americans can, should and will debate how to remedy inequality. But if we lose the sense of a shared national “we,” the talk may not be of “One America” but of “No America.”

Figure 1 – The American Public’s Confidence in Higher Education

This profound erosion of trust is not limited to adults. According to Monitoring the Future (MTF), a long-running survey of American 12^th graders, the percentage of high school seniors saying colleges do a “good” or “very good” job for the country cratered between 2007 and 2023, falling from 75% to 47%. As Figure 2 shows, the decline has occurred across political and demographic lines, with young Republicans, Democrats, Independents, whites, African Americans, and Hispanics, all souring on higher education over the last 15 years.

Figure 2 – Positive Evaluations of Colleges and Universities among High School Students

This “crisis of confidence” has unfolded alongside intensifying concerns about protest and free speech on college campuses. In particular, the growing popularity of the “heckler’s veto” – attempts by an individual or group to disrupt or restrict speech through excessive noise, intimidation, or violence – among college students has subjected universities to withering public scrutiny. According to data from the Foundation for Individual Rights and Expression’s (FIRE) Campus Deplatforming Database, the number of “disruptive” incidents has, with the exception of a brief pause during the COVID pandemic, been rising since 2014 (Figure 3). Particularly striking is the explosion in the “heckler’s veto” during the last three years, with events labeled by FIRE as “attempted disruptions” or “substantial disruptions” of campus speaking events growing from just 3 in 2021 to 79 in 2024.

Figure 3 – Attempted Uses of the “Heckler’s Veto” on College Campuses

But correlation is not causation and, to date, there has been little empirical evidence attempting to systematically link the growing use of the “heckler’s veto” on college campuses to the public’s changing attitudes about higher education. As a result, it is unclear whether university leaders should be concerned about the trend displayed in Figure 3 and what, if anything, they should do when disruptive protests shut down speech on their campuses.

This post presents the first empirical test of whether the rising use of the “heckler’s veto” on college campuses has contributed to the erosion of public trust in higher education. Drawing on an experiment embedded in a large, nationally representative survey, I examine whether exposure to video footage of a single disruptive protest aimed at deplatforming a speaker on a college campus alters how people feel about higher education.

The results are clear and striking. Viewing even a brief, 20-second snippet of footage showing students using the “heckler’s veto” at a campus speaking event significantly worsens attitudes toward higher education. Equally important, the results show that how universities react in the wake of a disruptive protest makes little difference. Whether university leaders respond by issuing statements condemning or praising the protesters does nothing to mitigate the harm caused by exposure to images of disruptive protest. In other words, once the “heckler’s veto” is seen, the reputational damage is done. Words from university leaders fall on deaf ears.

These findings arrive at a time when university leaders are struggling to address higher education’s “crisis of confidence.” In a recent survey conducted by Inside Higher Ed, 57% of college presidents said they have been “not at all effective” in responding to the public’s declining confidence in higher education. Yet the results presented here suggest a path forward for university leaders. Colleges and universities must couple clear, politically neutral policies with robust instruction in the norms of peaceful protest. Restoring the public’s confidence in higher education will depend not on how colleges and universities respond after disruptive protests occur but on whether they prevent them from happening in the first place.

As the data presented below also shows, in order to effectively mitigate the “heckler’s veto” problem, universities will have to confront the fact that support for silencing campus speech through disruptive action is not equally distributed across student groups. Counterintuitively, the groups that have benefitted most from robust norms emphasizing the value of free speech, political tolerance, and peaceful protest – progressives and members of national racial, gender and sexual minority groups – are now the least supportive of them. Additionally, the data reveal significantly greater willingness to endorse disruptive action as a means of preventing campus speech among Ethnic and Gender Studies students than among students of any other major. If universities would like to bolster the free speech climates of their campuses and restore their good standing with the public, they will have to design interventions that take these large attitudinal disparities into account.

Colleges and universities must come to grips with the fact that disruptive protests about speech are no longer merely internal campus disputes. Increasingly, instances of the “heckler’s veto” are visible public signals that undermine the fragile legitimacy of higher education in the eyes of the American public. Because statements issued by campus leaders in the wake of disruption do little to repair the reputational damage caused by protests, the only viable strategy for colleges and universities is prevention. In practical terms, this means creating campus cultures where protest is protected (even encouraged) but does not descend into illiberal shout downs, blockades and violence.

Testing the Impact of the “Heckler’s Veto”

There’s been no shortage of commentary blaming the “free speech crisis” on college campuses to the “crisis of confidence” in higher education. Unfortunately, there’s been more speculation than hard data about this connection. Despite widespread speculation, however, empirical evidence supporting this connection has been nearly non-existent.

In an initial attempt to determine whether the growing use of the “heckler’s veto” on college campuses might explain the public’s growing cynicism about higher education, I embedded an experiment in a large (N=4,188) nationally representative survey of the American public conducted by YouGov in late 2024.1

The primary treatment in the experiments (shown above) was a professionally produced, 30-second local news clip of real protest footage (taken from Milo Yiannopoulos’s 2017 appearance at UC Berkeley and Charles Murray’s 2017 appearance at Middlebury College) repackaged as occurring at fictional “Jefferson University.” In the clip, a news anchor explained to viewers that “a disruptive protest at Jefferson University forced the cancellation of a talk by author Steven Jones on Thursday night. While students outside the event blocked entrances, started fires and threw debris at attendees, students inside blew airhorns and shouted slogans whenever Jones spoke. At 8:15, campus police shut down the event, citing safety concerns.” In addition to a control group that did not watch this video, respondents in my experiment were randomly assigned to one of three experimental conditions connected to this video:

1. The “Video without a University Statement” Condition: Respondents assigned to this condition watched a version of the news clip that showed video of the disruptive protest but did not include a statement from the university’s president about the disruption.

2. The “Video with a University Statement Condemning Protesters” Condition: Respondents assigned to this condition watched a version of the news clip that showed video of the disruptive protest and was immediately followed by a statement from the university’s president condemning the protesters (i.e. a video including the statement “The University strongly condemned the protestor’s actions. In a statement, University President James Marshall said, “Last night’s disruptive actions are unacceptable. Throwing debris, blocking entrances, and shouting down speakers are serious violations of our Student Honor Code and antithetical to our core values of informed inquiry and the respectful exploration of diverse perspectives. The University is conducting an investigation and will hold student protestors accountable.”). This statement was modeled on actual statements issued by university leaders following campus disruptions over the last few years (e.g. the statements issued by the University of California, Berkeley in the wake of 2019 protests against Ann Coulter).

3. The “Video with a University Statement Praising Protesters” Condition: Respondents assigned to this condition watched a version of the news clip that showed video of the disruptive protest and was immediately followed by a statement from the university’s president praising the protesters (i.e. including the statement, “The University did not condemn the protestor’s actions. In a statement, University President James Marshall said, “Our campus values different ideas but we also believe in promoting an inclusive community…Thank you to the students who participated in Thursday’s event. It took bravery to stand in a challenging space. I am proud of you for demonstrating the value of free speech and the right to protest.”). This statement was also modeled on actual statements issued by university leaders following campus disruptions over the last few years (e.g. the statement issued by San Francisco State University in the wake of disruptive protests against Riley Gaines).

Immediately following exposure to this video clip, respondents were asked two questions about higher education: (1) Please tell me how much confidence you, yourself, have in higher education; and (2) Are colleges and universities having a positive or negative effect on the way things are going in the country these days? A control group was asked these questions without seeing video footage of a disruptive protest. By comparing responses to these questions across the treatment and control groups, it is possible to isolate the effect of exposure to images of the “heckler’s veto” on public attitudes toward higher education. Additionally, the design allows us to assess whether statements by university leaders (whether condemning or praising the disruptive protest) meaningfully influence those perceptions.

Figure 4 – Perceptions of Higher Education by Exposure to Videos of the “Heckler’s Veto”

Figure 4 shows variation in responses to these two questions across each of the four treatment conditions. As Figure 3 shows, exposure to video footage of the “heckler’s veto” harmed perceptions of higher education independent of whether or how universities responded. To illustrate in substantive terms, the percentage of respondents expressing “very little” confidence in higher education rose from 16.1% among respondents not viewing the video to 22.0% among those seeing any version of the video. Similarly, the percentage of respondents saying colleges and universities have a “negative effect” on “the way things are going in the country these days” increased from 43.3% among those not watching the video to 48.3% among those watching the video. In other words, simply seeing images of the “heckler’s veto” soured members of the public on universities. A presidential statement, regardless of its perspective on the protesters, did nothing to restore confidence.

Given that Republicans have become more negative in their assessments of higher education than Independents and Democrats over the last decade, we might expect the treatment effects in Figure 4 to vary by partisanship. In order to explore this possibility, my analyses included an interaction between party identification and treatment conditions. The predicted margins from these analyses are presented in Figure 5.

Figure 5 – Perceptions of Higher Education Among Partisans by Exposure to Videos of the “Heckler’s Veto”

As Figure 5 shows, perceptions of higher education in the “Control – No Video” condition are more positive than in the three video treatment conditions across all partisan groups. Although only Republicans show a consistently, statistically significant decline when exposed to the various video treatments, Democrats and Independents also show non-significant downward trends across conditions. A formal adjusted Wald test of the “treatment*party identification” interaction terms for both dependent variables revealed no evidence that treatment effects actually differ by party identification. Put less formally, exposure to the footage of the “heckler’s veto” reduced confidence in higher education, with the effect being strongest among Republicans.

Collectively, these findings challenge the widespread assumption that it is only Fox News- watching conservatives that are troubled by the growing use of the “heckler’s veto” on college campuses. In fact, the broader public, regardless of how university leaders respond, is negatively affected when it sees college students engaged in illiberal actions to stop campus speakers.

While the shifts shown in Figures 4 and 5 are relatively modest, they capture the effect of only a single exposure to a stylized incident of disruptive campus protest. In reality, the American public has been subject to a nearly relentless stream of these images in recent years. Over the past decade (Figure 3), the “heckler’s veto” has become a recurring storyline in news coverage and social media debates about higher education. The cumulative impact of this recurring exposure likely exceeds what this one experiment can measure. In other words, even small experimental effects like the ones shown in Figures 4 and 5 offer compelling evidence that the real-world explosion in the “heckler’s veto” on college campuses is playing an important role in worsening higher education’s legitimacy crisis.

What is to be Done?

The Educational Cure for the Heckler’s Veto

If the rising use of the “heckler’s veto” is contributing to higher education’s “crisis of confidence,” what can colleges and universities do to redeem themselves in the public’s eyes? The temptation for risk-averse administrators might be to eliminate the potential for reputationally damaging disruption by preemptively deplatforming controversial speakers altogether. Sadly, there’s evidence that some universities are giving in to this temptation. At the University of Southern California, for example, administrators canceled all of their scheduled commencement speakers after pro‑Palestinian social media posts by valedictorian Asna Tabassum triggered concerns about “a security threat.”

But preemptive cancellations are likely to do far more harm than good. Instead of restoring trust and confidence in colleges and universities, these cancellations risk reinforcing the damaging perception that institutions of higher education are unwilling to defend free expression. If colleges and universities hope to prevent the reputational harms caused by disruptive protests, they must teach their students how to protest in ways that are peaceful, principled and, most importantly, non-coercive.

There are already numerous models for providing this kind of education. FIRE, in partnership with New York University’s First Amendment Watch, for example, offers free orientation modules that discuss counter-speech strategies, the legal limits on disruptive action, and the difference between protest and censorship. Similarly, interested universities might follow the lead of schools like Washington University, the University of Wisconsin-Eau Claire, and Appalachian State University, and distribute free expression videos (describing principles and policies about campus speakers, student protests and classroom instruction) as part of their student orientation programs. Embedding these messages in orientation activities at the start of every student’s college education is likely to signal that civil protest is welcome but must be limited by rules against shout-downs, blockades, and violence.

Figure 6 - Do Disruptive Actions Increase or Decrease Support?

Students should also be educated about the ineffectiveness of illiberal political action. One study of more than 300 movements taking place since 1900, for instance, found that violent tactics repel potential supporters while non-violent ones can attract influential allies. A poll conducted after activists threw tomato soup at Van Gogh’s Sunflowers in 2022 found that 46 percent of Americans said they were less inclined to support efforts to address climate change as a result of the disruptive action (Figure 6). The message should be made plain to students: violence, intimidation, and disruption hurt the causes they claim to care most deeply about.

Targeting the Core of the “Heckler’s Veto” Problem

If these educational interventions are to succeed, they will have to pay particular attention to reaching the segments of the university community most supportive of illiberal, disruptive political action. Who are these students? Fortunately, we have good data to help us answer this question.

Every year since 2022, FIRE has conducted a large, nationwide survey of college students as part of their efforts to rank the free speech climates of American universities. The 2024 survey, for example, included 58,000 enrolled students at over 250 colleges around the country. In addition to questions about demographic characteristics (race, gender, sexual orientation, etc.), political identities (e.g. political ideology, partisanship, etc.) and areas of academic study (i.e. college majors), the FIRE survey also asked respondents “How acceptable [always, sometimes, rarely, never] would you say it is for students to engage in the following action to protest a campus speaker?”:

1. Shouting down a speaker to prevent them from speaking on campus.

2. Blocking other students from attending a campus speech.

3. Using violence to stop a campus speech.

Perhaps unsurprisingly given the data in Figure 3, the FIRE survey data shows a troubling increase in the number of college students who are supportive of illiberal attempts at limiting speech on campus. As Figure 7 shows, in 2024, over half of students (52%) said that blocking other students from attending a campus speech is at least “rarely” acceptable, up from 45% in 2023 and 37% in 2022. Even more concerning, nearly one-third of students (32%) believe that “using violence to stop a campus speech” is at least “rarely” acceptable, up from 27% in 2023 and 20% in 2022.

Figure 7 – Acceptability of Shout Downs, Blockades and Violence to Stop a Campus Speech

More importantly, FIRE’s survey data allows us to identify which types of college students are most likely to view use of the “heckler’s veto” as “acceptable” and, thus, which groups should be prioritized in speech-related educational interventions.

For ease of presentation, I used FIRE’s survey data to construct a measure of support for the entire repertoire of tactics falling under the broad umbrella of the “heckler’s veto” (i.e. shout downs, blockades and violence). Based on their responses to the three questions listed above, I classified each student in FIRE’s sample as either a “disruption-endorsing student” (i.e. someone who said it is “always,” “sometimes,” or “rarely” acceptable to use shout downs, blockades, AND violence to stop a campus speech) or a “disruption-rejecting student” (i.e. someone who said it was “never” acceptable to use shout downs, blockades OR violence to stop a campus speech). It should be pointed out here that this an extremely conservative definition of “rejection.” Indeed, even students who fully endorse shout downs and blockades would still be categorized as “disruption-rejecting” if they said “using violence” was “never” acceptable.

Figure 8 – Percent of “Disruption-Endorsing” Students by University

Overall, in 2024, 26.8% of college students were “disruption-endorsing” and 73.2% were “disruption-rejecting.” Yet, “disruption-endorsing” students are far more numerous at some campuses than at others. As Figure 8 shows, there are eight schools (Macalester College, Oberlin College, Northeastern University, University of Pittsburgh-Pittsburgh Campus, Boston University, California Institute of Technology, University of Minnesota-Twin Cities, University of California-Davis) where more than 40% of students say that shout downs, blockades and using violence are ALL at least “rarely” acceptable tactics to stop a campus speech.

As Figure 8 also shows, however, there is nothing inevitable about having a student population that widely endorses use of the “heckler’s veto.” Only 6% of students at BYU and only 11% of students at Hillsdale College say that shout downs, blockades and using violence are all at least “rarely” acceptable.

Actual instances of attempted deplatforming via the “heckler’s veto” depend, of course, on much more than just how students feel about disruptive action. Some campuses regularly invite controversial figures to speak on campus (creating opportunities for students to employ the “heckler’s veto) while others do not. As a result, even schools with the largest number of “disruption-endorsing” students may never experience the “heckler’s veto.” In other words, since the opportunities for disruptive protest are unevenly distributed across institutions, student attitudes and real-world incidents of the “heckler’s veto” are poorly correlated. Nevertheless, the data in Figure 8 should prompt universities with large numbers of “disruption-endorsing” students to proactively institute programs focused on addressing their campus’s attitudinal illiberalism. Support for shout downs, blockades and violence is malleable and campuses at the top of Figure 8 should reflect on features of their educational environment that are leading so many students to endorse the “heckler’s veto.”

Figure 9 – “Disruption-Endorsing” Students by Major

It is also important for university leaders to realize that, within campuses, some groups of students are more likely than others to endorse shout downs, blockades and violence as an acceptable means of preventing campus speech. Figure 9, for example, shows differences across broad categories of academic concentration. Generally speaking, students majoring in Ethnic and Gender studies (e.g. “African American Studies,” “Ethnicity and Race Studies,” “Women’s and Gender Studies,” etc.) are significantly more likely than students from other disciplines to say that illiberal disruptions are “acceptable.” For example, in 2024, while more than half of Ethnic and Gender studies majors say that using violence is “always” (5.5%), “sometimes” (22.2%) or “rarely” (24.1%) acceptable in order to stop a speech from being delivered on campus, less than a third of Humanities, Visual and Performing Arts, STEM, Social Sciences, and “undecided” or “other” majors said violence is ever “acceptable.” As Figure 9 also shows, these numbers are trending in the wrong direction, with the percent of students qualifying as “disruption-endorsing” doubling within nearly every major since 2022.

The differences shown in Figure 9 could reflect self-selection or socialization. For example, students holding more “disruption-endorsing” attitudes may be drawn to disciplines (such as African American Studies, Women’s Studies and Chicano/Chicana Studies) that relentlessly critique existing norms, structures, and institutions. At the same time, however, the course content, pedagogical approaches, and campus subcultures associated with programs in Ethnic and Gender Studies may further legitimize disruptive protest tactics, including shout downs, blockades, and violence. Regardless of the exact nature of the causal relationship, majors with disproportionately high numbers of “disruption-endorsing” students should be prioritized in university efforts to reduce support for the “heckler’s veto.” Carefully targeted educational interventions in these departments may be especially effective in reducing embarrassing instances of disruptive protest and restoring the university’s role as a forum for open inquiry and reasoned debate.

Figure 10 – “Disruption-Endorsing” Students by Gender

Attitudes toward the “heckler’s veto” vary by more than just chosen major, however. There are also profound differences in support for shout downs, blockades and violence based on identity and ideology. Generally speaking, students who identify with historically marginalized groups are significantly more likely than students who do not to view the tactics of the “heckler’s veto” as “acceptable.” For example, nearly 60% of agender students (N=379) and almost half of genderqueer (N=705), non-binary (N=1,249) and students who were unsure about their gender identity (N=299), believe that “using violence” can “always,” “sometimes,” or “rarely” be an acceptable response to an offensive speech on campus. By contrast, only 32% of men and 29% of women share this belief (though, as Figure 10 shows, the percentage of male and female students who qualify as “disruption-endorsing” has doubled since 2022).

Figure 11 – “Disruption-Endorsing” Students by Race

Similarly, support for the “heckler’s veto” varies considerably based on racial identity (Figure 11). While non-white support for violence is relatively low overall, Asian (39.8%), Hispanic (34.8%), Black (39.9%), and Middle Eastern students (40.9%), were significantly more likely than white students (27.1%) to say that “using violence” was “always,” “sometimes” or “rarely” acceptable in 2024.

Figure 12 – “Disruption-Endorsing” Students by Sexual Orientation

The same pattern holds across sexual orientation. As Figure 12 shows, students not identifying as heterosexual express far more support for disruptive action than students identifying as heterosexual.

Figure 13 – “Disruption-Endorsing” Students by Political Ideology

Finally, political ideology appears to play a role in shaping how students view disruptive action in response to campus speech. Left-leaning students (i.e. those identifying as “Democratic Socialists” and “Liberals”) are more accepting of shout downs, blockades and “using violence” than moderate and right-leaning students (i.e. those identifying as “Conservative” and “Libertarian”). As Figure 13 shows, this ideological gap has declined slightly since 2022, with more moderate and right-leaning students embracing the “heckler’s veto” in recent years.

Figure 14 – Determinants of “Disruption Endorsing” among College Students

In addition to being correlated with support for the “heckler’s veto,” college major, gender identity, race, sexual orientation and political ideology are also likely related to each other. It is possible, therefore, that some of the relationships illustrated in the above graphs are spurious. In order to account for this possibility, Figure 14 shows the results of a multivariate logistic regression model predicting the probability of qualifying as a “heckling apologist.” More specifically, the coefficients in Figure 14 show the effect of moving from the “excluded” category of each variable to the category listed in the left-hand column of the plot. For example, the coefficients for each major (e.g. “Business,” “Humanities and Performing Acts,” “Social Sciences,” etc.) indicate the effect of majoring in one of those fields relative to majoring in “Ethnic, Gender or Area Studies.” Similarly, the coefficients for “man” and “woman” show the effect of identifying as a man or a woman relative to identifying with one of the other gender identities (i.e. “agender,” “genderqueer,” “non-binary,” “unsure”). The coefficients under the headings “Race,” Ideology,” and “Sexual Orientation,” show the effect of each category relative to “Whites,” those ideologically on the “Left,” and “Heterosexuals,” respectively. Negative coefficients indicate that individuals in the labeled category are less likely to say that shout downs, blockades and violence are ALL at least “rarely” acceptable and positive coefficients indicate that individuals in the labeled category are less likely to say these things.

Taken together, the data shown in Figures 8 through 14 demonstrate that support for the “heckler’s veto” is not randomly distributed across college campuses but, instead, reflects deeper patterns connected to social identity, political ideology and academic environment. While the multivariate regression analysis shown in Figure 14 confirms that each of these dimensions exerts an independent impact on disruption attitudes, it also underscores the complexity of addressing the problem of the “heckler’s veto.” If institutions of higher education are serious about reducing speech disruptions and restoring public trust, they must design educational interventions that reflect this complexity. This will mean focusing their efforts on the disciplines and campus communities where support for shout downs, blockades, and violence have taken hold. The success of any strategy to protect free speech on campus will ultimately depend on engaging these students directly.

From Crisis to Credibility: Rebuilding Trust in Higher Ed by Reducing Disruption

Higher education’s “crisis of confidence” has many midwives (including, but not limited to spiraling tuition costs, disappointing job‑market prospects, and partisan assaults from journalists, elected officials, and political activists). As the results presented above show, however, no accounting of the declining reputation of colleges and universities is complete without an understanding of the “free speech crisis” on America’s campuses. Viral shout downs, blockades and acts of violence in lecture halls have become one of the public’s principal lenses for viewing campus life, and, as the results shown above demonstrate, the picture Americans are receiving is corrosive. To be more precise, the experimental results discussed above clearly show that witnessing a single use of the “heckler’s veto” on a college campus significantly hurts public trust in higher education. Just as importantly, statements from university leaders alone cannot repair this damage.

Public patience with universities is running out. Even left-leaning and once-friendly voices are raising alarms about the speech problems on college campuses. Most notably, in an April 2025 forum at Hamilton College, former president Barack Obama took a strong and explicit stand against the “heckler’s veto,” arguing that “The idea of cancelling a speaker who comes to your campus, trying to shout them down and not letting them speak, even if I find their ideas obnoxious, well, not only is that not what universities should be about, that’s not what America should be about. You let them speak, and then you tell them why they’re wrong. That’s how you win the argument.”

The good news is that proactive educational programs that channel dissent into non-disruptive forms of action have a chance at reducing disruptive protest and, in the process, slowing the reputational decline of higher education. Colleges and universities should loudly and frequently remind students that, in the marketplace of ideas, persuasion beats coercion.

College students will likely always feel the impulse to protest speech they find offensive. The challenge for higher education is not to suppress that impulse but, instead, to channel it into forms of expression that uphold liberal values and protect the legitimacy of the institution itself.

Veterans and the Unprecedented Recruitment Crisis

The armed forces are struggling to address an unprecedented recruitment crisis, with three of America’s four major military services failing to hit enlistment targets in 2023. The military’s recruitment problems are posing a growing threat to national security. The active-duty Army currently stands at 445,000 soldiers, 41,000 fewer than only three years ago and the smallest it has been since 1940. As Gil Barndollar and Matthew C. Mai argued in September, “America isn’t ready for another war — because it doesn’t have the troops.”

Why is the military shrinking? Most diagnoses of the recruitment crisis focus on the declining number of young people who are willing and able to serve. There’s relatively little attention paid to the fact that the military is a “family business,” with 80% of recruits coming from extended military families. In 2023, however, the Wall Street Journal (in an article entitled, “The Military Recruiting Crisis: Even Veterans Don’t Want Their Families to Join”) argued that a major contributor to the recruitment crisis is that “disillusioned” former servicemembers are no longer advising their children to enlist. The military, in other words, is in danger of losing its “main pipeline for new servicemembers.”

The Wall Street Journal’s reporting presented no hard evidence that service recommendations from veterans have meaningfully declined over time. Instead, the article relied on a small number of interviews with disaffected former servicemembers to advance the claim that the recruitment crisis is being driven by increasing skepticism among veterans. According to these interviews, veterans are no longer advising their family members to join because the military has been forced to endure “20 years of war in Iraq and Afghanistan with no decisive victories, scandals over shoddy military housing and healthcare, poor pay for lower ranks that forces many military families to turn to food stamps, and rising rates of post-traumatic stress disorder and suicide.” In addition, the story suggested that service recommendations are declining among liberal veterans because “some on the left see the military as a redoubt of fringe conservatism.”

Surprisingly, the Journal’s reporting made almost no mention of the complaints frequently voiced by conservative veterans, an oversight pointed out by numerous commenters on the article. Three of the most “liked” posts in the comment section, for example, highlighted frustration with the military’s leadership during the Biden administration and its perceived focus on “progressive” social issues rather than warfighting:

“I served in Vietnam. Based upon what I've seen of Obama/Biden and now Biden's 3 years, no way would I serve or recommend anyone serve in today's military. As the SC ruling showed about Harvard and Biden's follow up comments diminished the SC and illustrates how much he hates the country and has diminished the quality of the military.”

“The focus on “progressive issues”, to put it lightly, has a far greater adverse impact on recruiting and retention, especially in the South, than the buried single sentence in this long piece would indicate. So long as the folks in charge fail to admit this and adjust accordingly, the situation will fester.”

“The article misses the point of what's really going on. I served, and was proud to do it because I believed in the what this country stood for, and I was willing to stand up and defend the Constitution. I believed that the leaders and politicians that commanded the military likewise had the same values. When I hear General Milley talk about climate change and white rage being the biggest threat to our security - I have no faith in our leaders. I would steer my own son away, because what's the point of doing the bidding of this bunch of fools.”

Comments such as these are typical of those found underneath online discussions of military issues. For example, Stand Together Against Racism and Radicalism in the Services (STARRS), which describes itself as a group of “US military veterans and citizens concerned about the divisive racist and radical CRT/DEI ideology infiltrating the military,” has assembled a document containing over 1,000 comments (published in response to news articles, online videos, and social media messages) from “currently in or retired from the military, or from a military family.” According to STARRS, the comments provide “overwhelming evidence that the CRT/DEI/Woke agenda being pushed in the Department of Defense DOES harm morale, recruiting and retention.”

The Wall Street Journal’s reporting and the analysis from STARRS advance competing claims about the root causes of veterans’ increasing reluctance to recommend military service to young people. Appropriately adjudicating between these claims requires far more, however, than a handful of interviews or an aggregation of quotes selectively culled from Internet comment sections. Nationally representative surveys of the veteran population (preferably gathered over time) would make it far easier to assess whether (and why) veteran service recommendations have fallen over the last five years.

While the Pentagon has made no effort to systematically track the country’s veteran population on these vital questions, a new survey - the 2024 Survey of Military Veterans - does. In this post, I draw on this nationally representative sample of 2,100 former servicemembers to determine how and why service recommendations have changed over time. As I show below, the willingness of veterans to advise young people close to them to join the military has declined dramatically since 2019. In other words, the premise of both the Journal’s and STARRS’s reporting is correct.

As I also point out, however, the “recommendation recession” is not a problem that has equally afflicted all parts of the veteran community. The overall decline in veteran service recommendations is almost entirely attributable to the diminishing enthusiasm for military service found among conservative veterans, who make up roughly half of the overall veteran population. In fact, service recommendations among liberal veterans have remained relatively constant since 2011.

Why has the decline been distributed so asymmetrically across the political spectrum? The data from the 2024 Survey of Military Veterans tells an unambiguous story whose broad outlines affirm the conclusions of the STARRS report. Recent Department of Defense policy changes and dissatisfaction with the direction of the military have drastically reduced the enthusiasm conservative veterans have for recommending service to their family members. For conservative veterans, mistrust of political and military leadership and the Department of Defense’s DEI initiatives are far more important factors in not advising their family members to enlist than “pay and benefits,” the “potential for war,” or the possibility of physical, psychological or emotional harm. There’s also little evidence that the military’s performance in Iraq and Afghanistan has deterred conservative vets from recommending service.

By contrast, the 20% of veterans who identify as politically liberal are relatively happy with the direction of the military. They express high levels of support for DEI initiatives and favor further efforts to diversify the force. This satisfaction has not significantly increased their overall willingness to recommend service relative to 2011 or 2019 but they view recent Department of Defense policy changes as largely positive developments.

Much like conservative veterans, very few liberal veterans cite the military’s performance in Iraq and Afghanistan as a reason to avoid making service recommendations. Liberal veterans are, however, far more sensitive to the harms associated with military service than conservative veterans. When asked why they are unwilling to advise young people to enlist, large numbers of liberal veterans cite concerns highlighted in the Journal’s reporting - PTSD, physical injury, death, and the potential for future war.

This polarization presents a profound recruitment challenge. Based on the data from the 2024 Survey of Military Veterans, eliminating DEI would help restore the trust of conservative veterans and increase sagging recruitment recommendations. Such a move would, however, would likely disillusion many liberal veterans who support these policies. While there is potential for remedial approaches that target the shared concerns of sexual harassment, budgetary cuts, and the country’s poor treatment of veterans, most paths forward risk alienating one segment of the veteran population in service of another. One of the primary challenges for the incoming Trump administration, then, will be to restore the trust of conservative veterans by refocusing the military on warfighting and lethality while not driving away large segments of the left-leaning veteran community.

Subscribe now

Previous Research on Adult Influencers and Military Recruitment

There are only three existing data sources that provide any insight into whether military service recommendations have changed over time. Unfortunately, the limitations of each of these data sources render them incapable of identifying the extent to which veteran service recommendations are declining and what factors might be driving such a decline.

The Department of Defense’s JAMRS Influencer Poll

The Department of Defense conducts “periodic polling of the influencer market” in order to “track attitudes, impressions and behavioral intentions as they relate to and affect military enlistment.” The most recent release from this “JAMRS Influencer" poll (July–September 2021) describes the sample of respondents as consisting of 924 “influencers,” including “233 Fathers, 258 Mothers, 221 Grandparents, 212 Other Influencers.”

Figure 1 - JAMRS “Influencer” Likelihood to Recommend Military Service

The JAMRS data shows very little movement in military service recommendations over time. As Figure 1 shows, there was no meaningful change in how willing adult “influencers” were to recommend military service between 2016 and 2021.

The JAMRS data is not particularly helpful in assessing the relationship between veteran service recommendations and the current recruitment crisis. The last release of the JAMRS results is now more than three years old, missing the worst years of the recruitment crisis. Although the publicly reported data is limited, JAMRS does not appear to ask respondents to explain the reasons why they might advise against a young person joining the military. What’s more, JAMRS does not sample the most influential “influencers”: military veterans. The JAMRS surveys, in other words, provide few useful insights into the issues that currently animate debates about the recruitment crisis.

Blue Star Families Military Family Lifestyle Survey

Since 2016, Blue Star Families (an organization founded with the goal of “empowering military families to thrive by connecting them with their civilian neighbors to create strong communities of support”), has conducted an annual “Military Family Lifestyle Survey” (MFLS) of active-duty military family members. The MFLS relies on a convenience sample of respondents obtained by distributing the survey “through Blue Star Families’ networks and partners in the military community.” As reported on the MFLS site, the survey “cannot be considered representative of the entire [active-duty military family member] population.” In 2023, the MFLS obtained responses from 7,431 active-duty military family members.

The MFLS has tracked the willingness of military family members to recommend military service over the last seven years. The questions used by the MFLS to assess this have evolved over time, focusing on a “young person close to you” between 2016 and 2020 and asking about “a young family members” in 2023.

Figure 2 – Blue Star Families MFLS Likelihood to Recommend Service                           

The data collected by the MFLS shows a significant decline in service recommendations over the last eight years. In 2016, the MFLS found 55% of active-duty family members would recommend service to a “young person close to you.” In 2023, the number of active-duty family members who would recommend service to “young family members” had fallen to only 32% (Figure 2).

The MFLS’s results are suggestive of growing skepticism about military service among those connected to the armed forces. Much like JAMRS, however, the MFLS is not a dedicated survey of veterans and it does not include the kinds of politically-focused questions that might help us evaluate the extent to which recent military policy and performance changes might be creating this skepticism.

Pew Research Center Survey of Veterans

The Pew Research Center has periodically conducted nationally representative surveys of the veteran population. In addition to a series of questions about political identity (e.g. partisanship, self-reported ideology), the Pew surveys ask “Would you advise a young person close to you to join the military?”

The Pew surveys are infrequent and badly outdated, with the most recent two iterations fielded in 2011 and 2019. The fact that these surveys precede the onset of the recruitment crisis, however, provides a useful baseline for measuring how enthusiastic veterans are about recommending service. If the recruitment crisis is driven, in part, by declining service recommendations from veterans, we should see clear evidence of it when comparing survey results from 2019 to survey results from today.

The 2024 Survey of Military Veterans

Fortunately, the 2024 Survey of Military Veterans, which was a large (N=2,100), nationally representative survey of veterans conducted by YouGov between August 19 and September 3, 2024, addresses the limitations of this previous research. The survey included a broad range of demographic and political identity questions that help paint a detailed picture of today’s veteran population.

Who are Today’s Military Veterans?

Before diving into the question of veteran service recommendations, it is important to point out that there are more veterans on the political right than on the political left, with 43% identifying as conservative and 52% expressing an intention to vote for Donald Trump in 2024. This makes the veteran population more right-leaning than the American public on the whole.

It is also important to point out, however, that the 2024 Survey of Military Veterans sample probably skews liberal. Consider data from the 2022 and 2023 Cooperative Election Study (CES) surveys. The CES surveys have massive, nationally representative samples (e.g. 60,000 respondents in 2022) that return huge numbers of veteran respondents (e.g. more than 6,000 former servicemembers in 2022).

Figure 3 - Self-Reported Political Ideology among Military Veterans

In both 2022 and 2023, the CES found that half of veterans were conservative (and only 17% were liberal). As Figure 3 shows, these results are slightly out of step with the estimates from the 2024 Survey of Military Veterans, which finds 43% of veterans are conservative and 24% are liberal. It is possible that 7% of the veteran population shifted from conservative to liberal (either through population replacement or attitude change) between 2023 and 2024 but it is more likely that the 2024 Survey of Military Veterans is slightly biased in a left-leaning direction.

Figure 4 - Presidential Voting Preferences among Military Veterans

The same left-leaning bias can be seen in presidential voting preferences (Figure 4). The 2024 Survey of Military Veterans was conducted in August 2024 and found that 52% of veterans intended to vote for Donald Trump in the presidential election. Data from an August 2024 Pew Research Center survey of veterans, however, found 61% of veterans were going to vote for Trump. Further, exit polls from the election found that 65% of veterans ended up voting for Trump.

All of this is to say that the results presented below probably underreport the number of conservative veterans and, in doing so, underestimate the scope of the overall recruitment problems posed by conservative dissatisfaction with the military.

Service Recommendations among America’s Veterans

As mentioned above, the 2024 Survey of Military Veterans included a number of questions related to the state of the American military. Most importantly, following in the footsteps of the MFLS and the Pew veteran surveys, respondents in the 2024 Survey of Military Veterans were asked whether they would “advise a young family member (child, niece, nephew, etc.) to join the military today?”

As Figure 5 shows, there was no movement at all in veteran service recommendations between 2011 and 2019, with approximately 80% of veterans saying they would advise service in both years. Since 2019, however, there has been a dramatic change in how veterans feel about advising young people to join the military. Specifically, the overall percentage of veterans saying they would NOT recommend military service has nearly doubled over the last five years, growing from only 20% in 2019 to 38% in 2024.

Figure 5 - Service Recommendations by Military Veterans Since 2011

The topline results, however, hide significant political variation in how veterans feel about military service. As Figure 6 shows, growing skepticism about military service is heavily concentrated among the 40%-50% of veterans who identify as politically conservative. In 2019, 88% of conservative veterans said they would advise a young person close to them to join the military and 12% said they would not. In 2024, those numbers were 53% and 47%, respectively. This steep decline stands in stark contrast to trends among liberal veterans, whose collective willingness to recommend service was 59% in 2019 and 64% in 2024.

Figure 6 - Veteran Service Recommendations by Political Ideology Since 2011

Veteran Views on Recent Military Developments

So why have conservative veterans (but not liberal veterans) become so reluctant to recommend service to their young family members? The 2024 Survey of Military Veterans helps us answer this question by measuring how veterans feel about the military’s recent policy initiatives and performance.

To be more precise, a large number of questions included in the 2024 Survey of Military Veterans make explicit reference to military controversies from the last four years. In 2017, for example, Secretary of Defense Mattis declared that lethality was the military’s overriding focus. Signaling a new approach built on the assumption that diversity is “a necessary part of warfighting and all successful military operations,” President Biden issued an executive order “embedding diversity, equity, inclusion, and accessibility (DEIA)” efforts in “all parts of the Federal workforce,” including the military.” “Diversity officers” were summarily installed throughout the ranks, systematically replacing Colin Powell’s “color blind” philosophy with racial and gender reporting up the chain-of-command. Reflecting these new priorities, Joint Chiefs of Staff Chairman C.Q. Brown issued a memorandum in 2022 setting race and sex quotas as “Diversity & Inclusion goals” for commissioned officer applicants to the Air Force.

The data reveal just how polarizing these decisions have been among veterans. As Figure 7 shows, 64% of conservative veterans (but only 15% of liberal veterans) say that the military has become “less focused on warfighting and lethality” over time.

Figure 7 - Military’s Focus on Warfighting and Lethality by Ideology

In direct opposition to President Biden’s statements, 87% of conservative veterans say DEI efforts are “not essential parts of warfighting and are not necessary for achieving success in military operations” (Figure 8). By contrast, 81% of liberal veterans say that DEI efforts are essential for military success.

Figure 8 - DEI Efforts And Military Success by Ideology

Similarly, 80% of conservative veterans say that the military pays “too much” attention to increasing diversity, equity, and inclusion. Only 8% of liberal veterans said the same (Figure 9).

Figure 9 - Attention Paid to DEI by Ideology

Opposition to racial and gender preferences was also widespread among conservative veterans, with 98% saying conservative veterans “race and gender should not be a factor” in military promotion decisions (Figure 10). Liberal veterans, by contrast, were more supportive eight times more likely than conservative veterans to say “there should be racial and gender preferences.”

Figure 10 - Racial and Gender Preferences in Military Promotions by Ideology

Collectively, these views have fueled a broad pessimism about the military among conservative veterans. As Figure 11 shows, while only 25% of liberal veterans and 44% of moderate veterans say the military is headed in the wrong direction, nearly 90% of conservative veterans do.

Figure 11 - Military Headed in the Right/Wrong Direction by Ideology

Unsurprisingly, these perceptions are highly correlated with the willingness to make service recommendations. Only 51% of those who see the military headed in the “wrong direction” would advise a young family member to enlist (compared to the 78% of those who see the military headed in the “right direction.”

Impact of Recent Developments on the Likelihood of Service Recommendations

While these results are suggestive, the 2024 Survey of Military Veterans provides more direct evidence about how important DEI initiatives are relative to other, more longstanding concerns. Specifically, the survey asked two sets of questions connecting the complaints identified in the Wall Street Journal and the STARRS report to service recommendations.

First, the survey asked respondents whether seven recent developments connected to the US military have made it more or less likely that they would recommend service to a young family member. Figures 12, 13, and 14 show the net impact of each development on the likelihood of service recommendations (i.e. the percentage of respondents who said the development made them “much more” or “somewhat more” likely to recommend service minus the percentage of respondents who said the development made them “much less” or “somewhat less” likely to recommend service).

Figure 12 - Impact of Recent Developments on the Likelihood of Service Recommendations among Conservative Veterans

Figure 13 - Impact of Recent Developments on the Likelihood of Service Recommendations among Moderate Veterans

Figure 14 - Impact of Recent Developments on the Likelihood of Service Recommendations among Liberal Veterans

Figures 12-14 show the incredibly positive impact the Post-9/11 GI Bill and reduction in deployment time have had on veterans’ willingness to recommend service to their children (regardless of political ideology).

The data also show the incredibly negative impact reports of sexual assault have had on service recommendations, with large percentages of veterans across the political spectrum saying these have made them less likely to advise young people to join the military. Similarly, the possibility of budget cuts have reduced the likelihood of service recommendations from a significant number of veterans of all political stripes.

Contrary to conventional wisdom, the “performance of the military in Iraq and Afghanistan” does not appear to be a net drag on service recommendations. 25% of veterans said the military’s performance in these conflicts made them less likely to recommend service but 28% said it made them more likely to recommend service.

The most striking finding to emerge from the data presented in Figures 12-14 is the polarizing nature of the military’s DEI efforts. 76% of conservative veterans said “the increasing focus on DEI” made them “somewhat” or “much” less likely to recommend service to a young person close to them (10% said it made the “somewhat” or “much” more likely to and 14% said it had “no effect”). This was the most important factor in reducing the likelihood of service recommendations among conservative veterans. Among liberal veterans, however, DEI efforts have had a net positive effect on the likelihood of service recommendations. While only 9% of liberal veterans said “the increasing focus on DEI” made them “somewhat” or “much” less likely to recommend service, 51% said it made them “somewhat” or “much” more likely to. DEI initiatives, in other words, polarize the veteran population along ideological lines when it comes to service recommendations.

Veteran Reasons for Not Advising Military Service

Second, the 2024 Survey of Military Veterans asked those who said they would not recommend service to explain the considerations that went into their decision. Specifically, respondents were asked to rate how important each item on a list of eight factors was (“a major factor,” “a minor factor” or “not a factor”) in their choice to not recommend service.

Figure 15 - Reasons for Not Advising Military Service among Conservatives

Figure 16 - Reasons for Not Advising Military Service among Moderates

As Figure 15 shows, dissatisfaction with the country’s political leadership and the military’s DEI policies were “major factors” discouraging conservative veterans from recommending service.  More than 85% of conservative veterans who would not advise military service said “mistrust of political leaders” and “the military’s current DEI and other social policies” were “major factors” in their decision. Poor treatment of veterans” was also identified frequently, being mentioned as a “major factor” for not advising service by 65% of conservative veterans. The same conclusions hold among moderate veterans (Figure 16).

Contrary to the reporting of the Wall Street Journal, pay and benefits, the potential to go to war, time away from family, and the possibility of psychological problems, physical injury or death were far less important, with less than a third of conservative veterans saying each of these considerations was a “major factor” in their unwillingness to advise service.

Figure 17 - Reasons for Not Advising Military Service among Liberals

Similar to conservative veterans, nearly 60% of liberal veterans mentioned “poor treatment of veterans.” As Figure 17 also shows, much like conservative veterans, a relatively small percentage of liberal veterans say “pay and benefits” are a “major factor” for not recommending service. It does not appear, in other words, that the military’s compensation for soldiers is acting as a significant deterrent to veteran service recommendations.

Despite these similarities, there were important differences between liberal and conservative veterans on the importance of DEI and harm-based considerations. Unsurprisingly, liberal veterans were far less likely than conservative veterans (17% to 86%) to say that the military’s “DEI and social policies” were “major factors” in their unwillingness to recommend service. More surprisingly, however, liberal veterans were roughly twice as likely as conservative veterans to say the potential to go to war (66% to 31%), the possibility of psychological problems (69% to 28%), and the possibility of physical injury or death (65% to 34%) are “major factors” for not recommending service. Liberal veterans were also significantly more likely than moderate veterans to list these concerns as “major factors.” Put differently, the Wall Street Journal’s diagnosis is a much better description of what’s going on for liberal veterans than what is going on for conservative or moderate veterans.

What is to be Done?

Some observers have recently argued that “veterans are the key to solving the US military’s recruitment crisis.” While veteran service recommendations have declined significantly over the last five years (coinciding with the onset and acceleration of the military’s enlistment problems), it has not been entirely clear what can be done to reverse this decline. There simply hasn’t been enough good data to identify a path forward.

The 2024 Survey of Military Veterans provides a clear picture of the scope and causes of the “recommendation recession.” Unfortunately, as the results presented above show, there are no silver bullets that can solve this problem.

The backdrop of the decline in service recommendations among veterans is a deep mistrust of the country’s political leadership. After watching four presidents lose two wars, buffeted by polarizing policy changes from one administration to another, veterans are no longer confident that their children and grandchildren will enjoy proper leadership. Indeed, 91% of conservative veterans, 83% of moderate veterans and 63% of liberal veterans say that “distrust of political leaders” is a major factor for not advising service.

Perhaps the easiest approaches to restoring trust and reinvigorating veteran service recommendations are the ones that receive widespread support across the political spectrum. Specifically, the 2024 Survey of Military Veterans points to five things the country’s political and military leadership can do to increase the willingness of veterans to advise young people close to them to enlist.

First, efforts must be made to improve the treatment of the country’s current veteran population. The perceived poor treatment of veterans was a “major factor” why many former servicemembers (independent of political ideology) would not advise a young person close to them to join the military. In short, taking care of today’s veteran population is both a moral and strategic imperative for the military.

Second, there must be stronger advocacy for defense spending. More than 30% of liberal, conservative, and moderate veterans said that the possibility of budget cuts made them less likely to advise a young person close to them to join the military. Protecting defense spending can help address the recruitment crisis by increasing veterans’ comfort with advising service to their young family members.

Third, the military must take steps to root out sexual harassment. Half of veterans said that reports of sexual harassment in the military make them hesitant to recommend military service. These concerns come up among liberal and conservative veterans alike. If the military would like to address its service recommendation problem, it must begin by addressing its sexual harassment problem.

Finally, the military should continue to expand its benefits and adjust its deployment schedule. The Post-9/11 GI Bill and the military’s decision to increase the deployment-to-dwell ratio made veterans of all political stripes significantly more likely to recommend military service. Emphasizing these benefits and pursuing similar policies should provide a much needed boost to veteran service recommendations.

Beyond these steps, however, things become tricky. Nearly any other approach to increasing veteran service recommendations will have to begin with the recognition that the veteran population is deeply polarized and addressing the concerns of veterans from one end of the political spectrum might alienate veterans from the other end.

Most notably, the results of the 2024 Survey of Military Veterans suggest that ending DEI initiatives and explicitly promoting a more meritocratic approach centered on the prioritization of warfighting and lethality would go a long way to increasing the willingness of moderate and (especially) conservative veterans to recommend military service to their young family members. The opinions of conservative veterans should not, of course, be the only consideration when deciding the future of DEI in the military. Whether DEI initiatives increase or decrease readiness, unit cohesion and the willingness of young people to enlist must also be considered. Nevertheless, the data presented above suggest that rolling back some of the military’s recent DEI efforts will increase many veterans’ enthusiasm for recommending military service.

There’s a legitimate concern that limiting the military’s DEI efforts may drive down service recommendations from liberal veterans. As the data presented above show, many DEI efforts are popular with a majority of liberal veterans. It is also important to point out, however, that the military’s embrace of DEI over the last five years has not significantly increased service recommendations from liberal veterans. The percentage of liberal veterans who would recommend service was 59% in 2019 and 64% in 2024. It does not appear, in other words, that DEI initiatives have offset conservative losses with liberal gains.

Veteran recommendations are only one part of the recruitment crisis. While the incoming Trump administration must try to restore the military’s “main pipeline” for new recruits by boosting service recommendations among veterans, it must also be aware that a focus on veterans cannot come at the expense of the interests of young people who are not part of a military family. Whether it makes sense to eliminate DEI initiatives, for example, will have to be determined by examining the net effect of those initiatives among both veterans and potential recruits. Solving the multi-faceted recruitment crisis will almost certainly require multi-faceted and narrowly targeted solutions.

There were many analyses I could not include in my Heterodox Academy blog post. Most importantly, I was unable to show that universities with larger DEI bureaucracies have more student self-censorship, more student discomfort with discussing controversial ideas, and more skepticism about the administration’s commitment to speech.

Below, I present this evidence and the entirety of the slides from my presentation:

Between the publishing of Lukianoff and Haidt’s Coddling of the American Mind in 2015 and the summer of 2020 there was a lively debate on the question whether there was a “free speech crisis” on college campuses. While some argued that “cancel culture” was destroying free expression on campus, others dismissed concerns about college free speech as a “moral panic.”

By 2023, however, most observers were willing to admit that the “campus free speech crisis” was not just a “myth.” As Voice of America put it, “Everyone Agrees Campus ‘Free Speech’ is in Crisis.”

But why was there a “crisis” of free speech on college campuses? The most popular arguments (e.g. Lukianoff and Haidt) centered on the idea that Gen Z’s unique political, cultural, emotional and/or psychological characteristics made them intolerant of speech.

But these accounts cannot help us identify why the “crisis” of campus free speech is not evenly distributed across places. While some universities find themselves in the news constantly for controversies connected to speech, others do not.

The cross-campus variation in campus free speech environments has led to a focus on factors internal to the university. Most notably, campus observers have started linking DEI efforts on campus to the ever-worsening “crisis” of college free speech. Consider Abbot and Marinovic in Newsweek:

Lawrence Krauss struck a similar tone in the Wall Street Journal:

More recently, high-profile conflicts over campus speech have involved DEI administrators as central players. Federal Judge Kyle Duncan’s speech at Stanford Law School was quite literally blocked by DEI (in the form of a DEI director):

Additionally, Claudine Gay’s response to speech conflicts at Harvard was immediately linked to the DEI bureaucracy:

The assumed link between the free speech crisis and the rise of DEI is curious. From a naïve perspective, what could be so wrong about institutional support for “diversity, equity and inclusion”? The mystery only deepens when one examines university DEI department websites. Consider the University of Michigan’s DEI page. After reading this, it would not be unreasonable to expect that the increasing size and significance of university DEI bureaucracies might significantly improve the speech climates on college campuses.

So we are left with an empirical question: are larger DEI bureaucracies really associated with worse campus speech climates?

The first challenge in answering this question is to measure DEI across campuses. In order to do this, I follow the approach spelled out in Greene and Paul’s 2021 paper, “Diversity University: DEI Bloat in the Academy.”

Greene and Paul measured the size of DEI bureaucracies at 65 “Power 5” universities:

Using keyword searches, Greene and Paul were able to identify how many employees are devoted to DEI efforts across each of their “Power 5” universities:

Greene and Paul find that universities employ an average of 45 DEI employees. In their words, “Promoting DEI has become a primary function of universities.”

Greene and Paul’s sample excludes Ivy League universities, liberal arts colleges and many other high-profile schools. In order to address this oversight, I extended their analysis to include an additional 44 schools. Specifically:

Perhaps unsurprisingly, the size of DEI bureaucracies varies tremendously across schools. Liberal arts colleges tend to have the largest DEI bureaucracies while lower ranked public universities tend to have the smallest DEI bureaucracies.

But can variation in the size of university DEI bureaucracies explain variation in campus speech environments? In order to answer this question, we need to measure campus speech climates. FIRE’s Campus Rankings Survey allows us to do this:

Assessing the impact of DEI on these measures, however, requires an analysis that can control for other factors that might impact campus speech. These factors can be individual-level or institutional-level. The list of variables I chose to include are listed here:

A multi-level modeling approach allows us to identify the impact of individual and institutional level factors:

The results of this analysis are summarized here (where “+” indicate a statistically significant and positive effect, “-” indicates a statistically significant and negative effect and “blanks” indicate results that were not statistically significant). Remember, the multilevel model also included controls for individual-level influences on speech attitudes.

To illustrate these findings, I present the following graphs (which show the predicted impact of increasing the number of university DEI personnel while holding all other variables constant at their means):

Overall, then, we can conclude that larger DEI bureaucracies are associated with worse campus speech climates. More employees with DEI responsibilities means:

• More self-censorship

• More discomfort

• More support for disruptive action

• More skepticism of the administration’s commitment to speech

• More intolerance of conservative speakers

There are important caveats, however:

• Effects are relatively small

• Eliminating DEI would not solve the free speech crisis

• Says nothing about the wisdom or effectiveness of DEI on other dimensions

Additionally, there is little reason to believe that “greening” speech codes or endorsing statements like the “Chicago Statement” would improve things. These variables are not consistently related to better campus speech climates.

In this post, I show that very few American high school students currently believe the military, higher education, the media, religious organizations, law enforcement, big business, public schools, or labor unions are serving the country well. But, as I also show, not all high schoolers are equally negative about America’s institutions. Young Democrats and young independents are significantly less likely to say that institutions are doing a “good” or “very good” job for “the country as a whole” than young Republicans.

Things were not always this way. Twenty years ago, significant majorities of high schoolers across the political spectrum thought highly of all of the country’s major institutions. The number of young Democrats with a positive view of American institutions has collapsed during the last two decades, however. Since 2002, for example, there has been a 44-point drop in positive assessments of the military, a 42-point drop in positive assessments of colleges and universities, and a 36-point drop in positive assessments of the news media among young Democrats. Smaller, but important declines, have occurred among young independents and Republicans.

It’s easy to be dismissive of the attitudes of high school students. But ignoring young people’s growing cynicism about our key institutions would be a serious mistake. Nearly every American institution that relies on the voluntary participation of young people now finds itself facing an existential crisis. Colleges and universities, for instance, are on the verge of falling off an “enrollment cliff.” Religious organizations are experiencing a “great dechurching.” The US military is fighting an unprecedented “recruitment crisis.”

Unsurprisingly, the emergence of these crises has coincided with young people’s worsening perceptions of the armed forces, higher education, and religious organizations. As I show below, students who evaluate the military and higher education poorly are far less likely to express an interest in joining these institutions after graduation. Similarly, students who say churches are not serving the “country as a whole” are far less likely to attend church services. If American institutions hope to sustain themselves, they must find ways to rehabilitate their rapidly deteriorating images with young people in general and young liberals in particular.

Kids these days…

Monitoring the Future (MTF), a long-running, nationally representative survey of thousands of American 12^th graders, is the single best data source for evaluating how young people feel about the country’s institutions. Each year, the MTF asks a subset of their sample questions about the performance of eight key “organizations” in American society: the US military, colleges and universities, the national news media, churches and religious organizations, police and law enforcement, large corporations, public schools and labor unions. Specifically, the survey presents high school seniors with the following question: “Now we'd like you to make some ratings of how good or bad a job you feel each of the following organizations is doing for the country as a whole.” Respondents are allowed to rate the institution as doing a “very poor,” “poor,” “fair,” “good” or “very good” job for the “country as a whole.” The most recent available data comes from the 2022 iteration of the MTF.

Do today’s high school seniors think the country’s most important institutions are doing a “good job” for “the country as a whole”? The short answer is: no.

Figure 1 – Positive Evaluations of Institutions among High Schoolers in 2022

As Figure 1 shows, most high schoolers do not rate the key institutions of American society positively. In fact, the US military is the only institution included in the MTF that a majority of high school students say is doing a “good” or “very good” job for the “country as a whole.” Interestingly, young people are harshest in their assessments of the media, with only 18.8% of high school seniors rating the performance of national news outlets positively.

Figure 2 – Positive Evaluations of Institutions in 2022 by Partisanship

Figure 2 reveals that negative performance assessments are not evenly (or predictably) distributed across the political spectrum. Four things in particular jump out from the data.

First, young Republicans currently provide the most positive evaluations of colleges and universities. Surveys of adults consistently show that Republicans are more negative toward colleges and universities than Democrats or independents. Among high school students, however, Republicans are the least critical of higher education. In 2022, young Republicans were approximately 20% more likely than young independents and approximately 15% more likely than young Democrats to say that colleges and universities are doing a “good” or “very good” job.

Second, young Republicans are not more negative than young Democrats or young independents about the media. Survey after survey has found that adult Republicans are now far less trusting of news organizations than anyone else. Yet, in the 2022 MTF, young Republicans (16.4%) were just as likely as young Democrats (15.1%) and much more likely than young independents (8.8%) to say that the national news media is doing a “good” or “very good” job.

Third, labor unions now receive their most positive evaluations from young Republicans. Historically, organized labor has been a large and loyal part of the Democratic Party’s electoral coalition. Indeed, the recent speech by Teamsters President Sean O’Brien at the Republican National Convention was noteworthy because Republicans have, historically, been antagonistic toward labor unions. The long-standing and deep ties between organized labor and the Democratic Party might lead us to believe that young Democrats would evaluate unions more positively than young Republicans. In the MTF, however, the opposite is true; young Republicans rate labor unions far more positively than young Democrats (47.4% “good” or “very good” job ratings to only 28.1%).

Finally, there is a massive partisan gap among young people in evaluations of law enforcement. While we might expect that young Republicans would be more positive in their assessments of the police than young Democrats, the distance between the parties here is striking. As Figure 2 shows, young Republicans were nearly eight times more likely than young Democrats (54.6% to 7.2%) to say that the police are doing a “good” or “very good” job.

Overall, the data in Figure 2 make it clear that young Democrats and young independents are far more critical of American institutions than young Republicans. On average, across the eight institutions included in the MTF, only 22.4% of Democrats gave “good” or “very good” ratings. In fact, there is not a single institution that received positive performance evaluations from even 40% of young Democrats.

Things are only marginally better among young independents. On average, 27.1% of young independents evaluated the eight institutions positively. With the lone exception of the US military (60.5% “good” or “very good” ratings), a large majority of young independents rated all of the institutions asked about in the MTF negatively.

Young Republicans stand out for their relatively rosy assessments of the eight institutions. Majorities of young Republicans rate the performance of the military (78.5%), churches and religious organizations (62.7%), police and law enforcement (54.6%) and, shockingly, colleges and universities (50.1%) as “good” or “very good.”

Was it always this way?

Were young people (particularly young Democrats) always so negative about the country’s most important institutions? The MTF’s archive of survey data makes it easy to answer this question because it has asked the same series of questions about institutional performance over the last 20 years.

Figures 3, 4, and 5 show the yearly percentage within each party providing a positive evaluation of each institution’s performance. Figure 6 summarizes the overall changes during the last two decades by showing the difference between assessments in 2002 and 2022.

Figure 3 – Positive Evaluations of the Military, Colleges, and Churches

Figure 4 – Positive Evaluations of Media, Police, and Public Schools

Figure 5 – Positive Evaluations of Large Corporations and Labor Unions

Figure 6 – Change in Positive Evaluations of Institutions by Partisanship (2002-2022)

The trends illustrated in Figures 3, 4, 5, and 6 are staggering. Young people (independent of partisanship) have become dramatically more critical of the country’s major institutions since 2002. As Figure 6 shows, there is not a single institution that today’s independent and Democratic-identifying high school students evaluate more positively than their counterparts in 2002 did. Today’s Republican high schoolers, for their part, are also now providing significantly worse ratings of most institutions than Republican high schoolers from twenty years ago. In other words, things were not always this way and there is nothing about being an American teenager that inevitably makes one skeptical about the country’s major social institutions.

Although all young people have become more negative about the country’s institutions, this generational “crisis of confidence” is concentrated disproportionately among high schoolers on the political left. As Figure 6 shows, young Democrats have become far more negative over the last 20 years than young independents or young Republicans. There have been more than 30-point declines in positive evaluations among young Democrats for six of the eight institutions queried about in the MTF, including a 44-point drop in assessments of the military and a 42-point drop in assessments of colleges and universities. By contrast, there have been 30-point declines among young independents for only three institutions and no 30-point declines among young Republicans. Any attempt to address young people’s growing cynicism about institutions will have to start by recognizing this partisan asymmetry.

Who Cares?

As mentioned above, major American institutions that rely on the participation of young people find themselves in a state of “crisis.” Are these “crises” a result of the fact that so few young people now think highly of American institutions?

The MTF allows us to take an initial stab at answering this question. In addition to the items measuring performance evaluations, the MTF asks high schoolers questions about some of their behaviors and career aspirations. Three of these questions are particularly relevant for exploring whether deteriorating perceptions of institutional performance have consequences for the continued success of American institutions.

The United States Military

First, the MTF asks high school seniors whether they would want to “serve in the armed forces” if they could do just what “they liked” and “nothing stood in their way.” In a previous post (“The Military’s White Democrat Problem”), I used this data to reveal a growing partisan divide in the desire for military service over the last decade, with white Democrats becoming far less interested in joining the armed forces since 2015. Figure 7 expands and extends that analysis by including all young Democrats between 2002 and 2022.

Figure 7 - Desire to Serve in the Armed Forces between 2002 and 2022 by Partisanship

Eyeballing Figure 7 alongside the left-hand panel of Figure 3 suggests that declining performance evaluations among young Democrats may be driving their declining willingness to serve in the military, with both trendlines taking a nosedive after 2013. Additionally, Republican high schoolers have become simultaneously less positive toward the military and less interested in serving since 2019.

The best way to examine the relationship between performance assessments and behavioral intentions at the moment, however, is to model it with the appropriate statistical controls at the individual level using the most recent data. Drawing on the 2022 MTF, I ran a logistic regression analysis predicting the probability of wanting to serve in the military based on performance assessments of the military (controlling for race, gender, partisanship, parental education, and religiosity). Figure 8 shows the predicted probability of wanting to serve in the armed forces across military performance scores.

Figure 8 – Predicted Probability of Wanting to Serve in the Military by Military Performance Assessments

As Figure 8 shows, performance evaluations exert a powerful influence on a high schooler’s interest in serving in the military (net of race, gender, partisanship, parental education, and religiosity). A student who evaluated the military’s performance as “very poor” is predicted to have less than a 1% chance of wanting to serve in the military. By contrast, a student who evaluated the military’s performance as “very good” is predicted to have 17% chance of wanting to serve in the military. Importantly, the values in Figure 8 show the effect of performance evaluations after controlling for partisanship (i.e. the relationship is not simply a consequence of the fact that Democrats are less interested in serving and less likely to rate the military positively). In other words, the increasingly negative perceptions of the military are a significant drag on military recruitment.

Colleges and Universities

The MTF also asks high school seniors whether they would want to “graduate from college” if they could do just what “they liked” and “nothing stood in their way.” In my last post (“The Political Causes of the College Enrollment Crisis”), I used responses to this question over the last twenty years to show an emerging partisan gap in the desire for a college education. Figure 9 shows the differences in educational aspirations between high school Republicans and Democrats since 2002:

Figure 9 - Desire to Graduate from College between 2002 and 2022 by Partisanship

Comparing Figure 9 to the middle panel of Figure 3 suggests that performance evaluations of colleges and universities and desire to attend college have had a complicated relationship to each other over the last 20 years. The percentage of high schoolers (regardless of partisan affiliation) saying colleges and universities do a “good” or “very good” job has declined in a relatively linear fashion since 2002. It is not until 2012, however, that young Republican disinterest begins to take off. What’s more, interest in college has not budged among young Democrats over the last two decades despite their massive increase in cynicism about higher education.

Where are things now? Following the same approach used for my analysis of the military, I ran a logistic regression model predicting the probability of wanting to attend college based on performance assessments of colleges and universities (controlling for all other theoretically important variables) in the 2022 MTF.

Figure 10 - Predicted Probability of Wanting to Graduate from College by College and University Performance Assessments

Figure 10 shows that performance evaluations of higher education exert an independent influence on a high schooler’s interest in graduating from college. A student’s chances of wanting to attend college are, ceteris paribus, predicted to increase by more than 20% if they have a “very good” view of institutions of higher education rather than a “very poor” view of institutions of higher education. If colleges and universities are interested in bolstering their flagging enrollments, they must take steps to reverse the trends shown in Figure 3.

Churches and Religious Organizations

Finally, the MTF asks high schoolers how often they attend church (“never,” “rarely,” “once or twice a month,” “once a week or more”). On his consistently excellent Substack, Ryan Burge has used the MTF to examine the evolving correlation between church attendance and political ideology. In “How Has Religion Changed Among High School Seniors?,” Burge shows the following change in religious participation between 1995 and 2022:

Figure 11 - Religious Attendance between 1995 and 2022

The fact that Democrats have become more critical of churches (left-hand panel of Figure 3) while also becoming less likely to attend church (Figure 11) suggests a strong connection between the two. To explore the relationship at an individual level in 2022, I ran an OLS regression predicting frequency of church attendance based on performance assessments of “churches and religious organizations.” Figure 12 shows the predicted level of church attendance across church performance scores after controlling for race, gender, partisanship, and parental education levels.

Figure 12 – Predicted Frequency of Church Attendance by Church Performance Assessments

As Figure 12 shows, there is a predictably strong relationship between evaluations of churches and church attendance, with those rating the performance of churches “very poor” predicted to attend church significantly less than those rating the performance of churches “very good.”

It would obviously be a mistake to conclude from this analysis that poor performance assessments cause less church attendance given the potential for more frequent church attendance to produce more positive performance assessments of churches. It is probably safe to say, however, that these two variables self-reinforcing. Absent any other good options, churches may want to try addressing their attendance problems by persuading young people that they are doing a “good job” for the “country as a whole” (even if, empirically speaking, the causal evidence to recommend this approach is weak).

What is to be done?

Youth dissatisfaction is a river fed by many tributaries. The onset of COVID, for instance, had a profoundly negative impact on how high schoolers think about all of the country’s institutions (whether they were involved in managing the pandemic or not). Between 2020 and 2021, positive evaluations for every one of the eight institutions included in the MTF declined among every demographic group of American high schoolers.

It is also clear that major political events have driven down performance evaluations for some institutions. Consider assessments of the police among young Democrats. In 2014, Michael Brown, Eric Garner and Tamil Rice were killed by police, sparking the Black Lives Matter movement. Between 2014 and 2015, Democratic positive assessments of law enforcement in the MTF were cut in half, with the percentage of young Democrats saying the police were doing a “good” or “very good” job falling from 41.3% to 21.8%. Positive evaluation evaluations of the police among young Democrats were halved once again between 2020 and 2021 in response to the death of George Floyd, dropping from 15.7% to only 5.5%.

Not all of the declines are easily attributable to such events, however. For example, Democratic perceptions of the military worsened considerably after 2014 without any obvious changes to the military’s performance. Similarly, perceptions of higher education and media have been in steady decline since 2002 among high schoolers of all political stripes.

These trends point to larger, structural causes feeding young people’s increasing disaffection with major institutions. Today’s youth are growing up in an information environment that incentivizes news that is “one-sided, critical, partisan, and reaction-generating.” When coupled with rising levels of political polarization, this media ecosystem encourages “Americans to adopt as second nature a more critical, if not cynical, view of systems and structures.”

All of this is to say that there is no one cause for the trends described above. Unfortunately, this also means that there is no panacea for the “crisis of confidence” among today’s young people. To make matters worse, it’s not obvious how institutions can effectively address the rapidly worsening perceptions of performance among young Democrats without also losing the support of young Republicans. Attempts to appeal to progressive values through controversial and ineffective “woke” DEI policies have almost always backfired in recent years, leading to widespread conservative frustration with the organizations that pursue them. There’s also not much evidence that further moves in a progressive direction will do much to reverse the trends among young liberals. High school Democrats, for example, have evaluated colleges and universities worse as colleges and universities have become more “woke.” The US military’s embrace of DEI has been similarly unsuccessful in winning over young Democrats.

Young people are unwilling to participate in institutions they view as performing poorly. Unless our key institutions can find a way to persuade young people from across the political spectrum that they are doing a “good job” for “the country as a whole,” the future looks grim.

Low birth rates are part of the story behind this “enrollment cliff.” But, they are only part. Just as important is the growing disinterest that many high school students have in graduating from college. According to data from Monitoring the Future, a long-running survey of American 12^th graders, the percentage of high school seniors reporting no interest in graduating from a four-year college rose from 18% to 28% between 2011 and 2022 (Figure 1).

Figure 1 – Increasing Disinterest in College

So why are high school students suddenly disinterested in college? The most commonly cited culprits are non-college career opportunities, costs and COVID. Yet, these explanations ignore two critical facts about growing disinterest in pursuing higher education: nearly all of the increase has happened since 2010 and nearly all of the increase has happened among Republican high school students.

The figure below shows results from Monitoring the Future’s survey question on college disinterest between 2002 and 2022 (the most recently released data) disaggregated by gender and partisanship. While Democratic boys and girls have expressed roughly the same desire to graduate from college since 2002, Republican high school students have significantly less interest in higher education than they did twenty years ago. What’s more, the disaffection of conservative high schoolers is a relatively recent occurrence, with the percentage of Republican boys uninterested in college increasing 16% and the percentage of Republican girls uninterested in college increasing 9% since 2010 (Figure 2). Disinterest among Independents (not shown) is also up among boys (20% to 26%) and girls (14% to 18%) since 2010.

Figure 2 – Increasing Disinterest by Gender and Ideology

Economics and pandemics alone cannot explain the timing or ideological asymmetry of these preferences. The central question, then, is why have Republican high school students become so disinterested in higher education since 2010?

One possibility is that a political realignment has led high schoolers with no interest in attending college to increasingly identify as Republicans instead of Democrats or Independents. There’s no perfect way to explore this possibility with the data provided by Monitoring the Future but it’s worth noting that there’s not any evidence that Republican high schoolers are now coming from families with an aversion to higher education. The percentage of Republican high school students with two college educated parents has increased by 12% since 2002 and by 9% since 2010.

Figure 3 – Parental Education among Republican High School Students

The more likely explanation is that the rising disinterest among Republican high schoolers is a function of the fact that, over the last 15 years, universities have become more ideologically homogenous, more overtly activist, and more monomaniacally focused on power and identity. These changes, and the extensively covered political opposition mobilized in response, have eroded trust in higher education among those on the political right and decreased young conservatives’ desire to attend college.

Consider the growing ideological homogeneity of universities. In 1995, the liberal-conservative ratio among faculty members was roughly 2 to 1. By 2010, the ratio was 5 to 1.

Figure 4 - Ideology among Faculty and the General Public (1989-2014)

The most recent data suggests even less political diversity within the professoriate. According to a 2018 survey of liberal arts faculty, for example, the Democrat to Republican party registration ratio was nearly 13 to 1. In 2023, the liberal to conservative ratio among Harvard faculty was 26:1. Nor is the political homogeneity confined to faculty. More than 70% of college administrators and more than 60% of students now identify as liberal.

Figure 5 - Self-Reported Ideology among Harvard Faculty Members

It’s not merely the lopsided distribution of political preferences in American universities that may have alienated conservatives, however. It is the increasingly activist nature of campus progressivism. According to a 2020 survey, for example, approximately 40% of social science and humanities faculty members agreed with the statement “I would consider myself as an activist.” As Derek Thompson recently wrote in the Atlantic, “in the past few years, many elite colleges and universities have cast themselves as “anti-racist” and “decolonial” enterprises that hire “scholar activists” as instructors and publish commentary on news controversies, as if they were editorial boards that happened to collect tuition.”

Growing activism from the campus left is also evident among students. According to data from FIRE, there were an average of eight “deplatforming” attempts on college campuses from the political left each year between 1998 and 2012. Since 2013, however, there has been an average of 48 attempts to “deplatform” speakers each year from the political left.

Figure 6 - Deplatforming Attempts on College Campuses (1989-2023)

Unsurprisingly, the scholarship produced by universities has shifted along with these broader political changes. A study of 175 million abstracts from scholarly articles published between 1970 and 2020 found that “the prevalence in the academic literature of terms denoting prejudice and social justice discourse (e.g. racism, sexism, transphobia, Islamophobia, ableism, ageism, fatphobia) has been growing steadily and diffusely since the 1980s but that such prevalence accelerated abruptly after 2010.” While “academic scholarship seems to have passed peak ‘woke,’” the research produced by universities is now far more focused on identity-based bias and discrimination than it was 15 years ago (Figure 3).

Figure 7 – Scholarship on Bias and Discrimination between 2000 and 2022

Predictably, these on-campus developments have triggered a reaction from the off-campus political right. Most importantly, Fox News coverage of the politics of colleges and universities (but not MSNBC or CNN coverage) spiked between 2011 and 2013 and, then again, between 2015 and 2017 (Figure 4). This increasing focus on higher education’s leftward shift is undoubtedly underneath some of the right’s growing skepticism toward colleges and universities. As Figure 4 also shows, however, Fox News coverage of campus politics fell significantly after 2017 (the moment when Republican high schooler disinterest in college began accelerating most). This fact suggests that growing distrust and declining interest in higher education among conservatives is not merely a function of Fox News’s coverage.

Figure 8 – Cable TV News Coverage of Higher Education and Political Ideology

The cumulative impact of all of this has been the evaporation of conservative trust in institutions of higher education. Between 2012 and 2019, the percentage of Republicans saying that colleges have a “negative effect on the way things are going in the country” jumped from 35% to 59%. From 2015 to 2023, the percentage of Republicans expressing “a great deal” or “quite a lot” of confidence in higher education fell from 55% to 19%.

Figure 9 - Views of Colleges and Universities by Partisanship

It is also important to note that none of the changes documented above have increased confidence among Democrats (down 9% since 2015) and Independents (down 16% since 2015). And this was all before Claudine Gay’s plagiarism scandal, the campus turmoil immediately following the 10/7 attacks, and the pro-Palestine encampment movement.

Conservative students, exposed to a torrent of news stories highlighting the ideological monoculture of colleges and universities and advised by an increasingly disaffected group of conservative parents, are looking for alternatives to higher education. This is bad news for conservatives, for colleges, and for the country. Conservatives should realize, for example, that universities are still, with some notable exceptions, a strong return on investment and one of the few, proven strategies for upward social mobility. Forgoing higher education will only serve to hurt their economic interests. More broadly, a lost generation of college educated conservatives will leave corporations, media, academia, and government without the deep bench of compelling conservative voices needed to persuasively argue for right of center approaches to public policy problems.

Similarly, institutions of higher education must realize that they cannot continue to willfully alienate a large part of their potential consumer base. Facing low birth rates and declining interest from international students, colleges and universities will not be able to travel along their current course and hope to remain fiscally solvent or socially relevant. Notably, high school boys are becoming more conservative, with nearly 50% of high school boys identifying as Republican in the 2022 Monitoring the Future survey. This demographic is simply too large for colleges to ignore.

Figure 10 - Partisanship among High School Boys

If Republican boys avoid college, it will only further accelerate the trend of growing illiberalism on college campuses. As I have discussed previously, “illiberal liberals” (i.e. self-identified liberals who support all forms of disruptive action to prevent objectionable campus speech and who support deplatforming all controversial conservative speakers) are more numerous on campuses that have less political diversity.

Figure 11 - Illiberal Liberalism and Ideological Homogeneity

Given that campuses with more “illiberal liberalism” are more likely to deplatform conservative speakers, colleges and universities are in danger of entering a “death spiral” (in which the absence of conservative students leads to more prominent manifestations of the kind of “illiberal liberalism” that further alienates conservative boys and girls).

Figure 12 - Disinvitation Incidents from the Left by Percentage of Illiberal Liberals on Campus

The stakes for the country are even higher. Ideologically captured echo chambers are unlikely to solve “big” problems and they are unlikely to engender trust among a politically diverse population. If conservatives opt out of higher education, colleges and universities will lack the political diversity required to fulfill their core purpose as truth-seeking and truth-disseminating institutions. As Keith Whittington has argued, “Universities make it harder rather than easier to fulfill their core function of advancing knowledge if they nurture orthodoxies and exclude those who might be skeptical of those orthodoxies. Intellectual homogeneity too easily gives rise to intellectual complacency. Intellectual diversity can foster productive dissension.”

Colleges and universities are not the only institutions to lose the trust of the American public broadly and conservatives in particular. They are, however, the most important. Whether the decline is reversible is anyone’s guess but there are three interventions that colleges and universities may want to implement if they care about addressing the problems identified here.

First, colleges and universities should consider adopting the “Chicago Statement” (i.e. the free speech policy produced by the Committee on Freedom of Expression at the University of Chicago). Endorsing the idea “the University is committed to free and open inquiry in all matters” and that “it is not the proper role of the University to attempt to shield individuals from ideas and opinions they find unwelcome, disagreeable, or even deeply offensive,” would begin rebuilding confidence in universities as “truth seekers” and undermine the notion that they are partisan political actors advancing their own ideological agendas.

Second, colleges and universities should consider addressing their faculty “viewpoint diversity problem.” This need not take the form of ideological quotas imposed during the hiring process but colleges and universities may, as Harvard Professor Tyler VanderWeele has argued, want to “try particularly hard to hire faculty who hold disfavored or controversial views when those views are held by a large portion of the population, have not been clearly refuted, and influence culture and policy.” Eliminating mandatory DEI statements in faculty hiring, which often serve as an ideological litmus test, is also an important step in this regard. 

Finally, colleges and universities may want to heed the Kalven Report’s reminder that they are “the home and sponsor of critics,” not critics themselves.

The Growing Tension Between Political Tolerance and Equality

According to Francis Fukuyama, “The most fundamental principle enshrined in [classical] liberalism is tolerance.” For Fukuyama and others, the essence of tolerance is the “willingness to put up with, or to permit, the [speech], actions or practices of others, which one disapproves of and which one otherwise would seek to prohibit or prevent from occurring.” One of the defining features of illiberalism, then, is political intolerance or, put differently, an unwillingness to permit the speech, actions, or practices that one disapproves of.

The so-called “classical liberalism” described here is distinct, of course, from the “liberalism” of the modern American left (which is most commonly understood to mean support for “activist government to help the poor and the middle class—taxes, transfer programs, and regulation—plus a concern for civil rights and equality”). Throughout the late 20th century, however, this distinction meant very little for most Americans; the vast majority of people who identified as “liberal” strongly endorsed the values implied by both the “classical” (e.g. free expression and political tolerance) and “modern” (e.g. civil rights and equality) forms of liberalism.

A growing emphasis on the importance of equality and an ongoing reevaluation of the harms associated with free speech and tolerance, however, has recently led many self-identified liberals to abandon the “classical” elements of their liberalism. Some of these so-called “modern liberals” have even come to view free speech and tolerance as impediments to “the correction of unjust distributions produced by the market and the dismantling of power hierarchies based on traits like race, nationality, gender, class, and sexual orientation.” As Henry Louis Gates explained in a 1993 article entitled “Let Them Talk: Why Civil Liberties Pose No Threat to Civil Rights,” concerns about “hate speech” have caused many on the left “to rethink their allegiance to the very amendment that licensed the protests and the agitation that galvanized the nation in a recent, bygone era.” Now, Gates laments, “Civil liberties are regarded by many [on the left] as a chief obstacle to civil rights.”

The emergence of equality as a substantial counterweight to tolerance has encouraged the proliferation of a previously rare species of political animal: the “illiberal liberal.” Although countless commentators from across the political spectrum in recent years have complained about growing illiberalism among liberals and attempted to draw clear rhetorical boundaries between proponents of classical and modern liberalism (e.g. the distinction between “liberals” and “leftists”), there have been few systematic attempts to pin down exactly how common this insurgent variety of illiberalism is among liberals and what its consequences are.

Illiberalism on College Campuses - An Overview

This post attempts to expand our understanding of “illiberal liberalism” (i.e. the combination of intolerant attitudes towards speech with support for “activist government” and civil rights) by leveraging the Foundation for Individual Rights and Expression’s (FIRE) unique data on American college students. Specifically, I ask: how widespread is “illiberal liberalism” on college campuses today, and does this “illiberal liberalism” matter for the quality of campus speech environments?

The data presented below tell a clear story: American higher education has a small but consequential “illiberal liberal” problem. Self-identified liberals who support all forms of disruptive action (including violence) to prevent objectionable campus speech and who support deplatforming all controversial conservative speakers make up approximately 3% of college students nationally and more than 10% of students at many of the nation’s top “liberal arts” colleges. This problem appears to be driven, in part, by a lack of exposure to alternative viewpoints. As I document below, the share of students who are “illiberal liberals” is highest in the places where ideological diversity is lowest.

Although the 3% and 10% numbers might appear underwhelming at first glance, it is important to remember that it only takes a tiny handful of committed activists to successfully disrupt a campus speech or intimidate students, faculty, and staff into silence. Indeed, some research shows that the active participation of 3.5% of a given population can produce profound social, political, and economic change within that population’s community.

Consistent with the idea that small groups can have large impacts, I find that even marginal increases in the number of illiberal liberals attending a college have significant and negative consequences for the quality of that campus’s speech environment. As I show below, for instance, it is the percentage of “illiberal” liberal students on campus – not the percentage of liberal students in general – that seems to encourage deplatforming efforts targeting insufficiently progressive speakers. The evidence presented here also suggests that the percentage of “illiberal” liberals on a campus may lead the campus to experience more protests, marches, walkouts, classroom controversies, and disruptions.

This post will focus primarily on “illiberal liberalism.” The reason for this narrow focus is that there is almost no evidence that “illiberal conservatives” exist anywhere on college campuses. In FIRE’s nationally representative sample of more than 55,000 students, only 40 students (less than one-tenth of a percent) were self-identified conservatives who view all forms of disruptive action as acceptable and support deplatforming liberal speakers. On college campuses, in other words, illiberalism is solely an affliction of the left.

How to talk young liberals out of their illiberalism is one of the primary challenges facing universities. While the evidence presented here does not offer any specific advice for persuading “modern liberals” about the virtues of “classical liberalism,” it does identify the scope, location, and consequences of higher education’s illiberalism problem. By doing so, I hope this post can help better focus the much-needed and long-overdue efforts now gathering steam to reform the nation’s universities.

The Political Ideology of College Students

Every year since 2020, the Foundation for Individual Rights and Expression (FIRE) has conducted a vast, nationwide survey of college students as part of their efforts to rank the free speech climates of American universities. The 2023 survey included 55,000 enrolled students at 254 colleges around the country.

In addition to a battery of demographic (race, gender, sexual orientation, etc.) and speech-related questions (discussed further below), FIRE’s survey asked students to describe their political beliefs. To no one’s surprise, the results show that most college students place themselves on the political left, with 58.8% describing their political views as “liberal” or “democratic socialist” and only 24.4% describing their views as “conservative” or “libertarian.” Interestingly, very few (16.7%) of college students label themselves as “moderate.” This distribution of political identification makes college students distinct from the American public broadly and non-college attending members of Gen Z (both of whom are less liberal and more moderate).

Figure 1 – Self-Reported Ideology by School

The national-level picture painted above masks significant cross-campus variation in political ideology, however. On 22 of the 254 campuses surveyed by FIRE, more than 80% of students identified as liberal (Figure 1). There were even two schools – Hillsdale College and Liberty University – where more than 80% of students identified as conservative.

Figure 2 – Self-Reported Ideology by School Type

A quick glance at the schools listed in the left-hand panel of Figure 1 suggests the distribution of political beliefs varies between universities and liberal arts colleges. As Figure 2 makes explicit, private liberal arts colleges have far more liberal students (72.9% of the student population) than private or public universities (63.1% and 53.9%, respectively).

As we start thinking about the prevalence and consequences of student “illiberalism” it is important to keep this school-level variation in mind. The potential population of “illiberal liberals” across college campuses generally and private liberal arts schools in particular is much greater than the potential population of “illiberal conservatives.” Even if liberals and conservatives are equally prone to illiberalism, the vastly greater number of liberals on college campuses will mean that their illiberalism is more consequential than that of conservatives.

Illiberalism among College Students

Measuring Illiberalism in the FIRE Survey

As mentioned above, illiberalism can be defined as political intolerance or, put differently, an unwillingness to permit the speech, actions, or practices that one disapproves of. When defined in this way, the FIRE survey includes two batteries of questions that provide useful proxies for measuring how strongly students subscribe to liberal norms of political tolerance in response to offensive (yet constitutionally protected) speech.

First, FIRE’s survey asked students a series of questions to assess how tolerant they were of controversial speakers. Specifically, students were asked whether the following three speakers who had previously espoused controversial liberal ideas should be allowed to speak on campus:

• The Second Amendment should be repealed so that guns can be confiscated.

• Structural racism maintains inequality by protecting White privilege.

• Religious liberty is used as an excuse to discriminate against gays and lesbians

Students were also asked whether the following three speakers who had previously espoused controversial conservative ideas should be allowed to speak on campus:

• Abortion should be completely illegal.

• Transgender people have a mental disorder.

• Black Lives Matter is a hate group.

FIRE’s question asked students to state their opinion on whether these speakers should be allowed on campus independent of their opinions about the speaker’s message (“Student groups often invite speakers to campus to express their views on a range of topics. Regardless of your own views on the topic, should your school ALLOW or NOT ALLOW a speaker on campus who promotes the following idea?”). In other words, the question was attempting to measure how principled students were with respect to speech on campus.

Second, the FIRE survey also asked respondents “How acceptable [always, sometimes, rarely, never] would you say it is for students to engage in the following action to protest a campus speaker?”:

• Shouting down a speaker to prevent them from speaking on campus.

• Blocking other students from attending a campus speech.

• Using violence to stop a campus speech.

These two groups of questions can be combined with self-reported political ideology to categorize a student as an “illiberal liberal” or an “illiberal conservative.” To be more precise, an “illiberal liberal” can be defined as a student who meets each of the following four criteria:

1. Identifies themselves as a “liberal” or a “democratic socialist.”

2. Says all three forms of disruptive action to prevent a campus speaker are “always,” “sometimes,” or “rarely” “acceptable.”

3. Says all three of the controversial liberal speakers described in FIRE’s survey should be allowed to give a speech on campus.

4. Says all three of the controversial conservative speakers described in FIRE’s survey should NOT be allowed to give a speech on campus.

Similarly, an “illiberal conservative” can be defined as a student who meets each of the following four criteria: :

1. Identifies themselves as a “conservative” or a “libertarian.”

2. Says all three forms of disruptive action to prevent a campus speaker are “always,” “sometimes,” or “rarely” “acceptable.”

3. Says all three of the controversial conservatives described in FIRE’s survey should be allowed to give a speech on campus.

4. Says all three of the controversial liberal speakers described in FIRE’s survey should NOT be allowed to give a speech on campus.

In short, an illiberal is both a “speech disruptor” (i.e. someone who endorses all forms of disruptive action to prevent speech) and an “out-group censor” (i.e. someone who favors speech rights for all of their ideological allies but opposes speech rights for all speakers from the other side of the political spectrum).

This approach to operationalizing “illiberalism” is exceedingly (and intentionally) stringent. For example, a liberal student expressing support for non-violent but illiberal tactics to limit speech (i.e. someone who endorsed shout-downs and blockades but not violence) would not be classified as an “illiberal liberal” under these rules. Similarly, if a liberal student supports allowing even one of the three hypothetical conservative speakers the opportunity to deliver a campus speech, they will not be classified as an “illiberal” using this approach. Even a liberal who says that one of the three hypothetical liberal speakers should NOT be allowed to speak on campus will fail to qualify as “illiberal” with these criteria.

As such, the data presented below constitute a “floor-level” or “bare minimum” estimate of illiberalism on college campuses. A looser definition would find much higher overall levels of illiberalism on the left and the right.

“Speech Disruptors” and “Out-Group Censors” on College Campuses

Before identifying how many students are “illiberal” it is useful to explore how many students are, separately, “speech disruptors” and “out-group censors.” As Figure 3 shows, nearly one in four liberal college students are “speech disruptors,” meaning they endorse all forms of disruptive action (i.e. shout-downs, blockades, and violence) to prevent campus speech. A similar percentage (22.2%) are also “out-group censors” (i.e. students who would like to prevent all controversial conservative speakers from appearing on campus but would like to allow all controversial liberal speakers).

Figure 3 – Illiberalism by Self-Reported Ideology

Figure 3 also shows that conservative college students are significantly less likely to hold illiberal attitudes toward speech than liberal college students. While nearly one in four liberals is a “speech disruptor,” only 17.7% of conservatives are. More importantly, liberal college students were almost ten times more likely to qualify as an “out-group censor” than conservative college students (22.2% to 2.4%).

Figure 4 – Liberal “Speech Disruptos” and Liberal “Out-Group Censors” by School

Liberal “speech disruptors” and “out-group censors” are not equally distributed across college campuses. As Figure 4 shows, there were ten schools where more than one-fourth of students were liberal “speech disruptors” and five schools where more than one-third of students were liberal “out-group censors.” More generally, the percentage of students who are liberal “speech disruptors” is highly and positively correlated with the percentage of students who are liberal “out-group censors” (r=.633, p=.00).

Figure 5 –Conservative “Speech Disruptos” and Liberal “Out-Group Censors” by School

The distribution of conservative “speech disruptors” and “out-group censors” presented in Figure 5 tells a much different story. While there were 148 universities where more than 10% of the student population qualifies as a liberal “speech disruptor,” there was not a single school in the country where conservative “speech disruptors” make up more than 10% of the student population. Similarly, more than half of the colleges and universities surveyed by FIRE (133 out of 254) had 0 conservative “out-group” censors and there was not a single school where conservative “out-group censors” exceeded 10% of the student body. By contrast, every one of the colleges and universities surveyed by FIRE had at least one liberal “out-group censor” and more than half (143 out of 254) have student populations composed of more than 10% liberal “out-group” censors.

In short, college-level “speech disruption” and “out-group censorship” are problems concentrated entirely on the political left.

Illiberal Conservatives

How many college students are “illiberal” conservatives (i.e. they are both “speech disruptors” and “out-group censors”)? Based on the above discussion, it should come as no surprise that “illiberalism” is nearly non-existent among college conservatives. Out of the more than 11,000 conservative college students in FIRE’s survey, there were only 40 (less than one-half of one percent) that can be classified as “illiberal” (i.e. students that endorse all forms of disruptive action and support bans for controversial speakers only when those speakers are from the other side of the political spectrum).

Unsurprisingly, therefore, illiberal conservative students were completely absent from 218 of the 254 schools included in FIRE’s survey. In the remaining 36 schools, illiberal conservatives were less than 1 in 200 students.

In both a substantive and statistical sense, illiberal conservatives are so rare on college campuses today that it makes very little sense to even discuss them as a distinct group. The rest of this post, therefore, is devoted to discussing a far larger and far more consequential source of illiberalism on college campuses: liberal students.

Illiberal Liberals

How many college students are “illiberal” liberals (i.e. they are both “speech disruptors” and “out-group censors”)? Taking the two measures of illiberalism (“speech disruption” and “out-group censorship”) together, more than 5% of left-leaning college students can be classified as “illiberal.”

Liberal students, then, have a much greater tendency towards illiberalism than conservative students. When coupled with the fact that there are far more liberal students than conservative students, this increased propensity towards political intolerance produces a vast disparity in the number of illiberal liberal students and illiberal conservative students. Specifically, illiberal liberals make up 2.9% of all college students while illiberal conservatives make up only .08%. 

As already suggested by the plot in Figure 4, there’s wide variation in illiberal liberalism across campuses. While more than 15% of the student body at some schools (e.g. Oberlin College, Scripps College, Grinnell College) are “illiberal liberals,” 0% of students at other schools meet the requirements for “illiberal liberalism” spelled out above (e.g. University of Mississippi, Montana State University, Northern Arizona University).

Figure 5 - Top 20 Schools Ranked by Percentage of Illiberal Liberal Students

Figure 5 shows the top 20 schools for “illiberal” liberalism in the country. Ironically, “illiberal” liberal students are largely creatures of liberal arts colleges, with more than half the schools in the Top 20 of “illiberal” liberalism being liberal arts colleges. More generally, liberal arts colleges have a significantly larger percentage of “illiberal” liberal students than public or private universities (Figure 6).

Figure 6 -Percentage of Illiberal Liberals by School Type

The relatively high rate of “illiberal” liberalism at liberal arts colleges is likely connected to the extreme lack of ideological diversity on these campuses. As Figure 2 shows, students at liberal arts colleges are far more liberal than students at public or private universities. Just because a school has more liberal students should not necessarily mean that it will have a larger percentage of illiberal liberals. Yet, as Figure 7 shows, campuses where liberals are a greater share of the student body also have a greater share of illiberal liberals (r=.63, p=.00). More ideological homogeneity seems to produce more illiberal liberalism.

Figure 7 - Illiberal Liberalism and Ideological Homogeneity

Does Illiberal Liberalism Matter?

Disinvitation Incidents

The big question about “illiberal liberalism” is whether it matters for what happens (or doesn’t happen) on college campuses. One way to answer this question is to assess the relationship between how much illiberal liberalism there is on a campus and how frequently it experiences liberal-led efforts to deplatform insufficiently progressive speakers.

FIRE’s Campus Disinvitation database has tracked the number of so-called “disinvitation incidents” (i.e. “controversies on campus that arise throughout the year whenever segments of the campus community demand that an invited speaker not be allowed to speak”) for more than two decades. For each of the database’s 576 listed “disinvitation incidents” since 1998, FIRE has coded whether the deplatforming attempt came “from the left of the speaker” or “from the right of the speaker” and whether it was successful in revoking the invitation or disrupting the speech.

Using FIRE’s disinvitation data from the last three years, I analyzed whether campuses with more “illiberal” liberals experience a larger number of illiberal attempts to limit speech offensive to progressives. More specifically, I ran an OLS regression model (with robust standard errors) predicting a campus’s number of total and successful disinvitation incidents coming from the political left based on its percentage of liberal students, its percentage of “illiberal” liberal students, and a range of other variables measuring features of its student population (including its racial, gender and religious composition).

Figure 8 illustrates the results of this analysis by presenting the predicted number of total (left-hand panel of Figure 8) and successful (right-hand panel of Figure 8) disinvitation incidents coming from the political left for different percentages of liberal students and illiberal liberal students (holding all other variables constant at their means).

Figure 8 - Disinvitation Incidents from the Left by Percentage of Illiberal Liberals on Campus

The impact of illiberal liberalism on disinvitation incidents is unambiguous: campuses with more illiberal liberals are significantly more likely to experience successful and unsuccessful disinvitation incidents originating from the left. As Figure 8 shows, increasing the percentage of students who are illiberal liberals from 0% to 10% is predicted to triple the number of disinvitation incidents from the left overall and quadruple the number of successful disinvitation incidents from the left. In other words, the prevalence of illiberal liberals on a campus is a major factor in determining the frequency and outcome of attempts to deplatform non-progressive speakers on that campus.

It is crucial to point out here that the number of campus disinvitation incidents coming from the political left (whether successful or not) is unrelated to campus liberalism more generally. In the multivariate OLS regression models described above, the percentage of students on each campus who self-identify as liberal was not a statistically significant influence on the total number of disinvitation incidents from the left (b=.00, p=.33) or on the number of successful disinvitation incidents from the left (b=.00, p=.47) after controlling for the campus’s illiberal liberalism. Together, these findings should help clarify that campus “cancel culture” is a consequence of the growing number of “illiberal liberals” found on campus and not a consequence of left-wing, ideological homogeneity in higher education alone.  

It is also important to be clear about the limitations of the FIRE disinvitation data. Disinvitations incidents can only occur if an invitation has been issued in the first place. One way that illiberal liberals might influence campus speech environments is to prevent disagreeable speakers from being invited at all.

Consider, for example, the most illiberally liberal college in the country: Oberlin College. While nearly 22% of students at Oberlin qualify as “illiberal liberals,” the campus has not experienced a single disinvitation incident since the attempted deplatforming of Christina Hoff Sommers in 2015. Rather than concluding that campus illiberal liberalism is not actually as influential as we originally assumed, it is probably best to interpret the lack of disinvitation incidents at Oberlin as evidence that illiberal liberalism works to constrain campus speech environments in many ways that are not captured by the simple number of disinvitation incidents (e.g. discouraging students and faculty from even considering hosting a speech by a right of center speaker on campus). To state all of this differently, the findings from the disinvitation database capture only one, very narrow dimension of illiberal liberalism's influence on college speech environments.

Incidents of Disruption

In an attempt to broaden my assessment of the impact of illiberal liberalism on campus politics, I also built upon the recent work of Nathan Honeycutt and Andrea Lan. In a fascinating and carefully executed study of recent campus “incidents” related to the Israel-Palestine conflict, Honeycutt and Lan show that protests, marches, walkouts, classroom controversies, disruptions, denouncements of university officials, and vandalizations were more common at “colleges where students are more accepting of illiberal forms of protest and where students have greater difficulty discussing the Israeli-Palestinian conflict.”

Figure 9 - On-Campus Incidents Related to the Israel-Palestine Conflict by Percentage of Illiberal Liberals on Campus

Using Honeycutt and Lan’s measure of campus incidents related to the Israel-Palestine conflict, I ran a logistic regression model with the same set of variables included in the disinvitation analysis above. Perhaps unsurprisingly, the results (Figure 9) show that campuses with a larger percentage of students who are illiberal liberals had a significantly higher probability of experiencing a disruptive, on-campus incident connected to the Israel-Palestine conflict between October 7 and November 6, 2023. Although the impact of illiberal liberalism on incidents for other issues might be more muted, these results are consistent with the possibility that more illiberal liberalism at a school produces more contentious campus political environments overall.

Concluding Thoughts

The above analyses are likely to significantly underestimate the extent and consequences of illiberal liberalism on college campuses. As discussed above, I adopt the strictest possible definition of illiberalism (requiring support for both “speech disruption” and “out-group censorship”) and I rely primarily on “disinvitation incidents” to assess illiberal liberalism’s impact (thereby missing the chilling effect it undoubtedly has on other dimensions of a campus’s speech environment). Put differently, as concerning as the above findings are, they are probably minimizing the significance of illiberal liberalism on college campuses.

Even with this in mind, the evidence presented here has documented a concerning asymmetrical illiberalism on college campuses. The asymmetry is largest on the campuses of private liberal arts colleges but its consequences are felt at universities around the country. Any attempt to address the campus “free speech crisis” and create healthier intellectual climates at our colleges and universities will need to address the small but significant problem of illiberal liberalism documented here.

Subscribe now

• Understanding the Barriers – and Solutions – to America’s Youth Mental Health Crisis

• Team USA player uses World Cup to address mental health crisis

• Baltimore City Schools sues Facebook and others for fueling mental health crisis

• How social media is contributing to a mental health crisis in youths

• Why Are So Many Young People Struggling With Mental Health?

• Study Delivers More Evidence of a Mental Health Crisis Among Teens, Particularly Girls

• We need new solutions to the crisis in youth mental health

• Reinventing the E.R. for America's Mental-Health Crisis

• Amid mental health crisis, school teaches students how to help each other

More systematically, a content analysis of mainstream news coverage reveals a dramatic surge in the media’s interest in mental health. As Figure 1 shows, major American newspapers have devoted a far greater share of their coverage to mental health issues in general and mental health issues among young people over the last two decades (with the number of yearly articles per 1,000 increasing by more than 10x between 2000 and 2022).1 As Figure 1 also shows, most of the change occurred after 2017. In the last five years, mental health (particularly among teenagers) has become a central fixation of the country’s public discourse.

Figure 1 - Major Newspaper Coverage of Mental Health Issues

The intensifying concern about mental health should not be surprising. The last few years have seen a seemingly endless stream of shocking new studies about the low and declining quality of young people’s mental health. For example, using the Monitoring the Future (MTF) studies (which are yearly, nationally representative surveys of high school students), Jean Twenge has shown a dramatic increase in depressive attitudes among American teens (Figure 2):

Figure 2 - Depressive Symptoms among American Teenagers

Drawing on the Behavioral Risk Factor Surveillance System (a yearly survey of 100,000+ U.S. adults administered by the CDC), Twenge has also shown that young people are experiencing a rapidly growing number of “poor mental health” days each month (Figure 3).

Figure 3 - Poor Mental Health Days Each Month by Age

Providing further confirmation of these trends, a 2022 report by the American Psychological Association found that young people (particularly young women) are incapacitated by daily stress. Specifically, a majority of those under 35 claimed that “stress is completely overwhelming most days.” By contrast, less than 9% of those 65 or over experience “overwhelming” stress (Figure 4).

Figure 4 - Overwhelming Stress by Age and Gender

Other data from 2022 paints a similar (yet equally disturbing) picture of the steady, intergenerational decline in Americans’ mental health (Figure 5). While more than 70% of the Silent generation (born 1928 to 1945) and 60% of Boomers (born 1946 to 1964) report “excellent” or “very good” mental health, only 29.4% of Gen Z (born after 1996) do. Gen Z’s mental health stands out even in comparison to Millennials, with members of Gen Z nearly 15% more likely than Millennials (born 1981 to 1996) to claim their mental health is “fair” or “poor.” Clearly, there is something unique going on with Gen Z. In other words, the “kids aren’t alright.”

Figure 5 - Self-Reported Mental Health by Generation

So what is wrong with Gen Z? I’m not the first analyst, of course, to attempt to answer this question (see, for example, Jonathan Haidt and Jean Twenge’s work). In this post, however, I’m going to be the first to use the novel data provided by the 2022 Cooperative Election Study (CES). The CES has a number of characteristics that may help shed new light on the dynamics behind young people’s (particularly Gen Z’s) deteriorating well-being. To be more specific, the CES surveys a large enough number of Americans (60,000) that there is a sufficient number of Gen Z respondents to draw meaningful conclusions (N=5,484), asks a question about mental health (“Would you say that in general your mental health is…excellent, very good, good, fair or poor?”), and includes a diverse array of questions about who respondents are and what they believe.

Let’s use the CES, then, to get a better sense of the kinds of attributes associated with better (and worse) mental health among Americans 25 and under.

Gender

Figure 6 - Self-Reported Mental Health by Gender among Members of Gen Z

The data displayed in Figure 6 provide show a massive mental health gap based on gender. Men were 16% more likely than women to report “excellent” or “very good” mental health and 33% more likely to report “excellent” or “very good” mental health than non-binary and “other” gender-identifying young people.

Sexual Orientation

Figure 7 - Self-Reported Mental Health by Sexual Orientation among Members of Gen Z

Gay, lesbian, and bisexual members of Gen Z claim their mental health is significantly worse than heterosexual members of Gen Z. As Figure 7 shows, heterosexual young people (35.1%) were more than twice as likely to say their mental health was “excellent” than gay, lesbian, or bisexual young people (13.6%).

Church Attendance

Figure 8 - Self-Reported Mental Health by Church Attendance among Members of Gen Z

According to political scientist, Ryan Burge (author of the indispensable Substack site “Graphs About Religion”), not every dimension of religiosity is equally important from a social, political, and individual perspective. Attendance at church services, as opposed to holding intense beliefs or identifying strongly with a particular religious tradition, is likely to offer the most significant and wide-reaching set of benefits. More directly, using Pew data from 2020, Burge has shown that church attendance is correlated with a lower probability of reporting anxiety, depression, and loneliness.

Consistent with Burge’s insights, church attendance is significantly and positively correlated with the quality of one’s mental health. As Figure 8 shows, those who attend church more than once a week are nearly three times as likely as those who never attend church (19.6% to 53.3%) to claim that their mental health is “excellent” or “very good.”

Marital Status

Figure 9 - Self-Reported Mental Health by Marital Status among Members of Gen Z

Very few 18 to 25-year-olds (8%) are currently married. Yet, marital status is strongly correlated with better mental health among Gen Z. As shown in Figure 9, married members of Gen Z are much more likely (43.8%) to report “excellent” mental health than unmarried members of Gen Z (28.1%).

Education

Figure 10 - Self-Reported Mental Health by Education among Members of Gen Z

Educational differences in well-being are slightly smaller than differences connected with gender, sexual orientation, religiosity, and church attendance. Additionally, contrary to these other variables, educational differences appear stronger on the low end of the mental health spectrum than on the high end. As Figure 10 shows, the educational gap is largest in the “fair” or “poor” mental health category, with those who have at least a college degree approximately 15% less likely to describe their mental health as “fair” or “poor” than those who do not.

Social Media

Figure 11 - Self-Reported Mental Health by Social Media Use among Members of Gen Z

The CES does not include the kind of finely-grained measure of social media usage that would allow us to assess whether there are differences in mental health based on the intensity (i.e. time spent, number of engagements, etc.) and location (i.e. Twitter, Facebook, Instagram, TikTok, etc.) of an individual’s social media activities. It does, however, include an item asking respondents whether they have used social media at all in the last 24 hours. Despite its limitations, I used this item as a rough proxy for social media activity.

Even this overly broad measure suggests a significant relationship between social media use and mental health. Paralleling the relationship with education, the primary difference between social media users and non-users is on the low end of the mental health spectrum. As Figure 11 shows, while 47.3% of daily social media users reported “fair” or “poor” mental health, only 29.2% of those who do not use social media each day did.

Political Ideology

Figure 12 - Self-Reported Mental Health by Ideology among Members of Gen Z

Figure 12 displays the widely discussed ideological differences among 18 to 25-year-olds, with a clear “staircase” pattern of improving mental health as young people become more conservative. As Figure 12 shows, very conservative Gen Z’ers (49.3%) are actually more than twice as likely to say that their mental health is “excellent” than very liberal Gen Z’ers (20.8%).

Military Service

Figure 13 - Self-Reported Mental Health among Active Duty and Civilian Members of Gen Z

As Figure 13 shows, there are vast differences in mental health among members of Gen Z based on whether they are currently serving in the armed forces. To be exact, while 54.2% of Gen Z’ers currently on active duty reported “excellent” mental health, only 28.5% of civilian Gen Z’ers did.

Figure 14 - Self-Reported Mental Health among Veteran and Non-Veteran Members of Gen Z

The differences in mental health between Gen Z’ers broadly and those connected to the military are not limited to those who are currently serving. A similar difference exists between those who previously served and those who have not, with veterans nearly 25% more likely to say their mental health is “excellent.”

What’s Different About Gen Z?

The above analyses show that being heterosexual, being married, being college educated, attending church frequently, serving in the military, and avoiding social media are all associated with better mental health among Gen Z’ers. Yet, all of these things are relatively rare and/or declining among the country’s 18 to 25-year-olds:

Figure 15 - Marriage Rates among American Women by Birth Decade

Let’s start with marriage. Marriage is strongly correlated with better mental health (Figure 9). In 2022, however, only 8% of Gen Z’ers were married. In addition to contributing to Gen Z’s current mental health crisis, this rate puts Gen Z on track for historically low levels of marriage. As indicated in Figure 15, the percentage of women married by age 28 has dropped from each cohort to the next, by about 11 percentage points per decade. If the marriage trends in Figure 15 hold and if marriage continues to be associated with better mental health, the country’s aggregate mental health is likely to become worse in the coming years.

Figure 16 - Church Attendance by Year among 18 to 25-Year-Olds

A similar story can be told about church attendance. Currently, 37% of Gen Z never attend church. As Figure 16 shows, young people were not always this way. Indeed, the share of 18 to 25-year-olds who never attend religious services has been steadily increasing over time. Fewer church attendees means less overall well-being among Gen Z (as indicated by the strong relationship between church attendance at mental health). To the extent that church attendance continues to be associated with improved mental health, this trend bodes ominously for the future.

Figure 17 - Sexual Orientation by Generation

One of the most widely discussed features of Gen Z is the comparatively high percentage of people who identify as lesbian, gay, and bisexual. Data from Gallup, for example, shows that nearly one in five members of Gen Z identify as LGBT (more than five times greater than the rest of adults). Given that these groups provide significantly worse assessments of their mental health than heterosexuals (Figure 7), the growing number of people identifying as gay, lesbian, and bisexual may (if nothing changes) lead to a noticeable overall decline in the country’s mental health.

Figure 18 - College Enrollment by Year

To date, only 15% of Gen Z have graduated from college. While this number will undoubtedly increase as Gen Z ages, there are some indicators suggesting this increase will be smaller than expected. Perhaps the most important is college enrollment (a leading indicator for college graduation). In Fall 2020, there was a 4.3% decline year-over-year. Preliminary data indicate that total postsecondary enrollment fell by another 14.7% between Fall 2020 and Spring 2022. Assuming college graduation continues its strong association with mental health (as shown in Figure 10), the decline in college graduates will mean fewer mentally well Americans.

Figure 19 - Likelihood of Serving in the Military among American Teenagers

While Gen Z’ers constitute a relatively larger share of the military than other generations, less than one-half of one percent of Americans serve in the armed forces at any given time. According to the JAMRS surveys, the propensity to serve among 16 to 21-year-olds has steadily decreased over the last few years. The small number of Gen Z’ers with a connection to the military limits the overall significance of military service to the country’s current and future mental health situation. Nevertheless, it is important to point out that, much like church attendance, marriage, and college enrollment, there is declining interest in military service among American youth.

Figure 20 - Social Media Use among Teenagers by Year

According to the Pew Research Center, approximately half of 18 to 29-year-olds are online “almost constantly,” with the bulk of this time being spent on social media apps such as Instagram, Snapchat, and TikTok. As Figure 19 shows, older Americans spend far less time on social media. While we obviously need better, more nuanced measures of social media use to draw firmer conclusions than the ones presented here (Figure 11), it does appear that Gen Z’s social media use has been (and will continue to be) an important part of the story of their mental health struggles.

Figure 21 - Political Ideology by Generation

Last but certainly not least is Gen Z’s rejection of political conservatism. As Figure 20 shows, Gen Z is currently the most liberal and least conservative generation, with Gen Z roughly twice as likely to identify as liberal than Silent Generation members or Boomers. If Gen Z does not get more conservative as they age and political ideology continues to be strongly associated with mental health, we could see a worsening national, mental health picture going forward.

Overall, then, the beliefs and behaviors of Gen Z: (1) are distinct from those of previous generations; (2) strongly associated with poor mental health; and (3) appear to be trending away from previous generations and away from those most strongly associated with better mental health.

What This Analysis Can (and Can’t) Tell Us About Gen Z’s Mental Health

There are five important things to point out about the analyses presented above. First, these bivariate relationships illustrated in each of the above figures hold up in a multivariate analysis that controls for other important causal variables. To be more precise, I ran a multivariate OLS regression model predicting self-reported mental health on the basis of race, gender, sexual orientation, political ideology, family income, education, religiosity, marital status, parenthood, social media use, and military service (active duty and veteran). As the coefficient plot in Figure 22 shows, every one of the relationships documented above remains significant even after the addition of statistical controls.

Figure 22 - Multivariate OLS Regression Analysis Predicting Self-Reported Mental Health among Members of Gen Z

Second, the findings presented here provide additional confirmation for the congealing conventional wisdom about the determinants of mental health among young people. Most recent accounts of Gen Z’s mental health have focused on the role that gender and political ideology play in shaping well-being. A recent paper by Gimbrone et al., for example, identified a large and accelerating gap in well-being between liberal and conservative teens. As their analysis also showed, liberal girls have consistently expressed more depressive attitudes than liberal boys (Figure 23).

Figure 23 - Depressive Affect Scores by Gender and Ideology among American Teenagers

Extending Gimbrone et al.’s analysis of the MTF data through 2021, Zach Goldberg finds widening gender and ideological disparities in measures of depression, self-esteem, life satisfaction, self-derogation, loneliness, and general well-being. Goldberg shows, for example, recent growth in the “female-to-male” and “liberal-to-conservative” gaps with respect to seeing a “doctor or other health professional at least once for some emotional or psychological problem of symptom in the past 12 months” (Figure 24).

Figure 24 - Treatment for Emotional or Psychological Problems by Gender and Ideology among American Teenagers

Similar to studies of American teenagers, these studies also reveal that women and liberals are more likely to experience mental health issues (with the largest differences existing among younger adults). Jonathan Haidt, for instance, used a 2020 Pew survey on formal mental health diagnoses to show that a majority of 18 to 29-year-old liberal women have been told by a “doctor or healthcare provider” that they “have a mental health condition.”

Figure 25 - Mental Health Diagnoses by Age, Gender, and Ideology

There are also gender and ideological differences in “life satisfaction.” As Brad Wilcox writes, “Liberals, especially liberal women, are significantly less likely to be happy with their lives and satisfied with their “mental health”, compared to their conservative peers aged from 18-55.”

Figure 26 - Life Satisfaction by Gender and Ideology

Summarizing this work in an incredibly thoughtful and thorough review of the literature that should be read by everyone with an interest in these topics, Musa al-Gharbi writes:

“the well-being gap between liberals and conservatives is one of the most robust patterns in social science research. It is not a product of things that happened over the last decade or so – it goes back as far as the available data reach. The differences manifest across age, gender, race, religion, and other dimensions. They are not merely present in the United States, but in most other studied countries as well.”

The data in Figures 6 and 12 are, of course, entirely consistent with each of these accounts. To further illustrate this consistency, Figure 27 shows the interaction between gender and ideology among members of Gen Z in the CES data.

Figure 27 - Self-Reported Mental Health by Gender and Ideology

Figure 27 again shows that conservatives report higher levels of mental health than liberals and that the relationship between ideology and well-being is conditioned by gender.

The coefficient plot in Figure 22 also helps us expand on these previous studies. For example, some have suggested that the well-being gap is simply a product of conservatives being more religious, more likely to be married, and more likely to be male. While all of these things are true (e.g. 17% of Gen Z’ers that attend church more than once a week are married but only 6% of Gen Z’ers that never attend church are), being more conservative exerts an influence independent of these factors. The benefit of more conservative views manifests even when controlling for all other influences on mental health.

While the analysis here helps us start to identify the aspects of conservatism that produce better mental health by ruling out the confounding effects of gender, marriage, and religiosity, there are a wide range of other possible explanations that our data cannot speak to. As al-Gharbi summarizes:

some have argued that the apparent psychological benefit of conservativism actually comes from feeling deeper connections with one’s country, one’s family and the Divine….In a similar vein, studies have repeatedly found that conservatives – both politicians and laymen – tend to be more conventionally attractive than liberals (and have better sex lives). Moreover, people who are healthier in childhood have been shown to be more likely to become conservative as adults. Meanwhile, people with high measured cognitive ability are also more likely to support economic conservativism (and cultural liberalism)…. conservativism and ideological fellow travelers (religiosity, patriotism) may help people make sense of, remain resilient in the face of, and respond constructively to, inequality and misfortune – irrespective of where they fall on the social strata. Liberal ideology, by contrast, may not provide the same benefits to adherents. 

As suggested above, the CES does not allow us to test all of these possibilities. What’s most important for our purposes here, however, is that these findings once again show that even among Gen Z, conservatism is correlated with better mental health (independent of gender, religiosity, and marital status).

Third, the findings here also allow us to highlight some previously under-discussed correlates of mental health among young people. Most notably, the results presented here allow us to make some initial observations about military service’s relationship to mental health. There is extensive and much-needed reporting on mental health issues among active duty members of the military and veterans. Most of this reporting focuses on the prevalence of serious mental health issues, such as PTSD and suicide. A consistent theme of this research is that those connected to the military (now or in the recent past) have a higher (though time-varying) risk of experiencing major mental health problems. For example, 7% of veterans will experience PTSD at some point in their life (compared to 6% of Americans in the general population). Similarly, after adjusting for the demographic differences between the military and the rest of the American population (e.g. accounting for the fact that the military is disproportionately young, white, and male), the suicide rate among members of the military has fluctuated from slightly lower than the public more generally (e.g. 17 per 100,000 compared to 20 per 100,000 in 2010) to much higher (e.g. according to Defense Department data, suicides among active-duty service members increased by more than 40% between 2015 and 2020).

Regardless of the actual numbers, however, the perception among many Americans is that military service is bad for mental health. In fact, the perceived negative mental health consequences of military service have become a major problem for military recruitment among young people. According to the JAMRS surveys (conducted by the Department of Defense on a tri-annual basis to measure young people’s propensity to serve in the military), the “possibility of PTSD or other emotional/psychological issues” are a primary reason many young Americans say they would not consider joining the military. In fact, concerns about these mental health issues have been the second most frequently cited reason to avoid military service since at least 2017 (consistently mentioned by more than half of those who do not want to serve).

Yet, the relationship between military service and other, less formal measures of well-being receives far less attention. Part of the reason why is that it is difficult, if not impossible, to use nationally-representative surveys to compare those with military backgrounds to those without. Even if a survey asks both a question about mental health and questions about military service, it typically includes only 20 to 30 members of the military and only 40 to 50 veterans. The samples will be even smaller (often a dozen or so) if attempting to look at the role that military service plays among young people. With sample sizes this small, it is impossible to draw conclusions about the mental health impact of military service. We are left, instead, with data about only the most extreme (and, often, formally diagnosed) mental health problems.

As shown above, members of Gen Z that are currently serving or that have previously served report significantly better mental health than members of Gen Z that have not. This relationship is large and exists net of statistical controls. As I discuss below, we should be cautious in interpreting these results. It is possible that this is a selection effect (i.e. more mentally healthy people choose to join the military in the first place) rather than a causal effect resulting from military service itself. It also should not be taken to minimize the serious and negative consequences that military service has had on a large number of young Americans. It should, however, be part of a broadening conversation about how both military and civilian young people are coping with mental health struggles.

Fourth, the role that gender, sexual orientation, religiosity, marriage, education, active duty military service, and social media use play in mental health is not unique to Gen Z. In fact, these variables appear to connect with mental health perceptions in the same way among all Americans. Indeed, running the same multivariate, OLS regression model for Americans over 25 years old that I ran for Americans between the ages of 18 and 25, produces strikingly similar results:

Figure 28 - Multivariate OLS Regression Analysis Predicting Self-Reported Mental Health among All Americans

As Figure 28 shows, being a man, being a heterosexual, being married, being college educated, attending church frequently, currently serving in the military, and avoiding social media are all associated with better mental health among members of the Silent, Boomer, Gen X, Millennial, and Gen Z generations.2 In other words, what has worked before from a mental health perspective appears to be working again.

This insight helps us assess why so many in Gen Z describe their mental health as “fair” or “poor” rather than “excellent” or “very good.” Before analyzing the data, we could not rule out the possibility that the variables shown to positively influence mental health among previous generations (e.g. a college degree, church attendance, etc.) do not exert the same impact on Gen Z. It was possible, for example, that marriage (which has historically been associated with better mental health) is no longer correlated with better mental health among Gen Z because it is no longer seen as a necessary condition for maintaining a meaningful relationship and raising children. If this were true, mental health outcomes could become worse even with an increase in marriage rates among Gen Z.

The evidence presented here deals a serious blow to this kind of argument. Indeed, given that the correlation between well-being and gender, sexual orientation, religiosity, marriage, education, active duty military service, and social media use is roughly the same for Gen Z as it is for other generations, we can be more confident in the argument that the worsening mental health of Gen Z is simply a function of having fewer people doing the things that are associated with well-being (which I documented in the previous section).

Finally, cross-sectional analyses like the one presented here have a number of problems when it comes to making compelling causal arguments. Two of these problems are particularly important for this analysis: selection effects (e.g. married people might have better mental health not because marriage produces better mental health but, instead, because people with better mental health choose to get married) and omitted variable bias (e.g. because the analysis does not control for the presence of permanent disability, serious illness, physical attractiveness, etc.) it is producing biased conclusions about the impact of the other variables on mental health). As a result of these two problems, I have avoided using causal language here and the findings should not be used to argue that marriage, a college degree, military service, adopting conservative political ideology, or social media use will lead to better mental health. As always, more work (and better) data is needed.

This, of course, is an absurd idea with very little empirical data to support it. For better or worse, American society is top-down and elite-driven, with a small number of people wielding a larger amount of influence through their positions in key social institutions. As discussed below, this relatively tiny group is often educated at a relatively tiny number of elite universities. If we want to understand how Gen Z will change American society, we need to look narrowly at the members of Gen Z that are enrolled in the nation’s most prestigious universities instead of looking at Gen Z more generally.

The findings presented below show that America’s elite universities are politically distinct from the country they will ultimately preside over. To be more exact, elite university students are far more liberal and far more Democratic than older Americans and their non-college attending, Gen Z counterparts. Furthermore, the analyses here show that there is a unique and somewhat contradictory pattern of thinking about political tolerance at America’s elite universities. Students attending Top 20 universities are significantly more likely than students from lower-ranked universities to say that they would allow both extreme liberal and extreme conservative speakers to deliver a speech on campus. Yet, elite university students’ greater willingness to allow controversial speech is tempered by the fact that they are far more willing than students attending non-elite schools to support shout-downs, blockades, and violence as a means of preventing an offensive speech from taking place. Elite university students are, in short, paradoxically less likely to express intolerance in the form of explicitly banning offensive speakers from campus and more likely to express intolerance in the form of supporting illiberal reactions designed to prevent speech they find offensive.

Over the course of describing the uniqueness of the country’s elite universities, we will also uncover a number of other important attitudinal characteristics of American college students. The analyses below show, for example, that fewer than 10% of women attending elite universities identify as liberal. Additionally, we will see that women are more likely to support de-platforming offensive speakers than men and “agender,” “genderqueer,” “non-binary” or “unsure” students are more likely to support disruptive action in response to speech than men or women. More generally, the analyses below show the groups that have historically benefitted most from robust norms emphasizing the value of free speech, political tolerance, and peaceful protest – progressives and members of national racial, gender, and sexual minority groups – are now the least supportive of them.

As the institutions of American life undergo rapid and dramatic compositional changes - becoming more liberal, more female, and more racially and ethnically diverse - and as political polarization intensifies partisan and ideological conflicts, the future of America’s distinctive brand of liberal democracy is unclear. This post hopes to provide some insight into the political values held by the group most likely to play a major role in this uncertain future: the country’s elite university students.

Why Elite Schools Matter?

Understanding the politics of elite universities is important because elite universities produce a vastly disproportionate share of the elites that control the country’s central political, economic, legal, technological, and educational institutions. Consider, for example, the educational pedigree of elected officials. A 2020 study of Congress found that while only 0.4% of undergraduates attend one of the Ivy League schools, 8% of members of the House and the Senate were Ivy League alumni. Furthermore, all but two of the top 10 most attended universities by members of Congress were featured in the Top 20 of US News and World Report’s College Rankings (Figure 1).

Figure 1 - Educational Pedigree of US Lawmakers

What about political appointees? While data here is harder to track down, a 2014 National Journal survey of the 250 top decision-makers within the Obama administration found that 40% were Ivy League graduates (with more than 60 attending Harvard alone). In fact, more Obama administration officials had degrees from England’s Oxford University than from any American public university. 

The same pattern emerges when looking at the federal court system. There are 204 American Bar Association-approved law schools in the United States. As Figure 2 shows, in the past four decades, more than two-thirds of all Supreme Court law clerks have graduated from one of five law schools: Harvard, Yale, Stanford, Columbia, and the University of Chicago, and 94% of the clerks have come from law schools ranked in the top 25 by U.S. News and World Report.

Figure 2 - Annual Composition of Law Schools from where Court Clerks Graduated (1980-2020)

It’s not just elite law schools that matter, however. A recent paper found that where a person received their undergraduate degree exerted a significant influence on whether they were ultimately selected to be a Supreme Court clerk (independent of where they received their law degree). For instance, those who graduated from one of 22 highly selective undergraduate institutions were more likely than undergraduates from less prestigious schools to be chosen as Supreme Court clerks (even among those who graduated from Harvard Law with honors).

Figure3 - Educational Pedigree of Supreme Court Justices

The makeup of the Supreme Court provides even further evidence of the dominance of elite educational institutions in the federal judiciary. In 2017, “Every Supreme Court justice went to Harvard or Yale Law School.” Currently, eight of the nine current Supreme Court justices attended law school at Yale or Harvard (Figure 3).

Figure 4 - Educational Pedigree of American Business Leaders

Unsurprisingly, business is no different than government when it comes to dependence on elite universities. In a study that looked for CEO in the job titles of more than 3.5 million LinkedIn profiles, Resume.io found that the most highly ranked universities in the country produce the most business leaders. As Figure 4 shows, nine of the ten universities that produce the most executives all land in the Top 25 of US News and World Report’s college rankings.

Overall, in a paper assessing the educational backgrounds of a wide variety of elite professions, Wai and Perina find that alumni from the 29 universities with the highest median combined SAT Math and Critical Reading scores and the 12 most highly ranked law and business schools were vastly overrepresented among leaders in business, government, the law, and journalism (Figure 5).

Figure 5 - Educational Pedigree of American Elites

We could go on and on and on like this. The main point is that a relatively small number of universities produce a vastly disproportionate share of our scientists, business leaders, lawyers, and policymakers. The values that students of these universities learn and practice during their time on campus inevitably seep into other major social institutions. If we are interested in the future of political tolerance and free speech, for instance, it is important to know not just what college students broadly think but also what students of elite universities believe.

A particular concern about elite universities today is that they have become more ideologically homogenous and less politically intolerant. To be more direct, elite universities are often criticized for their perceived hostility to right-leaning political figures, ideas, and speech. As Reihan Salam, for example, has pointed out that “many on the right see Ivy League universities and other similarly selective institutions as bastions of left-liberalism, where dissent from the right is tolerated only begrudgingly.” Similarly, Harvard law professors Jack Goldsmith and Adrian Vermeule have complained that “the distinctive progressive ideology of elite universities is relentlessly critical of, to the point of being intolerant of, traditions and moral values widely seen as legitimate in the outside world.” If this critique is accurate (i.e. elite universities are attracting and encouraging an unrepresentative and intolerant set of political views) and alumni of elite universities continue to dominate important social institutions, it could very well lead to a fundamental shift in the character of America’s liberal democratic order.

So, is there any basis for these criticisms? In the sections that follow, I use the Foundation for Individual Rights and Expression (FIRE) 2022 College Free Speech Rankings Survey data to identify what elite university students actually believe (and, importantly, whether it is different from what everyone else believes). More specifically, I answer three specific questions about America’s elite universities:

1. Are students at elite universities politically diverse?

2. Are students at elite universities politically intolerant?

3. Are students at elite universities supportive of disruptive action in response to constitutionally protected speech?

In order to answer these questions, I am going to examine the attitudes of those attending elite universities and compare them to the attitudes of those who attend non-elite universities. For the purposes of this analysis, I treated the top 20 schools in the US News and World Report rankings as “elite” while all other schools were treated as “non-elite.” The “elite” (or top 20) universities were: Princeton, MIT, Harvard, Stanford, Yale, University of Chicago, Johns Hopkins, University of Pennsylvania, Cal Tech, Duke, Northwestern, Dartmouth, Brown, Vanderbilt, Rice, Washington University, Cornell, Columbia, Notre Dame, Berkeley, and UCLA.

The decision to focus on the top 20 schools is, of course, fairly arbitrary and there are myriad problems with the US News and World Report rankings. The top 20 are responsible, however, for the vast majority of alumni identified in the graphics above.

Are students at elite universities politically diverse?

Measuring Political Beliefs

Every year since 2020, FIRE has conducted a nationwide survey of college students as part of their efforts to rank the free speech climates of American universities. The 2022 survey included 45,000 enrolled students at over 200 colleges around the country. In addition to questions about basic demographic characteristics (race, gender, sexual orientation, etc.), the FIRE surveys also asked respondents about their political views (i.e. their self-described political ideology and their partisan identification). When coupled with data about the university each student attends, it is possible to assess whether elite universities are more politically homogenous than non-elite universities.

The data from FIRE can also be compared against nationally representative surveys of the American public in order to assess whether college students generally or elite college students more specifically are similar to their non-college counterparts. While there are many potential surveys to use for this purpose, I will rely on the 2022 Cooperative Election Study (CES). The CES is useful for given my goals here because it has more than 5,000 Gen Z respondents and it poses questions about partisanship and ideology that are similar to those asked in the FIRE survey.

Eric Kaufman has used the FIRE data from 2020 and 2021 for a similar analysis. I highly recommend reading it for more discussion on these issues.

Partisanship and Ideology among All Americans

The left-hand panel of Figure 6 shows that Democrats have a major partisanship advantage (55.4% to 27.6%) among the young. The data for self-reported political ideology presented in the right-hand panel of Figure 6 tells a similar story. Young people are 20% more likely to identify as liberal than as conservative (43.2% to 22.2%).

Figure 6 - Partisanship and Ideology by Generations

There is a small but statistically significant gender gap among young people, with women being significantly more Democratic and significantly more liberal than men. As Figure 7 shows, the partisan gender gap is approximately 7% for partisanship (with 52.1% of men and 59.4% of women identifying as Democrats) and approximately 7% for ideology (with 37.8% of men and 44.6% of women identifying as liberal).

Figure 7 - Partisanship and Ideology among Gen Z by Gender

Partisanship and Ideology among College Students

The gaps illustrated by Figures 6 and 7 are present but far more pronounced among college students than among other members of Gen Z. The advantage for Democrats and liberals is particularly prominent among students attending elite universities. As the left-hand panel of Figure 8 shows, Democrats outnumber Republicans by a factor of two to one (24.8% to 57.0%) at non-elite universities and by a factor of four to one (15.4% to 70.0%) at elite universities. As the right-hand panel of Figure 8 shows, there are 33% more liberals than conservatives at non-elite universities (57.7% to 24.7%) and 55% more liberals than conservatives at elite universities (71.9% to 16.4%).1

Figure 8 - Partisanship and Ideology among College Students

Figure 9 shows party identification among men and women at highly-ranked and lower-ranked universities. The most important thing to point out here is the massive size of the gender gap among college students. At both elite and non-elite universities, the gender gap is nearly 20%. As Figure 9 shows, 45.9% of men at non-elite universities identify as Democrats while 65.1% of women at non-elite universities identify as Democrats. Similarly, 60.9% of men at Top 20 schools identify as Democrats while a whopping 79.1% of women at Top 20 schools identify as Democrats.

Figure 9 - Partisanship among College Students by Gender

As Figure 10 shows, there is also a roughly 20% gender gap in ideology for college students (regardless of how prestigious their university is). Perhaps the most striking finding from Figure 10, however, is the fact that fewer than 10% of women at elite universities identify as conservative.

Figure 10 - Ideology among College Students by Gender

Overall - America’s Elite College Students are Politically Unrepresentative of Other College Students, Other Young People, and Other Americans

Collectively, these analyses allow us to draw three important conclusions about elite university students. First, students at elite universities are vastly different in their politics from members of Gen Z in the country more generally. For example, Top 20 university students are 15% more Democratic than other members of Gen Z. Similarly, while 43% of Gen Z are liberal, nearly 72% of those attending elite universities are liberal. The left-leaning and strongly Democratic politics of elite university students will likely mean that most of our important social, economic, political, and legal institutions will fail to adequately represent the country in the years to come. This lack of political representation has the potential to accelerate an already harrowing collapse of institutional trust and precipitate a foundational crisis in the legitimacy of foundational elements of American democracy.

Second, although elite universities are not politically homogenous, they are certainly dominated by the political left. As discussed above, roughly two-thirds of elite university students identify as Democratic and liberal in their politics. This, of course, says nothing about the political homogeneity of the faculty. As Langbert, for example, found, the Democrat-to-Republican ratios at elite colleges and universities ranged from 1.6:1 in the field of engineering to 56:0 and 108:0 in communications and interdisciplinary studies (Figure 11).

Figure 11 - Partisan Identification of Faculty at Elite Universities

The dearth of Republican and conservative students and faculty at Top 20 universities is particularly striking in a country where Republicans frequently win control of the local, state, and federal government institutions. The fact that liberal, Democratic elite university students are unaccustomed to sharing institutional space with conservative Republicans may present serious governance challenges going forward.

Finally, the combination of the gender and educational gaps is creating a huge political divide - approximately 30-points in partisanship and ideology - between elite college-educated women and non-college educated men. This political divide may have ripple effects throughout other parts of American society. If Americans continue to remain reluctant to marry across partisan, ideological, and educational lines, for example, marriage rates may continue their precipitous decline. As Stone and Wilcox recently claimed in a widely circulated Atlantic Monthly piece (entitled “Now Political Polarization Comes for Marriage Prospects”), “fewer Americans are willing to date or marry across the aisle” and this is “bad news for marriage.”

The main point here, however, is that today’s elite college students are far more liberal and far more Democratic than previous generations and, more importantly, most other young people.

Are students at elite universities politically intolerant?

Measuring Intolerance of Controversial Speakers

In addition to the aforementioned questions about demographic characteristics and political identities, the FIRE survey also asked respondents a series of questions intended to assess how tolerant they are of controversial speakers. Specifically, students were asked whether the following five speakers espousing extreme liberal views should be allowed on campus:

1. The Second Amendment should be repealed so that guns can be confiscated.

2. Undocumented immigrants should be given the right to vote.

3. Getting rid of inequality is more important than protecting the so-called "right" to free speech.

4. White people are collectively responsible for structural racism and use it to protect their privilege.

5. Religious liberty is used as an excuse to discriminate against gays and lesbians

Students were also asked whether the following four speakers espousing views extreme conservative views should be allowed on campus:

1. Abortion should be completely illegal.

2. Transgender people have a mental disorder.

3. Black Lives Matter is a hate group.

4. The 2020 Presidential election was stolen.

FIRE’s question asked students to state their opinion on whether these speakers should be allowed on campus independent of whether they personally agreed with the speaker’s message (“Student groups often invite speakers to campus to express their views on a range of topics. Regardless of your own views on the topic, should your school ALLOW or NOT ALLOW a speaker on campus who promotes the following idea?”). In other words, the question was attempting to measure how principled students were with respect to speech on campus.

In the analyses below, I use two indices based on responses to these questions: (1) an index of intolerance for extreme liberal speakers; and (2) an index of intolerance for extreme conservative speakers. The scores of these indices range from 1 (meaning the respondent said every one of the hypothetical speakers should “definitely be allowed to speak”) to 4 (meaning the respondent said every one of the hypothetical speakers should “definitely not be allowed to speak”). By using this data, we can assess whether elite university students are more or less tolerant than students at non-elite universities.

Differences in Tolerance across Universities

In order to examine the political tolerance of elite university students, I first compared scores on the intolerance indices at “elite” and non-elite universities.

Figure 12 – Intolerance for Offensive Speakers at Non-Elite Universities

Figure 12 shows the mean intolerance for extreme liberal speakers (x-axis) and extreme conservative speakers (y-axis) for every respondent in the FIRE data (with each small “x” indicating an individual respondent). The dark black “line of equality” running at a 45-degree angle through the plot areas in Figure XXX indicates where a student would fall if he or she were equally intolerant of all speakers. Students positioned above the “line of equality” are more intolerant of extreme conservative speakers than intolerant of extreme liberal speakers. Students falling below the “line of equality” are more intolerant of extreme liberal speakers than intolerant of extreme conservative speakers.

As Figure 12 shows, there are relatively few “free speech absolutists” (i.e. students who say that all speakers should “definitely be allowed” to give a talk on campus) at American universities today. At the same time, however, there are also relatively few “consistent censors” (i.e. students who say that all speakers should “definitely not be allowed” to give a talk on campus) enrolled in the country’s universities. Most students tend to apply principles of political tolerance in a highly selective fashion.

As Figure 12 also shows, at both elite and non-elite universities, the areas under the “lines of equality” are sparsely populated (indicating there are very few students who are more intolerant of extreme liberal speakers than they are of extreme conservative speakers). In fact, only 8% of students at Top 20 and 14% of students at lower-ranked universities are more tolerant of conservative speakers than liberal speakers. By contrast, the area above the “line of equality” (particularly between 1.5 and 2.5 for extreme liberal speakers and 2.5 and 3.5 for extreme conservative speakers) is densely populated.

Another way of illustrating the relationship between university prestige and political tolerance is to aggregate responses within each university. As Figure 13 shows, there is only one university in the country – Hillsdale College – where students are more intolerant of speakers offensive to conservatives. Every other university in the country is more tolerant of speakers offensive to conservatives than speakers offensive to liberals. As Figure 6 also shows, the differences in tolerance for speakers offensive to conservatives and those offensive to liberals are usually quite large. Among the Top 20 universities, the University of Chicago stands out as being the most tolerant.

Figure 13 – Indices of Intolerance for Offensive Speakers

In order to best describe the role that attendance at elite universities plays in shaping political tolerance, I estimated an OLS regression model that predicts intolerance to extreme liberal and extreme conservative speakers on the basis of an individual’s gender identity, sexual orientation, race, and political ideology as well as each of these variable’s interaction with university ranking. In the figures below, I present the predicted “intolerance scores” associated with these various attributes after controlling for all of the other variables in the model.2 In other words, the graphs presented below depict the unique influence of the characteristic identified in the graph on a student’s “intolerance score” once the influence of other factors has been separated out.

Figure 14 – Indices of Intolerance for Offensive Speakers

Figure 14 shows the predicted “intolerance score” for elite and non-elite university students after controlling for all other influences. As Figure XXX further underscores, students are, on average, far more intolerant of speakers that communicate controversial conservative viewpoints (e.g. “Transgender people have a mental disorder,” “Black Lives Matter is a hate group,” etc.) than they are of speakers who communicate controversial liberal viewpoints (e.g. “White people are collectively responsible for structural racism and use it to protect their privilege,” “Religious liberty is used as an excuse to discriminate against gays and lesbians”).

More importantly, Figure 14 shows that attending an elite university instead of a lower-ranked university is associated with lower intolerance scores for both extreme conservative and extreme liberal speakers. The difference in intolerance scores between students attending Top-20 universities and students attending lower-ranked universities is particularly large for extreme liberal speakers.

It is possible that the higher levels of tolerance associated with attendance at a Top-20 school are a consequence of higher levels of tolerance only among certain subgroups within these universities. For example, it might be the case that elite universities make or attract women who are more politically tolerant than those attending non-elite universities but they actually make or attract men who are far less politically tolerant than those attending non-elite universities. Alternatively, it might be that more tolerant whites are selected for by elite schools but there are no differences in the kinds of non-whites admitted to elite schools. As we will see, what is so striking about the impact of attendance at elite universities is that it is associated with more tolerance among every major subgroup on campus.

Differences in Tolerance across Gender Identities

Gender identity is an important influence on political tolerance. As Figure 15 shows, after controlling for all other factors, men are significantly more tolerant than women and those who identify as other genders (e.g. “Agender,” “Genderqueer,” and “Non-Binary”). Interestingly, those identifying as “other” genders are also significantly more tolerant than women. Put differently, women are more intolerant of both extreme conservative and extreme liberal speakers than any other gender group (independent of race, sexual orientation, and political ideology).

Figure 15 - Intolerance by Gender and University

Attendance at an elite university is associated with less intolerance across all gender identities. As Figure 15 shows, students of all genders at elite universities are significantly less intolerant than their counterparts at non-elite universities. For example, men at Top 20 universities are more tolerant of extreme liberal speakers (mean intolerance = 1.8) than men at lower-ranked universities (mean intolerance =2.1). The same can be said of the mean intolerance levels of women (2.3 to 2.0) and “other” gender identity students (3.0 to 2.8). A similar pattern of intolerance exists with respect to extreme conservative speakers. Regardless of gender identity, in other words, students at elite universities are more tolerant than those at non-elite universities.

Differences in Tolerance across Sexual Orientations

Receiving an elite university education is also correlated with less intolerance across all sexual orientations. As Figure 16 shows, attendance at a Top-20 university is associated with significant decreases in intolerance among straight students and among lesbian, gay, bisexual, and students with “other” sexual orientations. Once again, the effect exists for both extreme liberal and extreme conservative students.

Figure 16 - Intolerance by Sexual Orientation and University

It is also important to point out here that straight students (regardless of the kind of university they attend) are more tolerant of extreme conservative students than lesbian, gay, bisexual, and “other” sexual orientation students. By contrast, lesbian, gay, bisexual, and “other” sexual orientation students (regardless of university prestige) are more intolerant of extreme conservative speakers. As a reminder, these differences are not a function of confounding with gender, race, or political ideology. Straight students are more tolerant of conservatives while non-straight students are more tolerant of liberals.

Differences in Tolerance across Race

Racial differences in tolerance are generally less pronounced than differences based on gender and sexual orientation. As Figure 17 shows, however, there is still an important effect on tolerance scores associated with attendance at an elite university among all racial groups. With the lone exception of African American students and tolerance for extreme conservative speakers, attending a Top-20 school is predicted to significantly decrease intolerance among students of every racial group for both conservative and liberal speakers.

Figure 17 - Intolerance by Race and University

Differences in Tolerance across Political Ideologies

Previous research has shown that political ideology is one of the largest influences on political tolerance. Typically, people are less willing to censor or punish speakers from their specific side of the political spectrum.

Figure 18 - Intolerance by Political Ideology and University

As Figure 18 shows, this is certainly true of students on the political left.3 Whether they attend Top-20 or lower-ranked universities, left-leaning students are far more intolerant of conservative speakers than of liberal speakers.

What is most surprising about Figure 18, however, is that right-leaning students at elite universities do not seem to discriminate based on ideology. For conservatives and libertarians at Top-20 schools there are no differences in predicted tolerance scores between extreme conservative speakers (tolerance score = 2.1) and extreme liberal speakers (tolerance score = 2.0). The same cannot be said of right-leaning students at lower-ranked universities. Much like left-leaning students, these students are more tolerant of those from their side of the political spectrum.

Figure 18 also illustrates the general pattern discussed above; namely, that attendance at an elite school reduces intolerance scores among all subgroups. Those on the left, those on the right, and those in the middle who are enrolled at a Top-20 university are less intolerant than their counterparts at lower-ranked universities.

Are students at elite universities supportive of disruptive action in response to constitutionally protected speech?

Measuring Support for Disruptive Conduct

The FIRE survey also asked respondents “How acceptable [always, sometimes, rarely, never] would you say it is for students to engage in the following action to protest a campus speaker?”:

1. Shouting down a speaker to prevent them from speaking on campus.

2. Blocking other students from attending a campus speech.

3. Using violence to stop a campus speech.

These questions provide a useful proxy for measuring how strongly students subscribe to norms of political tolerance in response to offensive (yet constitutionally protected) speech.

Differences in Support for Disruption across Universities

As Figure 19 shows, support for free speech norms varies significantly between elite and non-elite universities.

Figure 19 – Acceptability of Disruptive Conduct in Elite and Non-Elite Universities (Overall)

There is a 9% to 15% difference in support for each of the disruptive actions measured in the FIRE data between elite and non-elite universities, with students at elite schools significantly more supportive of violence, blockades, and shout-downs. 

In this section, I rely on a multivariate logistic regression analysis to predict the probability that a student will view shout-downs, blockades, and violence as “always,” “sometimes,” or “rarely” acceptable. As in the previous section, this analysis includes measures of gender identity, sexual orientation, race, political ideology, and school ranking.

Once again, the differences between elite and non-elite universities illustrated in Figure 19 hold across almost every demographic group. Below, I show how minority groups are more supportive of disruption overall and, importantly, more supportive in elite universities.

Differences in Support for Disruption across Gender Identities

Let’s first consider the differences in support for disruptive action in response to offensive campus speech based on self-reported gender identity. Figure 1 compares the percentage of male, female, and “other” gendered students (“agender,” “genderqueer,” “non-binary” and students who were unsure about their gender identity) who say that shout-downs, blockades, and violence are “always,” “sometimes,” or “rarely” acceptable.

Figure 20 – Acceptability of Disruptive Conduct in Elite and Non-Elite Universities (Gender)

As Figure 20suggests, there is widespread support for disruptive action in response to an offensive campus speech among “other” gender-identifying students. After controlling for race, sexual orientation, political ideology, and university ranking, the predicted probability that an “agender,” “genderqueer,” “non-binary” or “unsure” student will endorse shout-downs is 70.7%. The predicted probability that an “other” gender student will endorse blockades is 52.2%. The predicted probability they will endorse “using violence” is 37.1%. By contrast, the predicted probabilities of supporting these actions among men is 61.4%, 36.1%, and 21.8%, respectively.

Figure 20 also shows the differences between students at elite and non-elite universities. Men at elite universities are significantly more likely to endorse blockades and violence than men at non-elite universities. Female and “other” gender-identifying students at Top 20 schools are far more supportive of every disruptive action than their counterparts at lower-ranked universities. For example, while there is a 45.9% predicted probability of a woman at Top 20 university saying a blockade is acceptable, there’s only a 36.7% chance of a woman at a non-Top 20 university endorsing blockades. In other words, the nation’s most highly ranked universities have student populations that, regardless of gender, are far more supportive of disruptive action than the nation’s other universities.

Differences in Support for Disruption across Sexual Orientations

Across all universities, there are significant differences based on sexual orientation, with students identifying as lesbian, gay, bisexual, or “other,” expressing far more support for disruption than students identifying as heterosexual. To be exact, lesbian, gay, bisexual, and “other” students were 14% more likely to support violence (31% to 17%), 19% more likely to support blockades (51% to 32%), and 21% more likely to support shout-downs (78% to 57%) than straight students.

Figure 21 – Acceptability of Disruptive Conduct in Elite and Non-Elite Universities (Sexual Orientation)

Once again, these overall differences mask significant variation based on the kind of university a student attends. Regardless of sexual orientation, students at Top 20 universities are far more supportive of disruptive action in response to a campus speaker than students are lower ranked universities (Figure 3). As indicated in Figure 3, straight students at elite schools are significantly more likely to endorse shout-downs, blockades, and violence than straight students at non-elite universities. Gay, lesbian, bisexual, and “other” sexual orientation students at Top 20 schools were also more accepting of disruption than similar students attending schools outside of the Top 20.

Differences in Support for Disruption across Race

In the overall student population, differences in support for disruption based on race are somewhat smaller than differences based on gender identity or sexual orientation.

Figure 22 – Acceptability of Disruptive Conduct in Elite and Non-Elite Universities (Race)

There are no meaningful differences in how students of different races view “shouting down a speaker to prevent them from speaking on campus.” There are, however, large differences in perceptions of the acceptability of blockades and violence. As Figure 22 shows, non-white students are consistently more supportive of blockades and violence in response to offensive campus speakers. Notably, Asian (25.7%), Hispanic (25.1%), Black (24.3%), and students reporting “other” racial identities (26.9%), were all significantly more likely than white students (17.1%) to say that violence is “always,” “sometimes” or “rarely” acceptable.

As Figure 22 shows, elite university students are again more supportive of disruption than non-elite university students who share their race (with the one small exception of Asian American students on the question of using violence). Particularly striking are the differences between Hispanics at elite and non-elite universities. Hispanics at Top 20 universities were 21% more likely to support blockades and 17% more likely to support violence than Hispanics at lower-ranked universities.

Differences in Support for Disruption across Political Ideology

Ideology also appears to play a role in shaping how students view violent responses to campus speech. Left-leaning students (i.e. those identifying as “Democratic Socialists” and “Liberals”) are more accepting of “using violence” than moderate and right-leaning students (i.e. those identifying as “Conservative” and “Libertarian”) independent of the kind of university they attend.

Figure 23 – Acceptability of Disruptive Conduct in Elite and Non-Elite Universities (Political Ideology)

As Figure 23 shows, however, Top 20 university students are frequently more supportive of disruption than their counterparts at lower-ranked universities.

Overall - America’s Elite College Students are More Supportive of Disruption than Their Counterparts at Non-Elite Universities

The above analyses paint a complex picture of political tolerance at American universities. We might expect that the two sets of measures included in the FIRE surveys (one about allowing a speaker on campus and the other about supporting disruptive action) produce similar results about political tolerance across different kinds of universities. More directly, we should find that students that believe a speaker should be allowed to speak on campus should also reject disruptive action in response to that speech. This is not, however, what the FIRE data show. Students at elite universities are both more tolerant of offensive speakers and more supportive of disruptive action in response to speech than students at non-elite universities.

It is tempting to attribute these differences to the persuasive effects of faculty. It is important to point out here, however, that students - at both elite and non-elite universities - are actually far more intolerant of offensive speakers and far more supportive of disruptive action than faculty. In addition to their 2022 student survey, FIRE also conducted a survey of faculty members at public and private four-year degree-granting universities in the United States. While the faculty survey does not have a sufficiently large sample (N=1,491) to compare faculty at Top 20 universities to faculty at other universities, it does allow us to compare faculty responses on the FIRE tolerance and disruption questions overall to student responses.

Figure 24 - Tolerance of Offensive Speakers among Students and Faculty

Figure 24 shows the faculty, elite university, and non-elite university comparisons for four of the political tolerance questions.4 As Figure 24 shows, students at elite universities were more tolerant than students at non-elite universities and faculty were more tolerant than students of all four hypothetical speakers. Support among students and faculty was highest for a speaker espousing the view that “structural racism maintains inequality by protecting White privilege.” Support among students and faculty was lowest for a speaker suggesting that “Black Lives Matter is a hate group.” The main idea here, however, is that faculty are much more tolerant than students.

Figure 25 - Support for Disruptive Action among Students and Faculty

Figure 25 shows the faculty, elite university, and non-elite university comparisons for the questions on disruptive action. Compared to students at both elite and non-elite universities, faculty were significantly less accepting of de-platforming speakers using illiberal tactics. What jumps out from Figure 25 the most, however, is the massive gulf between faculty and students at elite universities. Students at elite universities were 15% less likely than faculty to say shout-downs are “never acceptable,” 31% less likely than faculty to say that blockades are “never acceptable,” and 24% less likely to say that “using violence” is “never acceptable.”

Magnets or Mints?

What can explain the fact that elite university students are more liberal, more tolerant, and more disruptive than their peers and the rest of American society? Although the question of whether elite universities are “magnets” (i.e. places that attract more liberal, more tolerant, and more disruptive individuals) or “mints” (i.e. places that create more liberal, more tolerant, and more disruptive individuals) is one of the most interesting and politically salient features of debates about higher education, it cannot be easily answered with the existing data. In particular, the analysis here (showing significant ideological and political tolerance differences across different kinds of universities) should not be used to imply a “brainwashing” (or, less pejoratively, a “socialization”) effect among elite university students.

Nevertheless, it is worth saying a few words about these two approaches (“magnets” or “mints”) to explaining the cross-university differences in ideology and tolerance described above. The “magnet” idea suggests that elite universities simply recruit, attract and enroll more liberal, tolerant, and disruptive students. To take just one example of this broader class of explanation, elite university students might display more tolerance for offensive speakers because they are, on average, more intelligent (and more intelligent people tend to be more tolerant). According to one recent study, for instance, “Cognitive ability was the single strongest predictor of political tolerance, with larger effects than education, openness to experience, ideology, and threat.” Far more work on this particular relationship is needed but it does illustrate the idea that elite universities might be magnets (i.e. places that attract more liberal, more tolerant, and more disruptive individuals) rather than mints (i.e. places that create more liberal, more tolerant, and more disruptive individuals)?

Alternatively, elite universities might be “minting” students that are more liberal, tolerant, and disruptive (or non-elite universities might be “minting” students that are less liberal, tolerant, and disruptive). The idea here is that something about the university environments (e.g. the curriculum, the political views expressed by faculty, the interactions between students, the incentives created by university administrators) transforms students during their four years at college. Faculty attempts to explicitly propagandize their students and DEI efforts aimed at curtailing “hate speech” have been the two most commonly cited concerns along these lines but there are many other flavors of this argument.

Independent of whether the disparity between elite and non-elite universities is closer to the “magnet” or “mint” model, however, the disparity remains. If our institutions continue to disproportionately represent elite university alumni, our institutions will become more liberal, more tolerant, and more supportive of disruptive action as a response to offensive speech.

Addendum: Support for Disruption by Major

The FIRE survey also asks each respondent about their major. In order to assess whether differences in support for disruptive action vary across majors, I lumped each major into one of eight broad categories: Education, Business, Humanities, the Visual and Performing Arts, Social Sciences, STEM, Ethnic/Gender/Area Studies and “Other” (which includes those who are “Undecided”).

As these graphs show, there are clear differences across majors, with students majoring in Ethnic, Gender, and Area studies far more supportive of disruption than students majoring in other areas.

Thank you to Project Luminas for this suggestion!

Here, “diversity” is defined (partly) as a commitment to representing a variety of “political perspectives” and “inclusion” is defined as a commitment to making sure these “different perspectives are respectfully heard.” When defined in this way, DEI sounds like a bedrock institutional commitment and operational blueprint for protecting free speech on campus. It would not be unreasonable, therefore, to expect that the increasing size and significance of university DEI bureaucracies might significantly improve the speech climates on college campuses.

Yet, the rise of DEI bureaucracies has actually coincided with the beginning of a “Free-Speech Crisis on College Campuses.” Careful observers of American higher education saw the tension between DEI and free speech early on. Most notably, in a 2016 lecture, Jonathan Haidt pointed out that universities were now attempting to simultaneously pursue “two incompatible sacred values”: truth and social justice. Haidt argued for a schism in higher education, with universities explicitly adopting either a John Stuart Mill-style commitment to the pursuit of truth through unfettered speech or a Karl Marx-ian commitment to the pursuit of “social justice” (even if it occurred at the expense of free expression). Although Haidt did not mention them explicitly, DEI bureaucracies were clearly implicated in his discussion (as they had become the primary institutional vehicles for pursuing the “social justice” values of “diversity” and “equity”). In other words, it was clear from the start that, regardless of what was on their websites, DEI bureaucracies were more likely to suppress than encourage free expression on college campuses.  

More recent rhetoric about DEI’s negative impact on free expression is far more direct. In a 2021 Newsweek editorial, for example, Professors Dorian Abbot and Ivan Marinovic argued that:

“American universities are undergoing a profound transformation that threatens to derail their primary mission: the production and dissemination of knowledge. The new regime is titled "Diversity, Equity, and Inclusion" or DEI, and is enforced by a large bureaucracy of administrators. Nearly every decision taken on campus, from admissions, to faculty hiring, to course content, to teaching methods, is made through the lens of DEI.”

Striking a similar tone in a Wall Street Journal op-ed (entitled “How ‘Diversity’ Turned Tyrannical),” Lawrence Krauss wrote that DEI had created “a climate of pervasive fear on campus” that was shutting down vitally important discussions. According to Krauss, “The DEI monomania has contributed to the crisis of free speech on campus.”

In a more recent piece from 2023, entitled “How DEI Is Supplanting Truth as the Mission of American Universities,” John Sailer claims that “the concepts of DEI have become guiding principles in higher education, valued as equal to or even more important than the basic function of the university: the rigorous pursuit of truth.” According to Sailer, “the entire experience of higher education—from earning a college degree to seeking a career in academia—now requires saturation in the principles of DEI.” As Sailer points out, some DEI offices even now warn against “weaponizing academic freedom.” In short, DEI and free expression have now become the antithesis of each other.

Summarizing these perspectives succinctly, Colin Wright claimed that “DEI departments, statements, and initiatives must be abolished at every school and university that receives public funding. They are costly cancers that stifle free speech and viewpoint diversity. Reform is not possible without first abolishing DEI.”

Are university DEI bureaucracies really “cancers that stifle free speech and viewpoint diversity”? Surprisingly, there have been no empirical studies of these claims to date. Without empirical evidence, we cannot assess whether DEI bureaucracies have any impact on university speech climates at all. Perhaps more importantly, without empirical evidence, we cannot be precise about exactly how DEI programs are shaping free expression on college campuses. It is possible, for example, that DEI programs degrade speech environments by creating an oppressive atmosphere where students from all over the political spectrum feel equally afraid to speak for fear of bureaucratic retribution. Of course, it is also possible that DEI programs have an asymmetrical effect on speech from the right and left, with conservatives feeling their views are unfairly demonized and progressives feeling their views are tacitly endorsed by the university administration. It is also possible that DEI’s effects are found only on expression in the classroom but not outside of it. Alternatively, the opposite (i.e. DEI’s effects are not felt in the classroom but are felt outside of it) could be true. Without data, we just can’t know.

Correctly diagnosing the effects (if any) of DEI programs is a necessary condition for devising effective and appropriately targeted solutions. In the absence of empirical evidence, however, we are left only with unpersuasive rhetoric that is far more likely to obscure the truth about DEI programs than it is to reveal anything useful about them.   

As I demonstrate below, more DEI personnel are, in fact, associated with notably worse attitudes toward free expression among college students. As I also show, however, DEI programs do not always matter in the ways that their critics assume. Specifically, using a combination of FIRE’s 2022 survey and the Heritage Foundation’s 2021 report “Diversity University: DEI Bloat in the Academy,” I find that larger DEI bureaucracies: (1) increase student tolerance for liberal speakers while exerting no effect on tolerance for conservative speakers; (2) raise levels of discomfort related to self-expression outside of the classroom but have little impact on self-expression inside of it; (3) encourage marginally more support for attempts to disrupt campus speech through violence, blockages, and shout-downs; and (4) leave students feeling that “open and honest” conversations are difficult on a slightly larger group of political issues.

Overall, then, there is a significant amount of evidence that DEI is bad for campus speech climates and an even greater amount of evidence that DEI does very little to promote viewpoint inclusion.

Measuring DEI in Universities: The Heritage Foundation’s “Diversity University” Data

The first (and only) cross-campus study of DEI bureaucracies was produced by Jay Greene and James Paul for the Heritage Foundation in 2021. The study, summarized in a report entitled “Diversity University: DEI Bloat in the Academy,” examined 65 universities from the “Power Five” athletic conferences: the Atlantic Coast Conference, the Big 10, the Big 12, the PAC 12, and the Southeastern Conference. According to Greene and James, the “Power Five universities tend to be mainstream institutions that students select—and state legislatures support—without much thought to their political and cultural aims.” Importantly, these 65 universities serve approximately 16% of all students enrolled in four-year universities across the United States (collectively amounting to 2.2 million students). As such, the results of this study paint a broad picture of the size and scope of DEI bureaucracies in American universities.

In order to identify the number of DEI employees working at each of these 65 schools, Greene and Paul relied on keyword searches of university websites. Specifically, they searched for terms such as “diversity,” “Multicultural Affairs,” “African American Culture,” “Asian Culture,” “Latino Culture,” “Native American Culture,” “Women’s Center,” and “LGBTQ Center.” Staff identified through these searches were added to their university’s count of DEI personnel. In an effort to “capture only the effort that these institutions want to devote to DEI, rather than what they must devote to DEI,” employees with primary responsibilities connected to Title IX, equal employment opportunity, or compliance with other legal obligations were excluded from the total count.

As the authors point out, the results produced from this study are likely an underestimate of the true size of university DEI bureaucracies. Not every DEI employee is listed on a university’s website, many employees have partial DEI responsibilities that may not be reflected in their formal titles and the list of search terms may have missed DEI personnel who are not associated with the narrow topics queried.

Even still, Green and Paul’s search procedures identified nearly 3,000 people tasked with promoting diversity, equity, and inclusion at the 65 universities of the “Power 5” conferences (for an average of 45.1 people per school). Of course, some universities have many more DEI personnel than others. The University of Michigan, for example, has 163 people working in DEI programming. Ohio State University and the University of Virginia each have 94. As Greene and Paul conclude, “Promoting DEI has become a primary function of higher education.”

The 65 schools studied by Greene and Paul were selected because they are “mainstream institutions.” Looking only at these schools, however, prevents us from analyzing how DEI may operate at more elite institutions. These elite institutions are important because: (1) they produce a disproportionate share of future leaders; (2) they serve as role models for lower-ranked universities (who might copy or mimic their policy approaches); and (3) they shape a disproportionate share of the public discourse related to higher education.

For these reasons, I decided to expand on Greene and Paul’s data by applying their template to the eight schools of the Ivy League: Harvard, Brown, Cornell, Columbia, Dartmouth, Princeton, Yale, and the University of Pennsylvania. Figures 1 and 2 show how these Ivy League schools compare to the 25 largest DEI bureaucracies in the “Power Five” conference universities.1

Figure 1 - DEI Personnel at “Power 5” Universities

Figure 2 - DEI Personnel at Ivy League Universities

Collectively, this data on university DEI bureaucracies reveals several patterns. First, private universities, on average, have relatively larger DEI staff than public universities. 14 of the top 20 universities ranked by DEI personnel per 1,000 students are private while 39 of the lowest 40 universities are public. Overall, private universities have an average of 4.5 DEI personnel per 1,000 students. By contrast, publicly funded universities have only 1.8 DEI personnel per 1,000 students.  

Second, increased racial and ethnic diversity among the student body is associated with larger DEI bureaucracies. Figure 3 shows the relationship between US News and World Report’s college diversity index (where scores range from 0 to 1 and higher scores indicate more diversity in the student population) and DEI personnel per 1,000 students. Higher levels of racial and ethnic diversity are highly correlated with more DEI personnel (r=.47, p=.00). The fact that more diverse campuses have more DEI employees might be a result of diverse campuses bulking up their DEI bureaucracies in an attempt to address the perceived needs of their students or a result of self-selection by non-white students into universities with more DEI services. The existing data does not let us sort these competing explanations out. For now, however, it is enough to point out that more racial and ethnic diversity tends to go along with more diversity, equity, and inclusion staff.

Figure 3 - DEI Personnel and Racial and Ethnic Diversity

Finally, the political ideology of the student body is related to how large a university’s DEI bureaucracy is. As Figure 4 shows, universities with more liberal students (measured through the FIRE survey discussed below) have more DEI personnel (r=.54, p=.00). Once again, this relationship may be a function of liberal students demanding more DEI (and universities responding) or a function of liberal students self-selecting into universities with larger DEI bureaucracies.  

Figure 4 – DEI Personnel and Student Ideology

Keeping these relationships in mind will be essential for thinking about how DEI bureaucracies shape campus speech climates. If we observe a relationship between DEI bureaucracies and speech, we will need to remember that it might be a consequence of the fact that private universities are less tolerant of speech and are, also, more likely to have bigger DEI bureaucracies. Similarly, if we find that large DEI bureaucracies make open and honest discussions of race more difficult, this could simply be a function of the fact that large DEI bureaucracies are more common at racially diverse universities. To state this more formally, we are going to need to control for these attributes of universities (and much more) to get an unbiased sense of whether DEI is helping or hurting free expression on college campuses.

Measuring Free Speech Attitudes among College Students: The FIRE Data

Each year since 2020, the Foundation for Individual Rights and Expression (FIRE) has conducted a vast, nationwide survey of college students as part of their efforts to rank the free speech climates of American universities. The 2022 survey included 45,000 enrolled students at over 200 colleges around the country. In addition to a battery of demographic (race, gender, sexual orientation, etc.) and political identity questions (e.g. political ideology, partisanship, etc.), the FIRE survey asked respondents about five, broad dimensions of free speech on campus:

1. Tolerance for allowing controversial speakers on campus

2. Comfort expressing ideas in the classroom

3. Comfort expressing ideas outside of the classroom

4. Support for disruptive action in order to prevent offensive speech

5. Openness to discussing challenging topics on campus

Using FIRE’s data on these five dimensions, we can provide the first-ever empirical assessment of the relationship between DEI bureaucracies and student attitudes about free expression on campus. Specifically, we can finally determine whether universities with more DEI personnel actually produce more or less “inclusive” speech climates.

In the discussion that follows, I will present the results of bivariate (i.e. the correlations between the number of DEI personnel at a university and student responses to FIRE’s questions on the five dimensions of speech) and multivariate analyses (i.e. regressions that explain variation in student responses to FIRE’s questions after controlling for a wide range of university-level variables, including the number of DEI personnel).

The bivariate correlations should be interpreted with caution. The most significant problem with bivariate analyses of any kind is that they cannot account for confounding influences. As discussed above, DEI bureaucracies are larger at private, liberal, and more racially diverse universities. If we find that DEI bureaucracies are associated with students having more negative assessments of their campus speech environments, we can’t be sure whether this is a consequence of DEI bureaucracies themselves or just spuriousness.  

In order to address this problem, I will also present multivariate models that explain variation in the five dimensions of campus speech. Each of these models controls for: (1) whether the university is public or private; (2) the racial and ethnic diversity of the university (measured through US News and World Report’s diversity index); (3) the average liberal-conservative political ideology of the university’s student population (measured through a self-reported question on FIRE’s 2022 survey); (4) the average perceived ideology of the university’s faculty (measured through a question asking students to place the “average” faculty member on a liberal to conservative continuum on FIRE’s 2022 survey); (5) the extent to which the university’s formal policies comply with First Amendment standards (measured through FIRE’s speech code ratings system); and (6) the size of the university’s undergraduate population.

The problem with these multivariate models, however, is the limited sample size of universities. Currently, there are only 71 universities where we have both DEI and free speech survey data. As a result, we can’t be sure whether a statistically insignificant result is a consequence of a true null effect or a lack of statistical power.

With these caveats in mind, let’s explore the data.

The Relationship between DEI and Tolerance of Controversial Speakers

The FIRE survey asks students a series of questions to assess how tolerant they are of controversial speakers. Students were asked whether the following five speakers espousing views potentially offensive to conservatives should be allowed on campus:

1. The Second Amendment should be repealed so that guns can be confiscated.

2. Undocumented immigrants should be given the right to vote.

3. Getting rid of inequality is more important than protecting the so-called "right" to free speech.

4. White people are collectively responsible for structural racism and use it to protect their privilege.

5. Religious liberty is used as an excuse to discriminate against gays and lesbians

Students were also asked whether the following four speakers espousing views potentially offensive to liberals should be allowed on campus:

1. Abortion should be completely illegal.

2. Transgender people have a mental disorder.

3. Black Lives Matter is a hate group.

4. The 2020 Presidential election was stolen.

FIRE’s question asked students to state their opinion on whether these speakers should be allowed on campus independent of whether they personally agreed with the speaker’s message (“Student groups often invite speakers to campus to express their views on a range of topics. Regardless of your own views on the topic, should your school ALLOW or NOT ALLOW a speaker on campus who promotes the following idea?”). In other words, the question was attempting to measure how principled students were with respect to speech on campus.

The FIRE report summarizing the survey’s results shows wide variation in tolerance for these hypothetical speakers:

Figure 5 – FIRE’s Summary of Tolerance for Controversial Speakers

The fact that the four least tolerated speakers espouse views offensive to liberals should not be surprising. Generally, people are far more intolerant of speakers and speech that challenge what they believe to be true and college students are overwhelmingly liberal. Specifically, in the 2022 FIRE survey, an average of 49.5% of students at the 71 universities included in this analysis self-identified as liberal (compared to 23.3% that identified as conservative). Perhaps more importantly, liberal students were the plurality in 64 of the 71 universities (with conservatives constituting the plurality only at Kansas State University, the University of Mississippi, Mississippi State University, Auburn University, the University of Tennessee, Clemson University, and Baylor University). As Figure 6 shows, liberal students outnumber conservative students by more than 40% (e.g. 60% to 20%) on 22 of the 71 campuses examined here while there is not a single university where conservative students outnumber liberal students by more than 20%. 

Figure 6 – Percent of Students Identifying as Liberal by University

The impact of ideology on tolerance is clear from examining the FIRE survey data at the individual level. As Figure 7 shows, liberals and conservatives are less intolerant of speakers who share their ideological predispositions. To be precise, liberal college students had an “average intolerance” score of 2.0 (where 1=the speaker should “definitely” be allowed and 4=the speaker should “definitely not” be allowed) when considering whether to allow liberal speakers on campus and conservative college students had an “average intolerance” score of 2.3 when considering whether to allow conservative speakers on campus. By contrast, liberal students had an “average intolerance” score of 3.4 when considering whether to allow conservative speakers and conservative students had an “average intolerance” score of 2.5 when considering whether to allow liberal speakers. Despite FIRE’s attempt to prime them to ignore their “own views on the topic,” in other words, students (particularly liberal students) are intolerant of those who contradict their beliefs.

Figure 7 – Intolerance of Controversial Speakers by Ideology

Even with this context in mind, the extent to which there is more tolerance of liberal speakers among the 71 universities studied here is striking. On every single one of the 71 campuses, there was more tolerance, on average, for liberal speakers than conservative speakers. Even at Mississippi State University and Clemson University (where conservatives outnumber liberals by nearly 20%), the average level of tolerance for liberal speakers was still greater than that of conservative speakers.

Figure 8 – Net Intolerance Toward Liberal and Conservative Speakers

Figure 8 gives us a broad sense of variation in tolerance across universities but it doesn’t help us assess whether relatively larger DEI bureaucracies increase or decrease intolerance. We can start to answer this question by looking at the average level of intolerance (i.e. the mean intolerance when left-wing and right-wing speakers are lumped together) among the 71 universities that we have both FIRE and Heritage data for. As Figure 9 shows, more DEI personnel is generally associated with less intolerance/more tolerance of controversial speakers (r=-.50, p=.00).

Figure 9 – Intolerance of Controversial Speakers

Does this mean that DEI bureaucracies are actually creating more “inclusive” speech climates for everyone? Not so fast. As Figure 10 shows, the overall reduction of intolerance is entirely a function of how DEI personnel increase tolerance for liberal speakers. Students at universities with more DEI personnel per 1,000 students are far more tolerant of liberal speakers (r=-.65, p=.00). Students at universities with more DEI personnel per student are not, however, more tolerant of conservative speakers (r=.14, p=.22).

Figure 10 – Intolerance of Liberal and Conservative Speakers

As Table 1 shows, the effect of DEI personnel is robust for intolerance of liberal speakers after adding statistical controls. Table 1 also confirms that the number of DEI personnel makes no impact on how much tolerance students will express for conservative speakers.

Table 1 – Intolerance of Liberal and Conservative Speakers

It is important to emphasize here that there is no evidence in the data that larger DEI bureaucracies increase intolerance for controversial conservative speakers. This point is an essential rejoinder to critiques of DEI that focus on the suppression of speech coming from the political right. Most of the energy behind free speech debates over the last few years has come from concerns that conservative speakers are being unfairly sanctioned for espousing their views. The typical narrative about “cancel culture” on college campuses before 2020 was that timid and ineffectual administrators were being pressured by a small group of left-wing malcontents (i.e. “social justice warriors”) into suppressing the communication of any ideas from the political right. As opposition to DEI has ramped up over the last two years, this old narrative has been supplanted with the idea that DEI bureaucracies are the primary causes of the deteriorating speech climate on college campuses. While conservatives remain the targets of censorship efforts in these accounts, explicitly hostile DEI personnel (rather than the combination of effete administrators and disgruntled leftist youth) have become the main villains of the story.

Contrary to this newly mobilized narrative, the data show that DEI bureaucracies actually work in favor of progressive ideas not by shrinking the Overton Window on the right but, instead, by expanding it on the left. To state this in more concrete terms, DEI bureaucracies appear to shape the campus speech climate not by stigmatizing ideas like “Black Lives Matter is a hate group” and “Transgender people have a mental disorder” but, instead, by normalizing ideas such as “Religious liberty is used as an excuse to discriminate against gays and lesbians” and “White people are collectively responsible for structural racism and use it to protect their privilege.” Overall, then, the “inclusion” in “Diversity, Equity and Inclusion” appears to mean more tolerance of left-wing speakers and not more tolerance of right-wing speakers.

The Relationship between DEI and Discomfort in Classroom

Let’s turn now to how DEI bureaucracies might shape the classroom environment. The good news for those who genuinely worry about its deleterious effects on education is that there’s not an immediately obvious relationship between having a relatively large DEI bureaucracy and having a more uncomfortable classroom environment. As Figure 11 shows, there is essentially no difference in how comfortable students are voicing an opinion during an in-class discussion or disagreeing with a professor (in public or in a written assignment) based on how many DEI personnel are working at a university (r=.01, p=.93). Figure 11, then, also contains some bad news for DEI supporters; namely, that more personnel ostensibly attempting to increase “inclusion” does not produce more classroom comfort for students.

Figure 11 – Discomfort Expressing Opinions in Class

The addition of controls for the campus’s racial diversity, the ideology of the student body, the ideology of the faculty, the severity of the campus’s speech codes, and the size and type (public or private) of the university does not change the story (Table 2). Based on this evidence, it appears that DEI bureaucracies have not (at least up until this point) significantly shaped the way students feel about expressing themselves in the classroom.

Table 2 – Discomfort Expressing Opinions in the Classroom

Interestingly, all of the other, non-DEI variables included in Table 2 were also uncorrelated (in either bivariate or multivariate analyses) with classroom discomfort across universities. This suggests that something not measured in this analysis might dictate how comfortable students feel in the classroom.

The most likely candidate is cross-campus variation in how earnestly professors attempt to create welcoming classroom environments. If most faculty members at a university make concerted efforts to conceal their political opinions or present competing perspectives in an even-handed fashion, it would not be surprising to find that most students are comfortable in their classrooms. By contrast, if professors aggressively proselytize their political values, many students will likely feel uncomfortable in the classroom. Unfortunately, there is no measure of faculty teaching methods (in the FIRE data or elsewhere) that might help us account for the heretofore unexplained variation in in-class comfort across universities revealed by the FIRE data. In other words, all we can conclude from the data we currently have is that larger DEI bureaucracies do not appear to negatively or positively impact the feelings of comfort students have in the classroom.

The Relationship between DEI and Discomfort Outside of the Classroom

The same cannot be said, however, for how students view the speech climate of their campuses once they leave the classroom. As Figure 12 shows, the greater the relative size of the DEI bureaucracy at a university, the more discomfort students feel expressing their views on social media and in informal conversations with other students in the campus “quad, dining hall, or lounge” (r=.49, p=.00).2 

Figure 12 – Discomfort Expressing Opinions Outside of the Classroom

Importantly, the relationship shown in Figure 12 does not appear to be spurious.3 As Table 3 shows, the size of the DEI bureaucracy remains positively correlated with feelings of discomfort outside of the classroom even after adding the necessary group of statistical controls.

Table 3 – Discomfort Expressing Opinions Outside of the Classroom

The Relationship between DEI and Disruption

The FIRE survey asked students how acceptable it is to engage in three different methods of protest against an offensive campus speaker: “Shouting down a speaker to prevent them from speaking on campus,” “Blocking other students from attending a campus speech,” and “Using violence to stop a campus speech.” Response options ranged from “always acceptable” to “never acceptable.” In the below analyses, higher scores indicate higher levels of acceptability for disruptive behavior.

Figure 13 shows the relationship between an index of responses to FIRE’s three “disruptive” conduct questions and DEI personnel per 1,000 students. The graph suggests a strong and positive relationship between the size of a university’s DEI bureaucracy and the acceptability of disruptive conduct to prevent a controversial speech from taking place on campus, with more DEI personnel being correlated with a stronger willingness to accept disruptive behavior (r=.55, p=.00). 

Figure 13 - Acceptability of Disruption to Prevent a Speech

As Table 4 shows, however, the relationship between DEI personnel per 1,000 students and acceptability of disruptive behavior falls just short of traditional levels of statistical significance after controls are added. This is likely due to the fact that student ideology and the size of a university’s DEI bureaucracy are so highly correlated with each other (see above) and with support for disruption. Indeed, as Figure 14 shows, universities with more liberal students are significantly more likely to endorse disruption (r=.74, p=.00).

Table 4 – Acceptability of Disrupting a Campus Speech

Figure 14 - Acceptability of Disruption to Prevent a Speech (Student Ideology)

Unfortunately, FIRE’s questions ask about support for disruption in the abstract, undetached from a particular trigger or target. As shown above, however, the content of the speech and the identity of the speaker needs to be considered if we are hoping to get a clear sense of how much students truly support disruption. For example, if FIRE had asked students whether they support “using violence” to prevent a speech arguing “transgender people have a mental disorder,” the results would likely look much different than if they had asked students whether they support “using violence” to prevent a speech arguing “Religious liberty is used as an excuse to discriminate against gays and lesbians.” With more nuanced measures such as these, it is possible that we would discover a statistically significant relationship between DEI bureaucracies and the acceptability of disruption that parallels the relationship we identified between DEI bureaucracies and intolerance for campus speakers (i.e. universities with more DEI personnel are more likely to express tolerance for and disruption on behalf of speakers espousing progressive views).

This is speculation, however. What is not speculation is that there is a strong, bivariate correlation between the size of a university’s DEI bureaucracy and support for disruption in response to offensive speech. The fact that this relationship disappears once student ideology is controlled for might be a function of this study’s limited sample size (N=71) or might indicate that there is truly no relationship between DEI and support for disruption. The only thing we can say with confidence is that DEI probably does not reduce support for disruption and may slightly increase it.

The Relationship between DEI and Honest Conversations

The FIRE survey asked students a simple “yes” or “no” question about whether it was “difficult to have an open and honest conversation” about 17 high-profile and potentially divisive political issues (including abortion, gender inequality, racial inequality, freedom of speech, gun control, transgender issues, affirmative action, vaccine mandates, and gun control).

There are many ways of analyzing whether responses to this question correlate with the size of a university’s DEI bureaucracy. The most straightforward approach is to explore whether the average number of topics students at different universities report “difficulty” discussing is related to the number of DEI personnel a university employs. Figure 15 presents the results of this analysis.

Figure 15 – Number of Topics that are Difficult to Discuss Openly and Honestly

As Figure 15 shows, the size of a university’s DEI bureaucracy is positively correlated with the overall number of topics students find it difficult to discuss “openly and honestly” (r=.36, p=.00). Students seem, in other words, to find it challenging to talk to each other about a larger set of issues when their universities employ a relatively large number of DEI personnel.

The 17 topics that are included in the overall measure found in Figure 15, however, are a diverse and wide-ranging group. Lumping these topics together makes sense if the goal is to identify broad variation in speech climates across a wide range of issues at different universities but it’s a poor approach for answering the narrower question of how DEI bureaucracies influence free expression on college campuses. Given their stated purpose and the primary focus of their campus activity, DEI bureaucracies are far more likely to have an effect (either positive or negative) on the campus speech climate connected to identity-related issues than on matters of public policy that are not so obviously connected to race, gender, and sexual orientation. We should expect, for example, that larger DEI bureaucracies will shape how difficult students believe “open and honest” conversations are on “transgender issues” and “racial inequality” but not on “gun control” and whether China is a geopolitical threat to the United States. In other words, we need to disaggregate these 17 topics in order to truly sort out whether (and on what issues) DEI bureaucracies are hindering campus speech.

Figure 16 shows the results of breaking the 17 topics down into two separate categories: topics related to identity (e.g. gender inequality, racial inequality, affirmative action, abortion, transgender issues, sexual assault) and topics not related to gender and race (e.g. gun control, China, COVID vaccine mandates, mask mandates, free speech, etc.).

Figure 16 – Number of Identity and Non-Identity Topics that are Difficult to Discuss Openly and Honestly

As the graphs demonstrate, larger DEI bureaucracies are correlated with more difficulty having open and honest discussions on both identity-related (r=.31, p=.01) and non-identity-related issues (r=.41, p=.00).

The size of university DEI bureaucracies is highly correlated with student ideology and the racial and ethnic diversity of the student body (see above). This just means that the bivariate relationship between DEI and the number of speech topics might be spurious. Consistent with this possibility, Table 5 shows that the size of the DEI bureaucracy is no longer a significant predictor of student feelings about identity-related and non-identity-related topic discussions once we control for other likely influences on student feelings. So, once again, we must conclude that although DEI bureaucracies may or may not make “open and honest” conversations on college campuses harder, they almost certainly do not make them easier. 

Table 5 - Number of Identity and Non-Identity Topics that are Difficult to Discuss Openly and Honestly

Conclusion: DEI and the “Free Speech Crisis”

There’s congealing conventional wisdom among critics of higher education that university DEI bureaucracies are “cancers that stifle free speech.” Yet, there have been no empirical studies to date testing these claims. Combining data from the Heritage Foundation and FIRE, this post shows that there is, in fact, some truth to these claims. Universities with more DEI personnel are only creating a more welcoming, accepting, and inclusive environment when the question is whether controversial liberal speakers should be allowed to speak on campus. On every other dimension of campus speech, larger DEI bureaucracies do not lead to better free speech climates. What’s more, along a number of other dimensions of campus speech (e.g. student comfort with expression outside of classrooms, support for disruptive action in response to offensive speech, the perceived difficulty of having “open and honest” conversations), universities with more DEI employees perform worse. In short, DEI bureaucracies commonly hurt and infrequently help free expression on college campuses. Whatever their benefits (and they may be considerable on other dimensions), DEI bureaucracies do not help speech climates.

These conclusions should be considered alongside three caveats. First, it is impossible with the data we currently have (i.e. snapshots of the DEI bureaucracy and attitudes towards free expression in 2022) to definitively conclude that the number of DEI personnel is leading to worsening campus speech climates and not the other way around. Indeed, one possible explanation consistent with the above data is that an oppressive DEI bureaucracy is making students less supportive of free speech. Another possibility, however, is that censorious students are successfully demanding larger and more intrusive DEI bureaucracies. While the latter explanation seems far less plausible than the former, we can’t, strictly speaking, rule it out with the existing data.

Second, the findings presented above only apply to the relatively small and deeply unrepresentative group of 71 universities studied here (a combination of Power 5 conference and Ivy League schools). If data on the number of DEI personnel was available for the more than 200 universities surveyed by FIRE, we might have arrived at a very different set of conclusions about how DEI shapes campus speech environments.4 Most notably, this analysis does not include the small, Northeastern liberal arts colleges that were initially responsible for elevating concerns about campus free speech beginning in 2015. Obviously, more work is needed to collect DEI information on the full set of American universities.

Finally, the findings discussed above only apply to a single moment in time (2022).  Campus speech climates can change quickly. Student populations experience tremendous turnover across time, with the student body being replaced in its entirety every few years. What’s more, a single, high-profile controversy on a campus might dramatically alter how comfortable students feel expressing themselves. Similarly, DEI efforts may expand or contract in ways that communicate different things to different student cohorts about the boundaries of acceptable speech. All of this is just to say that the strength of the relationship between DEI bureaucracies and free speech environments is a variable, not a constant, and it should be periodically reexamined.

Subscribe now

3

Some might be concerned that the displayed regression line is incorrectly suggesting a relationship that does not exist given the presence of outliers in the data. Stanford, for example, is something of an outlier. Harvard and Duke might also be outliers. Dropping these three “outliers” does not change the results of the analysis in any way. In other words, the bivariate analysis shows the same relationship even with the exclusion of the outliers:

The U.S. military is facing an “unprecedented” recruitment crisis, with the Army, Navy, Air Force, Coast Guard, and National Guard all failing to meet their enlistment goals in 2022. This recruitment crisis is partly a function of the fact that so few young people are now able to serve in the military due to obesity, educational deficiencies, mental health problems, or criminal records. In fact, according to the Associated Press, “only 23% [of American youth] are physically, mentally, and morally qualified to serve without receiving some type of waiver.”

The recruitment crisis is not entirely a consequence of these metastasizing physical, mental, and moral problems, however. It is also a direct result of the growing unwillingness of young people to serve in the armed forces. The most recent estimates “show that only 9% [of America’s youth] are even interested in military service.”

So why are so few young people willing to serve in the military? For some conservative observers, the answer is obvious: the Pentagon’s increasingly “woke” diversity, equity, and inclusion (DEI) policies are alienating the groups most inclined to serve in the armed forces (white conservatives, white Southerners, and whites living in rural areas).

There’s no shortage of pieces advancing this kind of argument. Consider, for example, Jimmy Byrn’s recent article in the Wall Street Journal (entitled “What if They Gave a War and Everybody Was Woke? The military’s embrace of faddish politics may make activists happy, but it’s driving away recruits”). Byrn begins by reviewing data on the military’s recruitment crisis:

Nearly every branch has struggled to meet its recruitment goals for 2022, with some falling as short as 40%. Worse yet, only about a quarter of America’s youth meet current eligibility standards—and recent surveys show only 9% are even interested.

Byrn then walks through a laundry list of “woke” policies initiated during the first few months of the Biden administration:

Defense Secretary Lloyd Austin mandated that every military unit conduct a “stand-down” to confront “extremism in the ranks.” The chief of naval operations, Adm. Mike Gilday, added Ibram X. Kendi’s “How to Be an Antiracist” to his reading list for sailors …The Navy has mandated gender-sensitivity training, and released a video encouraging sailors to closely police the use of pronouns as well as everyday language, declaring that those who fail to comply aren’t “allies” of their fellow sailors.

Byrn argues that these changes are a “repellent” for conservatives and Southerners:

“[S]uch measures have amounted to a form of antirecruitment for prospective enlistees. The Pentagon is appealing to activists at the expense of those most likely to serve. The military has historically drawn an outsize proportion of recruits from conservative Southern states….Unsurprisingly, military members privately skew conservative. In the 2018 midterm elections, nearly 45% of service members surveyed indicated they would back Republican candidates, versus 28% who favored Democrats. Support for Republicans among veterans was similarly strong in 2020.”

Striking a similar tone in Imprimis, Thomas Spoehr argues that:

“Wokeness in the military affects relations between the military and society at large. It acts as a disincentive for many young Americans in terms of enlistment…Recent reports show the military’s dismal failure to gain new recruits in adequate numbers...Is anyone surprised that potential recruits—many of whom come from rural or poor areas of the country—don’t want to spend their time being lectured about white privilege?”

Unlike Byrn, Spoehr sees “wokeness” in the military beginning prior to 2021. As he points out, “The push for it didn’t begin in the last two years under the Biden administration” and many of the military’s “woke” policies have been in place since at least 2015.

The JAMRS Surveys

Is there any evidence that the desire to serve in the military has declined among white Republicans, white Southerners, and white rural residents since 2015 as a result of “wokeness” within the military? The only easily accessible data on this question (i.e. data that an average citizen without advanced knowledge of statistical software packages can locate) comes from eight, short PowerPoint presentations based on the Department of Defense’s JAMRS surveys.

Three times a year, the JAMRS surveys gather a nationally representative sample of more than 3,000 16- to 24-year-olds in the United States. The survey asks its respondents, “How likely is it that you will be serving in the Military in the next few years? (Definitely, Probably, Probably not, Definitely not).” The aforementioned PowerPoint presentations display aggregate and subgroup (racial and gender) trendlines over time:

Figure #1 – Likelihood of Military Service (JAMRS)

As Figure #1 shows, the percentage of 16 to 21-year-olds saying they will “definitely” or “probably” serve in the military during the “next few years” averaged approximately 13% in 2015. After dropping somewhat between 2015 and 2016, the percentage stayed around 12% through 2019. In 2020, however, the aggregate likelihood of military service dropped to 10%. By 2021, the number reached a historic low of 9%.

The JAMRS presentations imply that labor market competition, not “woke” DEI policies, is behind young people’s diminishing propensity to serve. Indeed, the survey data on likelihood is immediately followed in the JAMRS PowerPoint presentations by a section entitled “U.S. Military and Economic Conditions.” In this section, results from a question on job prospects (“How difficult is it for someone your age to get a full-time job in your community?”) are presented: 

Figure #2 – Economic Perceptions (JAMRS)

These presentations send a clear message: the likelihood of serving in the military is declining because the difficulty of getting a “full-time job in your community” is also going down. In other words, the recruitment crisis has an economic cause and requires an economic solution.

Unsurprisingly, the publication of the most recent JAMRS PowerPoint slides (Fall 2021) was followed by a steady stream of calls for expanding military pay and benefits so that the armed forces might better compete with other career opportunities. Representative Anthony Brown, for example, issued a statement demanding the Pentagon provide “better pay, training, opportunities, connections, and benefits” as part of an effort to “do more” to encourage young Americans to join the military. Striking a similar tone, Representative Jason Crow told Politico that the Pentagon should “provide enlistment incentives and bonuses” to make military careers seem more desirable to young people than other job opportunities.

The most notable example of how the JAMRS data provoked calls for increased military benefits came with Brittany Dymond’s (the Associate Director of the National Legislative Service for the Veterans of Foreign Wars of the United States) congressional testimony on the “Status of Military Recruiting and Retention Efforts Across the Department of Defense.” In her statement, Dymond relied exclusively on the JAMRS reports to argue that “Congress must ensure military benefits such as pay, health care, tuition assistance, and retirement are competitive with the private sector, continuously improved, and come without cost increases to members and families as applicable.”

The runaway economic optimism illustrated in Figure #2 is undoubtedly an important factor in the military’s ongoing recruitment crisis and military pay, educational incentives, and benefits are essential for encouraging young people to enlist. What I want to explore here, however, is the idea (implied by the JAMRS presentations) that economic factors are the only drivers of the recruitment crisis. Specifically, I want to test Byrn and Spoehr’s claim that interest in serving in the armed forces is declining among white Republicans, white Southerners, and white rural residents as a result of the military’s growing “wokeness.”

Unfortunately, with the exception of what is contained in the eight short PowerPoint presentations posted on the JAMRS website, the JAMRS data is not publicly available. As a result, we do not know what other questions were asked in the JAMRS survey or how responses to those questions may have correlated with each other (including with the “likelihood of serving” question). Most importantly, we don’t know whether JAMRS asked questions about the desire to serve (as opposed to the mere likelihood of serving) or about political beliefs.

In this post, I’m going to take advantage of a largely unexamined dataset – the Monitoring the Future survey – to explore preferences for military service among whites since 2015. As I demonstrate, both the JAMRS reports and conservative observers are misdiagnosing the causes of the recruitment crisis. The JAMRS reports suggest the military is facing a short-term economic problem. As I will show, the military is actually facing a longer-term political problem. While conservative observers like Byrn and Spoehr correctly intuit the military’s political problem, they are mistakenly locating that problem among Republican, Southern, and rural whites. They are also mistakenly attributing that problem to the Pentagon’s “woke” DEI policies. In fact, the political problem exists almost exclusively among white Democrats and it is almost exclusively a function of this group’s increasingly negative attitudes toward the country and toward the military.

Misdiagnosing the causes of the recruitment crisis will mean policy interventions (e.g. higher pay and more generous benefits) to enhance enlistment will fail. Given the scale of the recruitment crisis already, this failure would be catastrophic. My goal here is to help avoid this catastrophe by drawing attention to the role that political “wokeness” is playing in killing military “willingness.”  The military cannot successfully fight what is partly an ideological war with economic weapons alone.

The Monitoring the Future Data

Every year, the Monitoring the Future project collects a “large, distinct, nationally representative samples of 12th-grade students in the United States.” The survey asks over 100 questions to more than 10,000 American high school seniors in order to “explores changes in important values, behaviors, and lifestyle orientations of contemporary American youth.”

Most important for my purposes here, the Monitoring the Future survey asks respondents a question about their partisanship and two questions about their views on military service:

1. How likely is it that you will do each of the following things after high school… Serve in the armed forces? (Definitely Won’t, Probably Won’t, Probably Will, Definitely Will)

2. Suppose you could do just what you'd like and nothing stood in your way. How many of the following things would you WANT to do? (Mark all that apply.) Serve in the armed forces?

These two questions track distinct, yet closely related dimensions of young people’s orientation toward military service. The “likelihood of serving” question is nearly identical to the item asked in the JAMRS surveys. This question avoids making any reference to the respondent’s preferences about military service. Instead of asking what the respondent “wants” to do, it asks them only what it is “likely” that they will do (independent of their desires). As a result, the question primes respondents to balance their preferences against their opportunities and constraints. It is possible, for example, that a respondent might answer this question by selecting the “definitely will” response because he or she imagines a reinstatement of the military draft (forcing them to serve regardless of whether they want to or not). Alternatively, a respondent might have a lifelong dream to serve in the military but realize that it will not happen because of familial obligations that require them to stay close to home. In short, this question asks respondents for a measured prediction about their future, not for a decontextualized disclosure of their preferences. 

By contrast, the question about “wanting” to serve in the military directly taps what young people feel about serving in the military. The question instructs respondents to ignore the opportunities and constraints they face when deciding what to do with their lives. Indeed, it begins with the sentence, “Suppose you could do just what you'd like and nothing stood in your way….”). The question also explicitly primes respondents to think about their desires. Specifically, it asks them, “How many of the following things would you WANT to do? [capitalization in original].”

Byrn and Spoehr’s claims are primarily about the desire of Republican, Southern, and rural whites to serve in the military rather than the likelihood that they will serve. I will focus my attention here, therefore, on this dimension of attitudes toward military service.

Relatedly, I should also point out that the MTF data underscores the problem with focusing on likelihood alone (as the JAMRS data apparently does). Empirically speaking, desire to serve in the military is not a sufficient condition for being likely to serve in the military. As Figure #3 shows, since 2015, only 58.3% of high school seniors who expressed a desire to serve also thought they would eventually enlist in the armed forces. Desire to serve in the military does seem, however, to function as a necessary condition for being likely to serve in the military. As Figure #3 shows, those who do not want to serve in the military stand almost no chance of saying that they will serve in the future. To be more precise, more than 96% of those who said they did not want to serve in the military also said that they “probably won’t” or “definitely won’t” serve in the military.

Figure #3 – The Relationship between Likelihood of and Desire for Military Service (MTF)

The fact that desire works as a necessary (but not sufficient) condition for being likely to serve in the military suggests that we need to pay particular attention to the size and growth of the “do not want to serve” category. If it becomes too large, there is very little the military can do (e.g. increase pay and benefits) to recruit the soldiers it needs. The JAMRS data, with its exclusive focus on the estimated likelihood of service, directs our attention away from this basic fact and necessarily leads us to focus entirely on economic solutions to recruitment problems. Any attempts to fully understand the military’s recruitment crisis should, therefore, use the MTF data (and, in particular, the desire to serve question). Surprisingly, however, I was unable to find even a single report, statement, or piece of congressional testimony related to military recruitment from the last 10 years that cited the MTF data.

Testing Byrn and Spoehr with the MTF

Byrn and Spoehr’s claims about “woke” DEI policies acting as a “repellent,” a form of “antirecruitment” and a “disincentive” for white Republicans, white Southerners, and rural whites are testable using the MTF data. Indeed, if the military’s increasing “wokeness” after 2015 (and particularly after Biden took office in early 2021) is driving individuals from these groups away, we should see their disaffection in the results of the MTF survey question asking respondents if they “want” to serve in the armed forces.

Figure #4 – Likelihood of and Desire for Military Service among Whites

Figure #4 displays responses to the MTF’s likelihood and desire questions among white respondents since 2015. The trend lines between 2015 and 2020 are essentially flat and indicate very little change in white orientations towards military service. Figure #4 also shows, however, a significant drop in white desire (14.8% to 10.0%) occurring between 2020 and 2021.

The drop in desire among whites is broadly consistent with Bryn and Spoehr’s claims but we need to investigate the trends among white Republicans, white Southerners and white rural residents to truly assess their argument. Let’s examine partisanship first. As the data in Figure #5 show, there has been no meaningful decline in desire among white Republicans (indicated, in part, by the fact that the 95% confidence intervals overlap with each other when looking across years). In 2015, for example, 18.1% of white Republicans expressed a desire to serve in the military. In 2021, nearly the exact same share (18.9%) of white Republicans expressed a desire to serve in the military. Despite some minor fluctuations in the exact estimates, there were no statistically significant changes across years in white Republican orientations towards military service. If white Republicans have been repelled from military service by “woke” DEI policies, there is no evidence of it yet.

Figure #5 – Military Attitudes among White Republicans

What about the argument that white Southerners are turning away from military service? Once again, there is nothing in the data to indicate that white Southerners have been disincentivized from serving in the armed forces by policies emphasizing “anti-racism,” “gender-sensitivity” or “white privilege.” In 2015, 16.2% of white Southerners wanted to serve in the military. In 2021, 16.6% of white Southerners wanted to serve in the military. Contrary to the claims of Byrn and Spoehr, in other words, there has not been a meaningful decline in desire for military service among white Southerners.

Figure #6 – Military Attitudes among White Southerners

Finally, there’s Byrn and Spoehr’s suggestion that woke DEI policies are delivering an “anti-recruitment” message for rural whites (regardless of which region they hail from). While there is no good measure of urbanicity in the MTF data, the project does provide a dichotomous measure of whether each respondent lives in a large metropolitan statistical area (MSA) or not. In 2015, 16.5% of whites living outside of large MSA’s wanted to serve in the military. In 2021, 16.6% of these whites did. Put differently, there’s really nothing in the MTF data pointing to a change in how rural whites view a career in the armed forces.

Figure #7 – Military Attitudes among White Non-MSA Residents

Collectively, these graphs provide initial evidence against the idea that “woke” DEI policies have significantly depressed white Republican, white Southerner and white rural resident interest in military service. I want to be clear, however, that these results do not mean that Byrn and Spoehr are necessarily wrong about the effects of “woke” DEI policies. It simply means that they are not right yet.

Young white Republicans strongly oppose the kinds of policies Byrn and Spoehr describe. Consider opinions on the question of whether there should be “preferential hiring and promotion of blacks.” Less than 5% of young white Republicans now endorse this kind of race-based affirmative action:

Figure #8 – Support for Affirmative Action Policies by Race and Partisanship

To the extent that policies such as these are implemented by the military, we should expect desire to serve in the military to decline among young white Republicans.

In order for “woke” policies to undermine interest in military, however, they first have to be well-known. Prior to 2022, woke DEI policies in the military received almost no sustained media attention. Indeed, Byrn and Spoehr’s articles were worth publishing in the summer of 2022 precisely because they directed attention to a set of issues that were previously ignored.

At this point, I should point out that the barriers to informing a nontrivial number of young white Republican, Southern, and rural voters of the existence of “woke” military policies are considerable. Most Americans pay little attention to politics and are ignorant about even the most basic and long-standing facts about the American political system. In 2022, for example, less than half of Americans could correctly identify all three branches of government and more than a quarter could not name a single right protected by the First Amendment. Similarly, only 57% of voters under the age of 30 could correctly identify which party controlled the House of Representatives in 2020. If the public has not learned about these foundational aspects of American politics, they are incredibly unlikely to learn about the details of the military’s DEI policies.

With that said, more sustained public discussion of “wokeness” in the armed forces might very well lead to the effects that Byrn and Spoehr describe. We just can’t know until we get more recent data.  

If Not White Republicans, Southerners, and Rural Residents, then Who?

While the above analysis helps us reject key parts of Byrn and Spoehr’s argument, it still leaves us with a puzzle: if white Republicans, white Southerners, and white rural residents are not expressing less desire to serve, which whites are responsible for the 5-point decline in desire illustrated in Figure #4?

As it turns out, white Democrats –the group most commonly ignored by conservative diagnoses of the recruitment crisis and the group most likely to support the military’s efforts to implement highly progressive DEI policies (see Figure #8) – are actually the ones driving the aggregate decline in white desire discussed above.

Figure #9 – Military Attitudes among White Democrats

In 2015, 12.2% of white Democrats wanted to serve in the military. That number fell to 3.6% in 2021 (Figure #9). This striking decline does not, however, tell the full story. Most of the collapse in white Democratic interest occurred among men (Figure #10). In 2015, 18.6% of white Democratic men expressed a desire to serve in the military (only slightly lower than the 19.9% of non-Democratic men who wanted to serve). By 2021, that number had dropped to only 2.9%.

Figure #10 – Military Attitudes among White Democrats by Gender

In fact, as Figure #11 shows, black men (12.1%), Hispanic men (14.3%) and white Republican men (25.3%) were all at least 4x more likely than white Democratic men in 2021 to “want” to serve in the armed forces. Democratic white men were also significantly less interested in military service than black women (9.5%) and white Republican women (10.9%).

Figure #11 – Desire to Serve in the Armed Forces by Gender

What’s Dampening Desire among White Democrats?

Why have white Democrats (especially young white Democratic men) soured on military service? Answering this question requires an examination of why people join the military in the first place. Fortunately, understanding the motives that drive young people to enlist in the American military has been a frequent focus of research since the introduction of an all-volunteer force. Survey and in-depth interview data have revealed “a range of motivations for military service, clustering around, on the one hand, the [extrinsic] benefits associated with service [e.g. pay, educational incentives, the opportunity to develop leadership skills, exposure to diverse nations and cultures via travel, an interesting job with variety, and experience to prepare for the future] and, on the other hand, normative commitments to the political community [e.g. patriotism and duty].”

The conservative critique of military “wokeness” alerts us to yet another possibility: orientations towards military service may be shaped by feelings about the military as an institution. It is possible that if young people feel negativelyabout or mistrustful of the military, they will not develop a desire to serve in it (independent of the benefits provided by enlisting and independent of their patriotism). This lack of interest might simply be a function of rational expectations given the perceived nature of the institution. If the military is seen to be untrustworthy, it is unwise to expect it to actually deliver the extrinsic benefits it is promising. If the military is disliked, serving it might not be consonant with one’s sense of patriotism and civic duty.

Overall, then, we want to examine whether perceptions of the military’s extrinsic benefits, the strength of normative commitments to the political community, and feelings towards the military as an institution have disproportionately declined among young white Democrats during the last few years. If so, these changes may be driving the declining desire to serve in the armed forces.

In order to truly assess whether these three attitudes are changing among young white Democrats, we would need to have survey data that: (1) asked questions about economic prospects, patriotism, and feelings towards the military; (2) repeated these questions at multiple points in time between 2015 and 2021; (3) collected sufficiently large sample sizes to yield a nontrivial number of young, white Democrats; and (4) is publicly available.

Unfortunately, there is no polling data that meets all four of these requirements. The best we can do, therefore, is make some educated guesses based on the imperfect data (i.e. meets one or two of the above criteria) that we do have.

Subscribe now

Extrinsic Benefits

The extrinsic benefits provided by military service are equal for every group. In order to believe that perceptions about extrinsic benefits are dampening enthusiasm for military service among young White Democratic men, there would have to be a reason why young white Democratic men were the only subgroup among the general population of young white people to change their perceptions of the military’s extrinsic benefits after 2015 (and particularly after 2020).

One’s perception of the military’s extrinsic benefits is likely shaped, in part, by one’s social position. Young people from more affluent backgrounds are likely to be less attracted to the pay, health care, retirement, job training, and other benefits provided by military service. If young white Democratic men are more likely to be raised in affluent backgrounds than other whites, they may find the military’s extrinsic benefits less appealing. More importantly, if the socioeconomic profile of young white Democratic men improved significantly since 2015, it may explain their declining desire to serve in the military.

Unfortunately, the MTF provides no data on the socioeconomic status of its respondents. It is not possible, therefore, to directly test the hypothesis that young White Democratic men’s declining desire to serve might be a consequence of their improving social position after 2015. Needless to say, however, the idea that the social position of young, white Democratic men (and no other group) shifted so uniformly and dramatically that it altered their short-term perceptions of economic benefits is implausible. It is far more likely that feelings towards the country and the military have evolved among this group over the last seven years.

Patriotism

As mentioned above, patriotism is an important predictor of military service. A majority of soldiers consistently describe their service as motivated by a deep-seated sense of patriotism and duty. For example, when soldiers in two infantry battalions were asked in 2002 to identify all the reasons that “were important in your decision to join the Army,” two of the most popular options were “serve country” (65.8%) and “patriotism” (54.9%).

Surprisingly, the MTF data contains no questions related to national pride or the strength of young people’s attachment to their national identity. Gallup has asked a question about patriotism (“How proud are you to be an American -- extremely proud, very proud, moderately proud, only a little proud or not at all proud?”) on a nearly yearly basis since 2001. While Gallup occasionally reports how responses to this question have varied over time and across groups, they do not provide access to the full dataset and only provide a limited number of group-specific analyses. As a result, we cannot identify changes among young white Democrats.

The data that Gallup has made public, however, are consistent with declining levels of patriotism among young White Democrats. As Figure #7 shows, “extreme pride in being American” have fallen more than 20% among Democrats since 2015 and more than 30% since 2013.

Figure #12 – National Pride by Partisanship

The Gallup data also shows that young people generally express far less national pride than older people. Specifically, as Figure #13 demonstrates, only 25% of 18-34 year olds are “extremely” proud to be American (compared to 51% of those 55 and older).

Figure #13 – National Pride by Gender, Age, and Education

Although the publicly available Gallup data cannot definitively prove that young white Democrats have become less patriotic since 2015, the tables and figures they have provided are consistent with this possibility. Indeed, if young Democrats are merely representative of older Democrats and young people more generally, there has likely been a greater than 25% decline in patriotism within the group over the last eight years. Such a decline could easily account for the 10% reduction in desire for military service documented above.  

Unfortunately, Gallup is the only survey organization that has asked a question about national “pride” over time. In 2016, 2020, and 2022, the American National Election Study (ANES) asked a slightly different question about national identity: “How important is being American to your identity?” While this question does not directly tap feelings of patriotism or national pride (e.g. respondents may believe that their national identity is important but dislike the country), it does provide a rough proxy for how people feel about their country.

The ANES data contains relatively few young, white Democrats. With this in mind, the ANES shows significant differences across both age and partisanship. As Figure #13 shows, American national identity is losing its salience among young white Democrats. Between 2016 and 2020, the percentage of white Democrats under 35 who claimed that “being American” was “extremely” or “very” important to their identity fell from roughly one-half (49.1%) to one-quarter (25.3%). Young white Democrats are now far less likely than older white Democrats and white Republicans of all ages to say that their national identity is important to them. Once again, this data is consistent with the idea that declining national attachment explains the declining desire to serve in the military among young white Democrats.

Figure #13 – Importance of American Identity by Age and Partisanship

A number of one-off surveys also indicate that young white Democrats may have much lower levels of patriotism than other groups. The Harvard Youth Poll periodically surveys more than 2,000 young Americans (aged 18 to 29). The Youth Poll asks a wide range of questions about politics, current events, and mental health. While the team behind the Youth Poll provides extensive cross-tabs (e.g. data showing variation in attitudes by age, race, gender, and other individual attributes), they do not provide full access to the data. As a result, it is not possible to examine how young, white Democrats respond to the survey’s questions.

The Fall 2021 iteration of the Youth Poll asked two questions that are helpful in understanding the patriotism of young Americans. First, the poll asked “Which statement comes closest to your own view? ‘America is the greatest country in the world’ or ‘There are other nations as great or greater than America.’” There was a massive partisan gap in responses to this question (see Figure #14). As the survey’s report summarizes, “The views of Democrats and Republicans are inverted with 21% of Democrats saying America is the greatest country and 64% saying other nations are as great or greater; 62% of Republicans believe that America is the greatest with 24% saying other nations are as great or greater.”

Figure #14 – “Great Country” Beliefs among Young People by Partisanship

Additionally, the Harvard Youth Poll asked, “Compared to your parents, do you think you more or less patriotic.” 32% of Democrats reported being “less patriotic” than their parents while only 11% reported being “more patriotic” than their parents. By contrast, 19% of Republicans reported being “less patriotic” than their parents but 21% reported being “more patriotic” than their parents. If young white Democrats are representative of young Democrats overall, the Youth Poll data is indicative of the kind of low and declining patriotism that could explain the collapsing desire of young white Democrats to serve in the armed forces.

More evidence about the declining patriotism of young people can be found in the surveys from the “More in Common Project.” The Project’s 2020 survey discovered that only 51% of Gen Z was proud to be American (compared to 94% of the Silent Generation). What’s more, identifying as “progressive” or “passively liberal” was associated with far less national pride (Figure #15):

Figure #15 – National Pride by Race, Age, Gender, and Ideology

More recently, a Morning Consult poll found that only a tiny percentage of Gen Z feels “proud to live in the United States.”

Figure #16 – Patriotic Sentiment by Age

These disparate surveys tell a common story: patriotism and national pride have evaporated among Democrats and young people over the last decade. Strictly speaking, the evidence summoned here cannot tell us anything about whether young, white Democrats are exceptions to or exemplars of these general trends. The data, however, are entirely consistent with the idea that the attitudes of young, white Democrats are changing in ways that could reduce their interest in military service.

Feelings towards the Military

There is no publicly available, over time data on how young white Democratic men feel about the military as an institution. Gallup has asked a yearly question about confidence in the military but they provide only the topline results (no subgroup analyses) to that question. The Gallup data shows a small and steady overall decline in the public’s confidence in the military since 2015 (dropping from 72% in 2015 to 64% in 2022). It is possible that this decline has occurred entirely among young white Democrats but we cannot know for sure with the data Gallup has provided publicly.

Better data is available from the Reagan National Defense Forum (RNDF). Each year since 2018, the RNDF has asked a nationally representative sample how much “trust and confidence” they have in the military. The RNDF provides crosstab summaries of the results (allowing us to examine how young people and Democrats feel about the military).

The RNDF results show a massive decline in “trust and confidence” among young people (18-29) and Democrats between 2018 and 2021. In 2018, 87% of young people and 92% of Democrats expressed “a great deal” or “some” confidence in the military. In 2021, however, only 67% of young people and 76% of Democrats expressed this sentiment. These results are broadly consistent with the possibility that young white Democratic men have expressed less desire to serve in the armed forces since 2015 as a result of more negative feelings towards the military.

In addition to the Reagan National Defense Forum data, there are a number of polls that provide a snapshot of how the public feels about various aspects of the. Generally speaking, these polls are also consistent with the idea that young white Democrats have disproportionately negative views towards the military. The Spring 2022 iteration of the Harvard Youth Poll, for example, asked young people to identify the most important areas for ensuring America’s global strength in the future. Only 12% of young Democrats identified the military (compared to 35% of young Republicans).

Figure #17 – Perceptions of the Military’s Importance among 18-29 Year Olds by Party

A number of other studies provide data suggesting generational (but not partisan) differences. In one recent survey, for instance, Morning Consult found only 55% of Gen Z (compared to 86% of Boomers) reported feeling  “some” or “a lot” trust in the military (Figure #18). In a different Morning Consult survey, 20% of Gen Z said they would stop “buying from a brand they currently shop at if the company supported the US military” (compared to only 8% of Boomers).

Figure #18 – Institutional Trust by Age

Recent data from the Pew Research Center finds similarly negative views towards the military among young people, with only 49% of 18-29 year olds saying that the military has a “positive effect” on the “way things are going in the country these days.”

Figure #19 – Views on the Effects of Institutions by Age

Once again, if white Democrats are merely representative of Gen Z in general, these surveys provide evidence that young white Democrats harbor very negative feelings towards the military. These feelings may be responsible for the fact that young white Democrats are showing little interest in military service since 2015.

The “Great Awokening”

The dramatic and relatively recent attitudinal shifts among young, white Democrats documented here should not be surprising to anyone who has followed American politics over the last eight years. As Zach Goldberg described in an influential 2019 article, white liberals experienced an unprecedented and profound “Great Awokening” sometime between 2012 and 2016:

“The rapidly changing political ideology of white liberals is remaking American politics….This is evident across a range of issues: the rapid growth in white liberals who favor affirmative action for blacks in the labor force; in the increase in white liberals who feel that we spend too little on helping blacks, and that the government should afford them special treatment; in the increase in white Democrats who think it’s the government’s job to ensure “equal income across all races”; and in the increase in white liberals and Democrats who think that white people have ‘too much’ political influence….[and] “recently they became the only demographic group in America to display a pro-outgroup bias—meaning that among all the different groups surveyed white liberals were the only one that expressed a preference for other racial and ethnic communities above their own.”

Goldberg was not alone in recognizing these changes. In a widely read 2019 article for Vox, Matthew Yglesias wrote:

“In the past five years, white liberals have moved so far to the left on questions of race and racism that they are now, on these issues, to the left of even the typical black voter. This change amounts to a “Great Awokening” — comparable in some ways to the enormous religious foment in the white North in the years before the American Civil War.”

While Goldberg and Yglesias’ analyses focused mostly on the changing racial consciousness of white liberals, the seismic attitudinal shifts described in their articles have had implications that go far beyond race. As this post has shown, in addition to becoming far more progressive on questions of race, gender, and immigration, young white Democrats have also become far less patriotic and supportive of the American military. A predictable consequence of these increasingly negative attitudes is a growing disinterest in military service. The interrelated set of highly progressive political beliefs, best characterized as “wokeness,” in other words, is beginning to represent an existential threat to the continuation of America’s all-volunteer armed forces.

Conclusion

The American military is facing a recruitment crisis driven, in part, by young people’s declining desire to serve in the military. According to some observers, the Pentagon’s recent moves to implement “woke” DEI policies are alienating young white Republicans from the military. This post has shown, however, there is no evidence to suggest that anything the Department of Defense has done since 2015 is exerting a negative effect on young white Republican desire to serve in the armed forces. This may change in time, of course. Growing wokeness within the military has already precipitated growing attention. Growing attention has the potential to precipitate declining desire among young white Republicans. Yet, as of 2021, there’s nothing to suggest young white Republicans are the source of the military’s recruitment problems.

There’s much to suggest, however, that young white Democrats are an important part of the military’s struggles. While the publicly available data is insufficient for providing definitive proof, there is a mountain of survey evidence consistent with the idea that diminishing enthusiasm for military service among young white Democratic men is a function of their increasingly negative views towards the country and the military.  

The military needs to take the “Great Awokening” of young white Democrats seriously and not assume that better pay and more generous benefits will persuade an increasingly reluctant and oppositional demographic group to enlist. “Wokeness” is killing desire to serve in the military and the death of desire might mean the death of an all-volunteer armed forces.

One final note, the evidence presented here about recruitment does not mean that there are no reasons to oppose “woke” military policies. Even if “woke” policies do not act as a form of “antirecruitment,” they still may have serious, negative consequences for the ability of the military to carry out its responsibilities. “Woke” policies may, for example, alienate active duty service members and hurt retention efforts. They may also seriously undermine military readiness. If so, the argument against “wokeness” in the military should be more narrowly framed around these claims and not connected to the ongoing recruitment crisis.

Subscribe now

Last week, Media Matters published a report entitled, “Fox’s coverage of violent crime dropped after the midterms: Weekday violent crime segments decreased by 50% in the week of the election compared to the weekly average since Labor Day.” According to the report, Fox News was “using violent crime as a political cudgel against Democrats throughout the midterms.” Comparing this issue to the 2018 “migrant caravan,” the report implied that Fox News was engaging in an excessive amount of violent crime coverage for no other reason than because of the perception that it would help Republicans win elections.

The evidence contained in this report (along with its claim that Fox News was responding to “perceived political benefits”) was tweeted out by Matthew Gertz (a senior research fellow at Media Matters) on November 17:

The tweet went viral, receiving over 40,000 likes in its first 24 hours (including a retweet from Alexandria Ocasio-Cortez). Mainstream news outlets quickly picked up the story as well. Yahoo News’ The Wrap, for example, ran the following story based on the Media Matters analysis:

The Daily Beast followed with a similar story accusing Fox News of “fearmongering”:

The implication in all of these reports was clear: Fox News was strategically emphasizing violent crime as a way of advancing the Republican Party’s electoral interests in the midterm congressional elections. Unfortunately for Media Matters, AOC, Yahoo, and the Daily Beast, the empirical evidence is woefully insufficient for drawing this conclusion.

In this post, I’m going to show that Media Matters’ claim about Fox News falls apart once we examine coverage of other campaign issues and on other cable television news networks. Specifically, I present evidence that Fox News did not focus on violent crime more than the other cable television news networks, they were not unique among cable television news networks in turning away from violent crime coverage after November 8, and they talked about other campaign issues at levels similar to that of other cable television news networks. In other words, contrary to Media Matters’ argument, we can’t conclude that Fox News was motivated by the “perceived political benefits” of covering violent crime.

The analysis I present below should have been the first step for any organization interested in a good-faith examination of media coverage during the 2022 election campaign. What’s more, the data I rely on was not difficult to obtain. Given that Media Matters chose to ignore very simple and very obvious pieces of contextual data in favor of presenting a single graph from Fox News, I have to conclude that they are operating in bad faith. Sadly, by employing this bad-faith approach (i.e. strategically reporting or ignoring information based on its “perceived political benefits”), Media Matters is doing exactly what they accuse Fox News of doing.

As always, I hope that this post provides a model for how to critically examine the empirical claims that are increasingly populating our political discourse.

Measuring Coverage of Violent Crime

How did Media Matters produce the graph contained in the viral tweet? In the report summarized in the tweet, Media Matters describes their methodology in the following way:

Media Matters does not make their data publicly available and it’s not possible to replicate their “internal database of all original, weekday programming on Fox News.” It is also not entirely clear how they operationally defined a news “segment” or what a “violent crime in general” refers to in their coding scheme (e.g. would the category “violent crime” include carjacking, rape, physical assault, robbery, etc.?).

The difficulty of replicating Media Matters’ analysis often works to their advantage. Without the resources to track news coverage on a daily basis, it can be difficult to counter their empirical arguments about what the media is covering (and what it is ignoring). The asymmetry between the data available to well-funded groups like Media Matters and the data available to casual observers interested in critically assessing the validity of popular empirical claims makes fact-checking the well-funded groups difficult if not impossible.

Subscribe now

Fortunately in this case, however, we can construct an alternative measure of the Fox News agenda that closely approximates the one employed in the Media Matters analysis by using the Stanford Cable TV News Analyzer. The Analyzer” searches the Internet Archive's TV News Archive of MSNBC, CNN, and Fox News coverage since January 1, 2010. Researchers can “write queries that compute the amount of time words are heard in cable TV news.”

Using the “Analyzer,” I collected data on three aspects of cable television news coverage between September 5 and November 14 (the time period covered in the Media Matters tweet and report): (1) the amount of violent crime coverage on Fox News, MSNBC and CNN; (2) the amount of election coverage on Fox News, MSNBC, and CNN; and (3) the amount of non-crime, non-election coverage on Fox News, MSNBC, and CNN. Collectively, these data allow us to assess the Media Matters claim that Fox News excessively focused on violent crime in order to advantage Republicans in the midterm congressional elections.

I built my measure of violent crime coverage by using the “Analyzer” to search the Fox News, MSNBC, and CNN captions for the words “murder(s),” “violent,” “violence,” “assault,” and “crime.” While this is an incomplete and imperfect list of terms for fully measuring the concept “violent crime,” it should capture the essence of what Media Matters was attempting to measure in their study. Even if the terms underestimate the overall amount of time devoted to “violent crime” discussions, they will provide a consistent measure across the time frame of the study (September 5 through November 14) and across the three major cable news networks.

In order to replicate the analysis presented by Media Matters, I aggregated the data from the “Analyzer” into weeks. The graph below displays the number of minutes each week Fox News devoted to “violent crime” during the last two months of the 2022 campaign (placed next to the graph from Media Matters for the purposes of comparison):

Graph #1: Measuring Violent Crime Coverage

As you can see, the data from the “Analyzer” closely matches the data from Media Matters, suggesting that the former is an exceptionally good proxy for the latter. Given that we have a good proxy measure of coverage, we can now begin to explore the quantity and dynamics of violent crime discussion on Fox News (relative to the dynamics of violent crime discussion on other cable news channels).

Did Fox News Cover Violent Crime More than the Other Networks?

If the Media Matters thesis that Fox News covered violent crime an excessive amount in order to help Republicans win the midterm congressional elections is correct, we should expect to find (1) that Fox News covered violent crime far more than MSNBC and CNN and (2) that violent crime was the most covered issue during the period of study on Fox News but not on MSNBC and CNN.

The graph below shows how much the three cable television networks covered violent crime during the last two months of the 2022 campaign:

Graph #2: Violent Crime Coverage on Fox News, MSNBC and CNN

The graph clearly shows that there were relatively small differences in the amount of coverage devoted to violent crime on Fox News and MSNBC between September and November. Indeed, during any given week, MSNBC only covered crime about 6 minutes less than Fox News (or, put differently, about a minute less each day). While CNN covered violent crime less than Fox News and MSNBC, it still spent a significant amount of time on the issue. For example, 25 minutes of CNN’s coverage each week during late October were focused on violent crime (compared to 35 minutes on MSNBC and 39 minutes on Fox). In short, violent crime was an important issue for all of the cable TV channels, not just Fox.

To show just how absurd the Media Matters claim is, we could easily create a graph of coverage on MSNBC and advance the exact same argument about the motivations behind their coverage (i.e. that they were focusing on violent crime in order to help Republicans do better in the election). Notice how this MSNBC graph could be easily substituted from the Media Matters graph:

Graph #3: Violent Crime Coverage on MSNBC

The Media Matters claim that Fox News was cynically focusing on violent crime to help Republicans might carry more weight if Fox News covered violent crime to the exclusion of other important political issues. In order to see how much attention Fox News paid to other issues, I used the “Analyzer” to track the time spent on four other major campaign issues: abortion, inflation/the economy, immigration, and threats to democracy. The “Analyzer” results for Fox News are presented below (with crime included as a reference):

Graph #4: Total Issue Coverage on Fox News

Fox News covered violent crime less than immigration between the weeks beginning September 12 and September 25. Fox News also covered inflation and the economy more than violent crime at numerous points during the two months of the Media Matters study (e.g. the weeks beginning September 12, September 19, and October 17). Coverage of violent crime spiked in late September and late October (reaching its peak with approximately 40 minutes of coverage a week immediately prior to the election). Overall, between September 5 and November 14, violent crime was discussed a total of 262.2 minutes on Fox News, abortion was discussed 78.2 minutes, inflation and the economy were discussed 270.9 minutes, immigration was discussed 111.3 minutes and threats to democracy were discussed 51.5 minutes. In short, violent crime was not the most discussed issue on Fox News overall or during significant periods of the 2022 campaign.

It is hard, however, to know what to make of these numbers in the absence of a broader context. One thing that would be helpful to consider, for example, is whether coverage reflects reality (e.g. is there a significant focus on violent crime because there is a significant problem with violent crime?). We should expect that rising violent crime rates will lead to rising news coverage rates.

While collecting and reporting violent crime data takes time, there is strong evidence that violent crime increased in the months before the 2022 election. As Axios summarizes:

Given that violent crime jumped 4.2% in the months prior to the election campaign and robberies “skyrocketed” by 12%, it is completely unsurprising that media attention to these issues also increased. In fact, if news coverage ignored these changes, we would probably be right to worry about media bias running in the direction of suppressing important stories for political purposes. Of course, it is difficult to know what the “right” amount of attention to social problems is but big increases in crime probably should be met with big increases in news coverage. In other words, there is an explanation for Fox News’ (and the other networks) coverage of violent crime that does not involve their “perceived political benefits.”

Alternatively, we might want to know whether the candidates running for office are making violent crime an issue. If news organizations are covering candidates running for office, their coverage will naturally (and appropriately) reflect what these candidates are talking about.

In order to determine what candidates were talking about during the 2022 campaign, I used data from the Wesleyan Media Project. The Wesleyan Media Project “provides real-time tracking and analysis of all political television advertising in an effort to increase transparency in elections.” In particular, it codes every ad run by Democratic and Republican candidates and groups for the issues discussed.

The data from the Wesleyan Media Project on House and Gubernatorial races are below (with “public safety” highlighted):

As the data show, both Democrats and Republicans were highly focused on crime during the final weeks of 2022 campaign (receiving attention in more than 25% of Democratic ads and 33% of Republican ads). Given this, it would have been extremely strange (and would represent bias of an entirely different kind) if Fox News and the other cable television news networks did not emphasize violent crime in their coverage during the last two months of the campaign. To state the matter more plainly, there are (again) good explanations for the pattern of coverage that do not involve attempts at electoral manipulation by Fox News.

A final way of contextualizing Fox News’ coverage of violent crime is to consider what other cable television news networks were covering. As shown above, Fox News was not the only network talking about violent crime. Still, it would be helpful to know other cable news networks were covering violent crime significantly less than the other main campaign issues.

The “Analyzer” results for MSNBC are below:

Graph #5: Total Issue Coverage on MSNBC

As the data show, MSNBC discussed violent crime more than any other issue during the period of study. Specifically, MSNBC discussed violent crime for a total of 198.5 minutes (compared with 145.5 minutes for abortion, 130.6 minutes for inflation and the economy, 50.4 minutes for immigration, and 122.7 minutes for threats to democracy). Put another way, both Fox News and MSNBC devoted roughly the same proportion of their issue coverage to violent crime between September 5 and November 14 (33% vs. 31%).

The “Analyzer” data for CNN are here:

Graph #6: Total Issue Coverage on CNN

The data for CNN tell basically the same story as the data from Fox News and MSNBC. Violent crime was a major issue during the last two months of the 2022 campaign and its salience increased rapidly during the final weeks of October. Overall, inflation and the economy were discussed more on CNN than violent crime (138.9 minutes to 124.2 minutes) but the differences were relatively small. In short, the other networks made decisions about what to cover that were very similar to those of Fox News. It is difficult, therefore, to believe that Fox News was uniquely motivated by “political benefits.”

Did Fox News Stop Covering Crime After the Election?

Media Matters accuses Fox News of abandoning its violent crime coverage after November 8 because the election “eliminated the perceived political benefit of talking about crime all the time.”

As Graph #3 clearly shows, violent crime coverage on Fox News did fall after the election. As Graph #3 also shows, however, it also fell in exactly the same way (approximately to 0) on MSNBC and CNN. Additionally, coverage of all issues (abortion, inflation and the economy, immigration, threats to democracy) on Fox News (Graph #4) fell after the election. This is also what happened on MSNBC and CNN for every issue except abortion (Graphs #5 and #6). Rather than cynically shifting their coverage in response to shifting “political benefits,” Fox News (and the other networks) merely replaced their election campaign coverage with election result coverage. The result was a declining focus on every kind of political issue.

Another way of conceptualizing the way that election results displaced issue coverage (including violent crime) is to track discussion of voting and elections over time. The “Analyzer” data from Fox News is here:

Graph #7: Election and Violent Crime Coverage on Fox News

The data are unambiguous: there was less coverage of violent crime after the election and more coverage of voting and election results after the election. This, of course, is completely unsurprising. Any major event (e.g. election, war, terrorist attack) would have meant less coverage of violent crime. Just because there was less coverage of violent crime on Fox News after the election does not necessarily mean (as Media Matters argues) that Fox News was covering crime only to shape the outcome of the election.

Below are the graphs detailing coverage of violent crime on MSNBC and CNN:

Graph #7: Election and Violent Crime Coverage on MSNBC and CNN

As the graphs show, there was almost no difference between the pattern of violent crime and election coverage on CNN (right) or MSNBC (left). It is hard, therefore, to believe the argument that Fox was motivated by a fundamentally different set of incentives than either of these two networks.

Conclusion

The data from Media Matters is consistent with the claim that Fox News was attempting to focus on violent crime in order to advance the Republican Party’s electoral interests. As I have shown above, however, the data was not sufficient to draw this conclusion. Examining coverage of other issues and other networks reveals that Fox News behaved in ways that were not surprising and not unique. In short, the Media Matters argument can only be sustained by ignoring incredibly important pieces of contextual information.

I should point out here that there is a way of explaining this overall pattern of results that is consistent with Fox News behaving in accordance with a set of politically-motivated incentives. If Fox News chose to cover violent crime in order to help Republicans and MSNBC’s and CNN’s coverage primarily follows the lead of Fox News’ coverage (responding to the issues raised on Fox News), we would have the same pattern of results that are presented in the graphs above. This explanation, of course, requires that (1) other networks respond to Fox News; (2) Fox News does not respond to the other networks; and (3) MSNBC and CNN are also responding to political incentives (i.e. they want Democrats to do well so they counter Fox News information that is beneficial to Republicans).

Assessing each of these claims is beyond the scope of this post. I merely want to point out here that litigating these more complicated and nuanced issues would require a much different analytic approach than the one presented by Media Matters. In the future, I hope they do better.

Share

Last month, the Department of Defense sent a survey to active duty service members last asking dozens of questions about race relations in the military. The survey was leaked to the media two weeks ago and immediately received wide coverage as a result of its focus on racial harassment and discrimination. The attention to the leaked survey is somewhat surprising; it is essentially the same set of questions posed by DoD in their 2017 Workplace and Equal Opportunity Survey of Active Duty Members.

This post is about how the “Workplace and Equal Opportunity Survey’s” approach to assessing racial harassment and discrimination produces data that are (deliberately or accidentally) amenable to misinterpretation. Without understanding exactly how the survey produces results of this kind, we are doomed to reach the same faulty conclusions in 2022 that we reached in 2017. Given the role that this survey plays in guiding the public’s perception of the military and DoD policy, these faulty conclusions can have catastrophic consequences.

Below, I’m going to talk about how the data from the 2017 “Workplace and Equal Opportunity Survey” were misinterpreted. As I will show: (1) the survey’s measure of harassment and discrimination is likely to leave us with a vastly inflated sense of how much racial harassment and discrimination are going on in the military; (2) even with this inflated sense, there is actually far less racial harassment and racial discrimination in the military than in the rest of American society; and (3) contrary to the implications of most reports on the survey, the group most overrepresented in claims about racial harassment and discrimination is African Americans. Once again, the goal of this discussion is to help us avoid drawing the same faulty conclusions when the 2022 data are analyzed. There is no “right” way to analyze the data but we must be aware of how the choices we make shape the stories we tell.

Reporting on the 2017 “Workplace and Equal Opportunity Survey” Results

How much racial and ethnic harassment and discrimination is going on in the military? In a 2021 article entitled “Long-withheld Pentagon survey shows widespread racial discrimination, harassment,” Reuters describes the results of the 2017 “Workplace and Equal Opportunity Survey” in the following way:

“Nearly a third of Black U.S. military servicemembers reported experiencing racial discrimination, harassment or both during a 12-month period, according to results of a long-withheld Defense Department survey that underscore concerns about racism in the ranks.”

The article focuses exclusively on the experiences of African Americans serving in the military (Latinos and Asian Americans receive only one passing reference) and frames the importance of the issue around anti-black discrimination (e.g. “Concerns about racial discrimination in the military - the largest U.S. employer - have taken on new urgency over the past year, as America undergoes a nationwide reckoning on racism”).

Only a month after the results of the Workplace and Equal Opportunity Survey were reported, Secretary of Defense Lloyd J. Austin III ordered commanders and supervisors to take an operational pause for one day to discuss extremism in the ranks with their service members. The memo announcing this “stand-down” alluded to the survey’s results. Specifically, the memo said, “Service members, DoD civilian employees, and all those who support our mission, deserve an environment free of discrimination, hate, and harassment.” The survey’s results were also referenced by other military officials justifying the stand-down. In an article entitled “Deep-rooted racism, discrimination permeate US military,” Pentagon spokesman Major César Santiago claimed that “we know that far too many service members indicate they experience discrimination.” These results, in short, have had a significant impact on the discourse and policymaking surrounding the American military.

Issue #1: How the Survey Inflates Racial Harassment and Discrimination Rates

The Choice to “Lump” Rather than “Split”

Does the report suggest a problem so large that dramatic interventions (e.g. temporary stand-downs) are required? The report describes the survey’s approach to measuring the “racial/ethnic harassment” rate in the following way: “To be included in the Estimated Past Year Racial/Ethnic Harassment Rate, members had to indicate that in the past 12 months they perceived experiencing at least one of the following 12 racial/ethnic workplace behaviors (i.e., the respondent indicated being “uncomfortable, angry or upset” by a behavior) by someone from their military workplace prohibited by EO policy.”

A similar approach was used to determine the “racial/ethnic” discrimination rate. As the report explains, “To be included in the Estimated Past Year Racial/Ethnic Discrimination Rate, members had to indicate that they perceived experiencing at least one type of differential treatment as a result of their race/ethnicity in the past 12 months.”

The first thing to point out is that the survey “lumps” together many different kinds of experiences into the broad categories of “harassment” and “discrimination.” All of the experiences are important but it is not obvious why they need to be “lumped” together and it’s not obvious that the lumping should be done in the way the analysts choose to do it. Let’s take the 12 “harassment” items. The “harassment” number includes behaviors that range from telling a joke and seeing a picture or flag being displayed to physically assaulting someone because of their race or ethnicity. These are clearly very different kinds of experiences. Given their differences, why not just list each item separately and avoid lumping them altogether? If some lumping must be done, why not create separate categories for verbal harassment, visual harassment, and physical harassment?

This becomes an essential point for how the results are interpreted. Lumping a large number of items together necessarily produces a larger topline number (i.e. the rate is inevitably larger because more items are included and each additional item will expand the number of people experiencing the behavior in question). Larger topline numbers are going to produce more public concern and more pressure for a policy response.

Below is the summary graph guiding most of the reactions to the survey. It is the source of the Reuters claim that “nearly a third of Black U.S. military servicemembers reported experiencing racial harassment.”

Even with this bizarre aggregation, does the report suggest “that minority service members face rampant discrimination” (as Senator Kirsten Gillibrand claimed in response to the survey)? Not exactly. The results show that 29.3% of blacks experienced “harassment” and 11.5% experienced “discrimination.” The results also show that 19.5% of Latinos experienced harassment and 6.5% experienced discrimination. Put differently, the overwhelming majority of blacks and Latinos did not experience a single instance of harassment (71.7% and 8.05%, respectively) or discrimination (88.5% and 93.5%, respectively) during the last year. Framing the issue around those experiencing harassment and discrimination is a choice that colors our understanding of the issue.

Consider what the story might look like, however, if we “split” the overly broad category “harassment” (which aggregates experiences with a large number of very different verbal, visual, and physical behaviors). As the data in the table below shows, there was not a single harassing behavior that was experienced by more than 13% of minority servicemembers. The most commonly experienced forms of harassment were hearing a stereotype, joke, or offensive racial or ethnic term (each experienced by approximately 10% of minority servicemembers over the last year). Being threatened or physically assaulted because of one’s race was reported by a fraction of a percent (.7%) of minority servicemembers.

Without aggregating these individual responses into the broad “harassment” category, the discourse around this issue might be much different. Rather than summarizing the survey’s results for African Americans as Reuters did, the reporting might say something like “18% of Black servicemembers heard a stereotype about their group but fewer than one in 20 were excluded from an activity because of their race.” It would also be possible to point out that less than one in ten Hispanics and Asian Americans reported experiencing nearly every behavior listed.

This is not to say that lumping is necessarily problematic and always inferior to “splitting.” Again, I could imagine differentiating verbal harassment (e.g. insults, jokes, and stereotypes) from physical harassment (e.g. “offensive action,” “exclusion” from activity, physical assault). While we don’t have the data needed to provide a precise estimate of the harassment rates under this alternative categorization, it is very likely that the verbal harassment category would include less than 25% of African Americans and the physical harassment category would include less than 5% of African Americans. To be clear, these numbers are concerning and do require a response. Avoiding overly broad lumping, however, might help the public better contextualize the source and scale of the problem.

The Time Frame Choice

The second thing to point out here is that the “racial/ethnic harassment and discrimination rates” (which receive all of the attention in these discussions) are influenced greatly by the selected time frame of the study. Over a long enough period of time, everyone will feel that they have been harassed or discriminated against in some way. Longer time frames will lead to higher percentages of people reporting that they experienced harassment or discrimination. For example, asking about harassment or discrimination over the last day, week or month would yield much smaller numbers than asking over a year. The point is simply that asking the question in this way returns a much higher number than other ways of asking the question and higher numbers create more pressure for policy interventions.

How “Important” is the Racial Harassment and Discrimination Experienced by Servicemembers?

As the discussion above indicates, the “lumping” of jokes with physical assault within the broader category of “harassment” obscures the fact that verbal harassment is vastly more common than physical harassment (and is, I would argue, of a less significant nature). Fortunately, the survey includes a measure of how “important” servicemembers believe the harassment and discrimination they experienced was. Specifically, the survey asks respondents who experienced harassment or discrimination to “consider the “One Situation” or set of related events or behaviors that was the most offensive or egregious to them.” The results are presented below:

As the results show, 48% of those experiencing harassment or discrimination did not think the behavior was important enough to warrant reporting the situation to DoD authority. Importantly, there were almost no racial differences in perceptions of how “important” the harassment and discrimination was (46% of whites, 46% of African Americans, and 49% of Latinos claimed their experiences were not “important” enough to justify reporting). If we were to reconsider the overall harassment and discrimination rate by filtering out those instances where the victims did not feel the harassment or discrimination was “important,” we would end up with rates that were essentially half of that reported in the media. To be more precise, we would conclude that 15% of African Americans and 10% of Hispanics experienced “important” harassment. Once again, this is unacceptable but paints the problem in a much different light than communicated in most of the media’s reporting.

Issue #2: There’s Far Less Harassment and Discrimination in the Military than Elsewhere

It is difficult to assess these rates of harassment and discrimination in the absence of some baseline of comparison. Obviously, we might hope for a perfect world in which 0% of service members of every race report experiencing harassment or discrimination. Putting this unattainable goal aside for the moment, however, are there any useful comparisons for how often African Americans and Latinos experience harassment and discrimination outside of the military? How many civilians report harassment and discrimination in their workplace?

Unfortunately, the items used in the Workplace and Equal Opportunity Survey to measure harassment are idiosyncratic to the report and have not been used by other organizations. There is, however, a survey conducted in 2020 by the Kaiser Family Foundation that provides a useful reference point for thinking about discrimination. The Kaiser Family Foundation survey asked respondents whether “there was a time in the last 12 months where they felt they were treated unfairly in the following places because of their race: in a store where they were shopping, their place of work, in dealings with the policy or while getting health care for themselves or a family member.” These questions share a time frame with the Workplace and Equal Opportunity Survey (the last 12 months) and, more importantly, include a question that covers workplace discrimination (providing a perfect comparison for the military).

As the results of the Kaiser Family Foundation survey, show that 28% of African Americans and 17% of Latinos reported being treated unfairly in their workplace. This means that African Americans and Latinos were approximately 2.5 times as likely to report discrimination in the civilian workplace than African Americans and Latinos in the military. The military is apparently far better at managing race relations than the rest of society. Once again, the numbers should be brought down. The military, however, should be lauded, not criticized, for the less harassing and less discriminatory work environment they’ve managed to create.

Issue #3: The Surprising Sources of Harassment and Discrimination

Most of the reporting on the Workplace and Equal Opportunity survey is framed around white supremacy and anti-black harassment and discrimination coming from whites in the military. For example, the previously mentioned AP news report entitled “Deep-rooted racism, discrimination permeate US military,” focuses entirely on instances of anti-black harassment and discrimination. Is this an accurate depiction of harassment and discrimination based on the data?

The Workplace and Equal Opportunity survey attempts to identify who the initiators of harassment are. The 2017 report states that whites are 55.9% of the total DoD population. Blacks are 14.6%. Hispanics are 17.7%. Asian are 4.5%.

Who is overrepresented among the offenders/initiators of racial harassment and discrimination? The question can be answered by comparing the overall percentage of whites, blacks, Latinos, and Asians in the military with the percentage of whites, blacks, Latinos, and Asians that were identified as initiating the harassment or discrimination. Subtracting the percentage who were identified as an initiator from the percentage of the overall force gives us a measure of overrepresentation:

As the data above shows, whites were overrepresented as an offender for every group but their own. While they are 55.9% of servicemembers, they accounted for 68% of the cases of harassment against blacks, 67% of the cases against Hispanics and 64% of the cases against Asians. This means they were overrepresented by 9-13% in interacting with out-groups. A similar degree of overrepresentation with respect to outgroups held for Latinos and Asians. Hispanics were overrepresented by 6%-12% and Asians were overrepresented by 7%-9%.

African Americans, however, were significantly more likely to initiate harassment and discrimination against outgroup members. Blacks are 14.6% of servicemembers but were identified as the offender in 60% of cases of harassment against whites, 35% of cases against Hispanics, and 38% of cases against Asians. This means they were overrepresented by 20%-45%.

The Future? Thinking Ahead to the 2022 Results

The Ever-Evolving Nature of Offense: The “Latinx” Cautionary Tale

Harassment and discrimination are, to some degree, in the eye of the beholder. As Eric Kaufmann has argued, “racism contains an important socially constructed component.” Indeed, there are large differences in perceptions of discrimination based on ideology (with liberals far more likely than conservatives to report being discriminated against) and social media use (with those reporting high levels of social media use much more likely to report being discriminated against). As ideas about race and ethnicity evolve, so too will the number of people who feel offended or aggrieved by certain kinds of interactions. Even the exact same set of behaviors, measured at two different times, can produce vastly different levels of reported harassment and discrimination.

It is very likely that more servicemembers will report experiencing harassment and discrimination in 2022 than in 2017. Sensitivity to mistreatment on racial and ethnic grounds has increased dramatically during the last five years. Even attempts to become more sensitive to racial or ethnic harassment can backfire. Consider, for example, the use of the term “Latinx.” Originally conceived as a way of being more “inclusive,” the term is now understood to be offensive to significant segments of the Hispanic community. For example, only 2 percent of Hispanics polled in 2021 reported referring to themselves as “Latinx.” More importantly, 40% said the term “Latinx” bothers or offends them to some degree.

This attempt at inclusivity, however, could produce a massive spike in the harassment rate the 2022 Workplace and Equal Opportunity Survey finds among Hispanics. For example, President Biden (commander-in-chief) has repeatedly used the term “Latinx” when referring to Hispanics. Similarly, Secretary of the Navy Carlos Del Toro has used the term in his public communications:

If every Latino servicemember heard Biden’s rhetoric or read the tweet from the Secretary of the Navy, approximately 40% might report that they were exposed to “an offensive racial or ethnic term.” Exposure to a single speech by Biden or a single tweet from the Secretary of the Navy could give the appearance of high (and rapidly increasing) levels of racial or ethnic harassment (coming from the military’s leadership no less). There are certainly many other examples of jokes, stereotypes, and racial or ethnic terms that could inflate the numbers in a way that might give a misleading perception of how much harassment and discrimination is occurring in the military.

Once again, these are serious issues that need serious attention. Misinterpreting what the data tell us, however, is likely to lead to bad and potentially harmful policy choices. Let’s hope we learn from the lessons of the 2017 survey.

Subscribe now

On October 31, the Supreme Court will hear oral arguments in two cases about the use of race-based affirmative action at Harvard University and the University of North Carolina. In deciding these cases, the Court may choose to end the decades-long practice of allowing universities to grant preferential treatment to some applicants on the basis of their race or ethnicity.

As part of the Supreme Court’s consideration of a case, interested parties (frequently referred to as “amicus curiae”) are allowed to author and submit briefs that offer information, expertise, or insight relevant to the issues in the case. In August, Google, Meta Platforms Inc., Apple Inc., HP, Inc. and more than 70 other major US-based companies filed an amicus curiae brief with the US Supreme Court in support of affirmative action programs being challenged at Harvard and the University of North Carolina.

In this post, I’m going to examine the arguments presented in this amicus curiae brief. As I will demonstrate below, the brief attempts to use social science research on workplace diversity to make the case for the continued use of race-based admissions in American universities. Unfortunately for Amici (i.e. the companies behind the brief), the cited social science research makes an incredibly shallow and weak case for affirmative action. Despite the fact that decades of research have attempted to demonstrate the benefits of racial diversity in the workplace, the brief can only summon a small collection of flawed studies that fail to demonstrate: (1) increasing diversity above the levels currently found in Amici’s companies will produce meaningful business benefits; and (2) allowing universities to continue with an illiberal and unpopular policy is the only (or even the best) means of achieving higher levels of racial diversity.

As always, my hope is that this post will provide a model for how to critique arguments that use social science research as a means of advancing controversial positions on matters of public policy.

Affirmative Action in the People’s Court

In the 1960s and early 1970s, affirmative action was justified on the grounds that it would serve as a kind of reparation for the harms imposed by America’s long-standing system of institutionalized discrimination against African Americans.1 Even at the outset, this argument rested on a somewhat shaky moral foundation given that the beneficiaries of preferential treatment were rarely the actual victims of discriminatory action and that those who paid the costs were often not those who engaged in the discriminatory conduct. Nevertheless, this “reparational” or “remedial” approach to affirmative action for African Americans maintained a sufficient level of public support to preempt a strong enough public backlash to overturn it.

Beginning in the late 1970s, however, the rationale for affirmative action began to evolve. Driven by the rapid demographic change arising from the 1965 Immigration Reform Act, affirmative action could no longer be framed as a remedial action required to make up for past or present discrimination. Indeed, recent immigrants could not claim to be the victims of past discrimination in the United States nor claim to be the descendants of such victims. As a result, affirmative action’s justification shifted to the idea that the country’s institutions needed to “represent” the country’s racial and ethnic diversity. Importantly, affirmative action as “representation” could proceed in the total absence of benefits for African Americans (e.g. if they were already sufficiently or over-represented in an institution, they may not benefit from that institution’s system of preferential treatment).

Once this shift from “reparational” to “representational” affirmative action happened, public support for the programs cratered. Survey experiments from 1998, for example, found that the American public was far more opposed to affirmative action as a means of increasing representation (82%) than there were to affirmative action as means of making up for past discrimination (59%):

At this point, it was clear that the idea that racial and ethnic diversity should be “represented” was an insufficiently compelling normative argument for most people. In order to provide an alternative foundation for “representational” affirmative action, the policy’s advocates began attempting to construct an empirical case for diversity. The idea here was to show that racial and ethnic diversity was not just a moral imperative but also a business and educational imperative. Businesses and universities that did not diversify would miss out on a range of benefits available only to those who did. The rationale for affirmative action, in other words, would be relocated from the realm of morality and ideology to the realm of science and pragmatism.

The institutional demand for research demonstrating diversity’s benefits has produced literally thousands of studies published on diversity EVERY YEAR. One article, for example, found that there were more than 45,000 papers on gender, racial, and disability diversity within the field of management in 2017 alone (see the figure below).

As I have pointed out previously, these studies are likely to dramatically overstate the benefits of diversity while minimizing its costs because of the rigidly enforced ideological monoculture within academia. Indeed, as Eric Kaufmann’s survey of faculty members shows, only half of professors in the social sciences and humanities would oppose a campaign to fire an academic who argues that “diversity is a negative.” Questioning the value of diversity (even if this questioning is motivated by a good faith interpretation of empirical data) has become disqualifying for faculty positions at contemporary American universities.

It is in this context, that we get the Supreme Court’s reconsideration of affirmative action and the social science-heavy brief authored by the tech companies. Currently, the idea that race should be a factor in who is accepted to a university (and who is not) is very unpopular. In March 2022, for example, Pew found that 74% of Americans said race “should not be a factor” in college admissions decisions (only 7% thought it should be a “major factor.”).

Similarly, a 2022 survey by Grinnel University found that 68% of Americans oppose universities “taking race into account in admissions decisions.” More importantly, the survey found no differences in support between whites and non-whites.

Blue Rose Research conducts surveys in which respondents are presented with the Democratic argument in favor of a given policy and the Republican argument. They can then pick which argument they find most persuasive. Using this method across 113 different issues, Blue Rose found that affirmative action is one of the least popular Democratic Party policy positions (only outpolling “lowering the voting age to 16”).

The public feels similarly skeptical about affirmative action in the workplace. While most value workplace diversity, very few want their employers to consider race when making decisions about hiring & promotion. According to the most recent Pew data, 74% want decisions about hiring and promotions to rely solely on “qualifications” “even if it results in less diversity.”

Perhaps most importantly, affirmative action has been voted down nearly every time it has appeared on the ballot. According to Richard Hanania, affirmative action has been voted on at the state level nine separate times between 1996 and 2020. It has been rejected 8 of those times (including, most recently, in 2020 in California):

All of this is just to say that the corporations are using their brief to promote a deeply unpopular policy. Their argument avoids presenting moral and ideological justifications for affirmative action and, instead, builds on decades of research designed to make a pragmatic and empirical case for allowing universities to discriminate on the basis of race.

This kind of argument, however, is only as strong as the academic research it cites. As I will show below, the business case for diversity is poorly supported by the research cited in the amicus brief and, more importantly, the evidence offered by Amici cannot be used to convincingly argue for the necessity of affirmative action in university admissions. Despite decades of research attempting to demonstrate otherwise, there’s no good evidence for allowing universities to discriminate on the basis of race.

How Much and What Kind of Diversity is Needed?

I’m going to focus my analysis of the tech company’s amicus brief on “Argument I - Racial and Ethnic Diversity Enhances Business Performance.” More precisely, I will discuss the studies presented in “Part A: Diverse Teams Make Better Decisions.” The authors of the brief begin this section in the following way:

this brief explains a wide variety of research-backed, tangible ways in which racial and ethnic diversity improves business. Empirical studies confirm that diverse groups make better decisions thanks to increased creativity, sharing of ideas, and accuracy. And diverse groups can better understand and serve the increasingly diverse population that uses their products and services. These benefits are not simply intangible; they translate into businesses’ bottom lines. For these reasons, it is no surprise that companies are investing substantially in diversity initiatives—a concrete acknowledgment of the value of a racially diverse workforce and leadership structure to business success.

In other words, the goal of this section is to provide “research-backed” evidence for the claim that diverse workforces produce benefits for employers.

Before evaluating the research presented in the brief it is worth asking why the diversity of an employer’s workforce is relevant to whether universities consider race and ethnicity when admitting students. According to the brief’s authors, the only way companies can recruit the kind of diverse workforce that would produce the previously mentioned benefits is if universities are sufficiently diverse themselves. As the brief states, “the “future of American business depends on universities’ ability to select racially and ethnically diverse student bodies.” Underneath this claim is an assumption about university admissions in the absence of racial preference; namely, that there will not be enough racial and ethnic diversity for businesses to hire a diverse workforce unless universities demonstrate a large preference based on an applicant’s race.

Is there any evidence that companies would struggle to recruit a workforce that is diverse enough to produce the workplace benefits they claim to seek? Answering this question requires specifying exactly how much and what kind of diversity is needed to generate positive business outcomes (e.g. to “make better decisions”). For example, does a team composed of whites, African Americans, and Asian Americans but no Hispanics produce the same amount, fewer, or more benefits than a team composed of whites and African Americans alone? Similarly, does a team that is perfectly representative of the American public in its racial composition (i.e. 65% white, 12% African American, 17% Hispanic, and 5% Asian American) perform better, worse or the same as a team that is majority-minority (e.g. 40% white, 60% non-white)? Without knowing what quantity and quality of diversity is minimal (never mind optimal) for producing “better decisions,” we can’t assess whether the companies even have a problem to be solved through more workplace racial diversity in the first place.

How Many College Graduates Do Amici Need?

A useful place to start in thinking through this issue is to identify how many college graduates Amici need to hire each year. The largest tech companies to sign on to the brief are Apple, Google, Meta, and HP. They employ 147,000 people, 135,301 people, 71,970 and 53,000 people, respectively. Since 2017, Apple has hired approximately 7,000 new employees a year, Google has hired approximately 20,000 new employees a year, Meta has hired approximately 13,000 new employees each year and HP, Inc. has hired an average of 0 new employees each year since 2017:

Collectively, then, the four largest tech companies in the country need approximately 40,000 new employees a year. Assuming that all of these positions require a college degree, will universities produce enough African American and Latino college graduates (who are the focus of Amici’s attention in the brief) to fill these needs if affirmative action is ruled unconstitutional?

How Many African American and Latino College Graduates Will There Be in the Absence of Affirmative Action?

Assessing the exact impact of ending affirmative action on enrollment and, more importantly, graduation rates across racial groups is difficult. Data from the state of California (which repealed affirmative action in 1996 with the passage of Proposition 209) suggests that drops in enrollment might be initially large but rebound somewhat over time. As the data below show, African American and Latino enrollment in the UC system declined in the immediate aftermath of Proposition 209 but increased slightly in recent years.

The most severe estimate suggests that African American enrollment would decline by 25% and Latino enrollment would decline by 19% in the absence of affirmative action:

Using this extreme estimate, how would the total number of African American and Latino graduates each year change if affirmative action stopped? In 2020, there were 532,720 Hispanic and 452,760 African American college graduates. If we assume there will be no improvement over time in African American and Latino enrollment after the end of race-based affirmative action (which is a highly questionable assumption), we would expect to have 10,121 fewer Hispanic and 11,717 fewer African American college graduates each year. Put another way, even with the end of affirmative action there would be nearly a million Hispanic and African American college graduates each year.

So, Will There Be Enough “Diversity”?

Even if the tech companies wanted to hire only African Americans and Latinos every year (to the complete exclusion of whites and Asian Americans), it seems they could easily meet their yearly hiring needs of 40,000 employees with the approximately 1 million new college graduates a year. If they wanted to adopt a slightly different standard for overall hiring, however, the task of hiring African Americans and Latinos would become even easier. Representing the national-level demographic composition, for example, would require hiring only 12,000 African American and Latino college graduates from the pool of nearly 1 million (1.2%). In short, with or without affirmative action, it seems that tech companies should have the labor supply they need to increase racial diversity in their workforces.

While it might be objected that the end of affirmative action would only hurt enrollment at “elite” or “select” universities and these are precisely the kinds of students Amici are attempting to recruit, raising this objection just begs the question of why Amici cannot adjust their standards for hiring if they see diversity as providing such value. Indeed, all of the arguments presented in Amici’s brief are about the importance of racial diversity alone (not about the importance of having diversity from the subset of students who attended elite universities). Corporations could hire without regard to education entirely and still accrue the benefits of diversity if they truly believed the research cited in the brief.

It is essential to point out here that private companies are already much freer to pursue affirmative action policies than universities under current federal law. In United Steelworkers v. Weber (1979), the Supreme Court ruled that Title VII of the 1964 Civil Rights Act “did not foreclose private race-conscious affirmative action plans.” This ruling gave the private sector far more flexibility than the government and academia in designing affirmative action programs. So, the question becomes: why can’t Amici use their own affirmative action programs to address diversity deficiencies rather than arguing for the government to implement a deeply unpopular policy? The solution to the problem of diversity in Amici’s workforce, in other words, might just be to have Amici lower their standards.

None of this answers the question, however, of exactly how much and what kind of diversity these companies are seeking to achieve. Throughout the brief, Amici argue for more racial diversity than they currently have in their workforces. For example, the authors of the brief write, “increasing racial and ethnic diversity throughout Amici’s workforces is the right thing to do.”

The idea that diversity needs to be increased (rather than simply maintained) raises the question of how diverse the companies are right now. To help answer this question, I tracked down data about the demographics of Google’s workforce:

As the graph shows, Google’s workforce had a significant amount of racial diversity in 2021. As the graph also shows, the composition of Google’s workforce has changed quickly in recent years. Between 2014 and 2021, the percentage of Google employees who are Asian American increased by more than 10% and the percentage of African Americans more than doubled. Hiring data from Google's 2022 "Diversity Annual Report" suggests these trends will continue to accelerate. According to the report, Google “achieved our best year yet for hiring from underrepresented communities” by increasing the percentage of hires going to individuals from the “Black+,” “Latinx+,” and “Native American+” communities. These increases have led to a significant decline in the percentage of “White+” employees at Google but have not led to a decline in “Asian American+” employees:

Google’s reference to “underrepresented communities” suggests they are evaluating their company’s diversity by comparing it to the country’s overall racial composition (i.e. 65% white, 12% African American, 17% Hispanic, and 5% Asian American). Given that the overwhelming majority of positions at Google select for a college education, a more appropriate baseline for comparison would probably be the racial composition of college graduates (see the graph of college degree holders over 25 years old below):

Regardless of the comparison (overall population or college graduates), it is clear that the company’s workforce vastly overrepresents Asian Americans (42.3% of current employees and 46.3% of new hires but only 5% of the population and 10.8% of college graduates). It is also clear that Google’s workforce only slightly underrepresents African Americans (5.3% of current employees and 9.4% of new hires while constituting 12% of the population and 9.1% of college graduates) and Hispanics (6.9% of current employees and 9.0% of new hires while constituting 17% of the population and 10% of college graduates). The most significant source of underrepresentation is among whites (who make up more than 60% of the population and hold 70% of the college degrees but account for only 48% of Google’s workforce and only 40.2% of their new hires).

Generally speaking, Google and the other companies who signed on to the brief are free to evaluate their employee’s racial composition against any standard they like. In the context of this section’s discussion of affirmative action, however, Amici is signing on to a very specific claim about racial diversity; namely, that they require more African American and Hispanic employees in order to perform better as companies (in the brief’s language, they need to “increase racial diversity” because “diverse groups make better decisions thanks to increased creativity, sharing of ideas, and accuracy”).

If it is to make a compelling case for affirmative action, then, the brief needs to present evidence that: (1) companies with more than the current level of African American and Hispanic representation (e.g. Google’s 5% of employees who are African American and 6% that are Hispanic) and less than the current level of white and Asian American representation (e.g. Google’s 50% of employees who are white and 42% who are Asian American) perform better than companies with different demographic racial compositions; and (2) the only way to achieve a workforce with this racial composition is by allowing universities to discriminate against applicants on the basis of their race.

To make a long story short, the brief does not do either of these things. More specifically, it fails to show any evidence at all that companies will benefit from moving beyond their current racial composition and it fails to demonstrate that affirmative action is a necessary path to achieving workforce diversity (independent of its benefits). Here, I will demonstrate this by reviewing four of the seven peer-reviewed studies the brief cites in support of its claim that “empirical studies confirm that diverse groups make better decisions thanks to increased creativity, sharing of ideas, and accuracy.” As I show below, each one of these studies has a set of problems that make it impossible for Amici to use them as part of an argument in favor of affirmative action in university admissions.

Study #1: “Do Pro-Diversity Policies Improve Corporate Innovation?”

Problem: The Study’s Measure of Diversity Conflates Race and Gender

The first cited study is Mayer et al.’s (2018) article “Do Pro-Diversity Policies Improve Corporate Innovation?” I will allow the authors of the brief to summarize this research in their own words:

“researchers found that companies with “a range of policies and characteristics” that indicate strength in diversity (including racial, ethnic, ability, and LGBT+ diversity) were positively associated with the “number of new product announcements per R&D dollar spent by a firm…..Companies with pro-diversity policies are more innovative.”

In order to measure the extent to which a company has “pro-diversity” policies, Mayer et al. use the MSCI’s Environmental, Social, and Governance (ESG) data. Mayer et al. do not use every variable contained in the MSCI ESG data. Instead, they choose the subset of variables related to gender, racial, and sexual orientation diversity (i.e. those classified as “Diversity Strengths” and “Diversity Concerns” by MSCI). To be exact, they rely on the following measures:

In the MSCI ESG data, “each firm is assigned a score of 0 or 1 for each strength or concern based on whether the firm meets the screen criteria.” Mayer et al. “calculate the net strengths of diversity as the difference between Diversity strengths and Diversity concerns (DIV = Diversity strengths – Diversity concerns).” Larger numbers on this measure indicate that a firm has more “pro-diversity” policies. Using this measure, Mayer et al. conclude that “pro-diversity policies are positively related to the number of new product announcements per R&D dollar spent by a firm.”

Does this study provide evidence that companies with more racial diversity perform better than companies with less racial diversity? No. The primary problem is that Mayer et al'.’s measure of diversity is that many of the items draw no distinction between race and gender. Consider the first “diversity” measure: “The company’s chief executive officer is a woman or a member of a minority group.” Bracketing the question of what constitutes being a member of a “minority group” under this measure, we have no way of determining whether all of the diversity included here is merely an artifact of companies hiring white women (not members of a “minority group”). The same problem applies to “Diversity Strength” measures B (“Promotion”), C (“Board of Directors”), E (“Women and Minority Contracting”), and H (“Employment of Underrepresented Groups”). Two measures under the rubric of “Diversity Concerns” even explicitly focus on the representation of women to the exclusion of members of “minority groups” (B and C).

In a nutshell, the problem with lumping race and gender together in this way is that it becomes impossible to determine whether the relationship between “pro-diversity” policies and firm performance is driven by racial representation alone, gender representation alone, or a combination of both. It is possible, for example, that gender representation is driving all of the pro-business effects that Mayer et al. describe as relating to “pro-diversity” policies more broadly. Representation for members of “minority groups” might have no effect or negative effects on business performance. We just can’t say given the measures used by Mayer et al.

Remember, Amici are attempting to argue for the necessity of affirmative action in university admissions by showing evidence that their businesses will benefit from increasing levels of racial diversity among their employees (e.g. this section of the brief is entitled “Racial and Ethnic Diversity Enhances Business Performance”). More directly, they are trying to make the case that increasing the percentage of African American and Hispanic employees above the current levels found in their companies will lead to “increased creativity, sharing of ideas, and accuracy.” Mayer et al.’s inability to distinguish between the effects of “diversity through female representation” and the effects of “diversity through racial representation,” however, means that we can’t conclude more racial diversity will lead to higher levels of performance. In other words, Mayer et al. fails to demonstrate that there is a business need for affirmative action.

There is another problems with using Mayer et al.’s measure of “pro-diversity” policies to make the case that companies need more diversity in their workforce overall; namely, that none of the 14 items in the overall measure is about overall workforce diversity. As the table above shows, the ESG data track racial and gender diversity for CEOs (Item A), Boards of Directors (Item C) and contracted firms (Item E) but not for the workforce overall. The measure of “Promotion” (Item B) can be scored as “pro-diversity” if a company “has made notable progress in the promotion of women or minorities” regardless of whether the levels of diversity are high or low among the workforce broadly. Even the measure labeled “Employment of Underrepresented Groups” does not tell us anything about the workforce diversity of a company (it only tells us about whether a company “makes efforts to recruit women and minorities”). In other words, the cited study fails (once again) to provide the evidence Amici needs it to provide in order to make a compelling business case for increased racial diversity.

Problem: The Study Measures Diversity as a Feature of HR Policy, not as a Feature of the Workforce

One last point deserves mention here: even if we believe that the “racial plus gender” diversity combination measure indicates a business benefit for racial diversity alone, Mayer et al.’s findings can’t help Amici make a convincing argument that the only way to achieve a workforce with more African American and Hispanic employees is by allowing universities to discriminate against applicants on the basis of their race. In fact, Mayer et al.’s research suggests that company policies, not the diversity of the broader applicant pool a company draws from, are the driving force in producing business benefits. Put differently, rather than arguing for the necessity of affirmative action, Mayer et al.’s work actually suggests that affirmative action is not needed at all.

Let me explain. If a business believed the results obtained in Mayer et al., they would want to advance their interests by adopting more “pro-diversity” policies. This would mean they should attempt to promote women and racial minorities currently working in the company (Items A and B), provide more “work-life benefits” to their employees (Item D), contract at least 5% of their business to women or minority-owned businesses (Item E), provide domestic partner benefits (Item G) and make an “effort” to recruit women and minorities (Item H). These adjustments would boost their “pro-diversity policy score” and, presumably, boost their performance.

Notice that all of these things are completely within a company’s control and require no changes to the eligible pool of potential applicants graduating from America’s universities. If Google wanted to, for example, they could make all of these changes tomorrow without ever needing to consider how universities accept or reject people. If universities were to end affirmative action or expand it, nothing would change from the perspective of companies attempting to follow Mayer et al.’s findings. In short, the brief’s reference to Mayer et al. actually undermines their claim that businesses can only accrue diversity benefits by allowing universities to discriminate on the basis of race.

Study #2: “Ethnic Diversity and Creativity in Small Groups”

Problem: The Study Compares “Equal Representation” Groups to “All-White” Groups

Next, the brief discusses McLeod et al.’s (1996) “Ethnic Diversity and Creativity in Small Groups” as evidence that diversity improves a company’s “creativity and innovation.” I’ll let the brief’s authors frame this study in their own words first:

“Another study illustrating this principle [how racial diversity improves creativity and innovation] asked small groups to brainstorm ideas to draw more tourists to the United States, and groups with the highest levels of racial diversity generated ideas that judges deemed “significantly more feasible . . . and more effective . . . than the ideas produced by the homogenous groups.”

Let’s examine whether McLeod et al.’s research really shows that diversity improves creativity and innovation. The study included 135 students (94 men and 41 women) from “a large midwestern university.” From this group, the researchers created 18 “ethnically diverse groups” (which included one white, one African American, one Hispanic, and one Asian American participant) and 16 “all-Anglo groups.” The groups were asked to “spend 15 minutes generating as many ideas as possible to get more tourists to visit the United States.” Each idea was then graded on a five-point scale by two judges “in the travel industry” for their effectiveness and feasibility. In addition, the researchers assessed how “attracted” each subject was to their group (measured through items such as “Group members seem to like each other”). The results are shown in the table below:

As even this short discussion suggests, the brief’s summary is misleading. The study did not attempt to assess whether “groups with the highest levels of racial diversity” would generate better ideas. Groups were either “diverse” (i.e. had one member of each racial group) or “all Anglo.” As such, the study cannot speak to the relationship between levels of diversity in a group and its performance. We cannot know, for example, whether a group that was racially representative of the population (i.e. 68% white, 12% African American, 16% Hispanic, and 5% Asian American) would produce superior work to an all-white, African American, or Hispanic group. More importantly, the study only tells us about the performance of groups that are completely equal in their racial composition relative to groups that are completely racially homogenous.

This caveat is essential for thinking about the applicability of this study to real-world workplace environments. As shown above, the tech companies are already racially diverse (i.e. not “all-Anglo”). As such, we cannot know whether more diversity in their workforces (e.g. moving from 6% African American to 12% African American or moving from 5% Hispanic to 16% Hispanic) would have any effect on their employee’s creativity and innovation. Any increase in diversity would have to be assessed for its impact on creativity and innovation relative to a workforce that is already quite diverse (see above). Indeed, even if companies somehow engineered their workforces to match the racial parity examined in McLeod et al. (an impossible task given the current demographics in the United States), the odds that this small increase in diversity would produce a measurable difference in creativity and innovation are nearly zero. In short, McLeod et al.’s study does not actually advance Amici’s claims about the necessity of diversity and does not, therefore, bolster Amici’s argument that universities should use affirmative action in their admissions processes.

I should point out here that McLeod et al. are actually quite explicit about the inability of their study to address these issues. As they clearly explain in the paper, a major limitation of their choice to include equal numbers of whites, African Americans, Hispanics, and Asian Americans in each “diverse” group is that they cannot identify “how much diversity is enough to achieve measurable differences in performance.” What they do claim, however, is that:

“the work on the influence of minority opinion would suggest that even small numbers of people whos opinions are radically different from the majority’s can stimulate creativity.”

Put another way, to the extent that McLeod et al. make any recommendations about the amount of diversity needed to produce creativity and innovation benefits, it is very likely that Amici has already exceeded it. There is no evidence in this paper, therefore, that affirmative action in university admissions is needed to help these companies.

Problem: The Study Emphasizes Diversity’s Costs as Well as Its Benefits

According to McLeod et al.’s study, the observed benefits for diversity are vanishingly small and come with non-trivial costs. As the table above shows, the “diverse” group outperformed the “all-Anglo” group on effectiveness and feasibility by a mere .2 points each on a five-point scale. What’s more, the benefits were likely a function of unique features of the experimental context. As the authors explain, this experiment did not require the kind of conflictual negotiations typically found in workplace contexts requiring a single solution to a problem because it “required the subjects to simply accept all the solutions offered by other group members.” The benefits of diversity, in other words, may never appear in the real-world workplaces Amici operates.

More importantly, these small and context-specific benefits came with a cost. As McLeod et al. explain:

“members of heterogenous groups may have had more negative affective reactions to their groups than did members of homogenous groups…..The results of this and other studies have suggested that increased heterogeneity may also be related to negative affect, communication difficulties and turnover.”

In other words, the relatively small benefits in work production came at the relatively large expense of group cohesion and satisfaction. Contrary to Amici’s claim, therefore, this study doesn’t actually provide clear evidence that diversity helps business. The net impact is more mixed. If diversity is not necessarily great for business, there is no strong rationale for affirmative action in university admissions.

Study #3: “Multicultural Experience Enhances Creativity: The When and How”

Problem: The Study Measures Diversity as a Feature of a Slideshow, not as a Feature of a Workforce

Next up in the brief is Leung et al.’s (2008) study on multicultural slideshows and Cinderella stories. As the brief’s authors write:

“One example [of how diversity produces creativity] is a study in which white undergraduate students watched a slideshow—either about Chinese culture, American culture, American and Chinese cultures in juxtaposition, or American-Chinese fusion culture—before writing a Cinderella story for Turkish children. Researchers found that those who had watched the slideshow juxtaposing American and Chinese cultures or the slideshow depicting American-Chinese fusion culture wrote more creative stories than other participants—providing “direct evidence for the causal role of exposure to a foreign culture in creative performance.”

Leung et al.’s study included 65 “European American undergraduates (29 females; mean age = 19.05 years) who “participated in two experimental sessions in exchange for course requirement credit.” There were between 12 and 14 students in each of the study’s five experimental conditions. The results (with creativity scored by “judges” on a seven-point scale) are shown below:

There are obvious concerns about external validity here - including the tiny and unrepresentative sample (65 white undergraduates at a single university), the bizarre task given to the experimental subjects (writing a version of the Cinderella story for a hypothetical group of Turkish children) and the highly individualized, non-competitive and inconsequential nature of the experimental context (writing the story alone with low stakes rather than working collectively on a high stakes project).

Independent of these concerns, the brief’s reference to this work is confusing given that it actually makes a strong argument against the need for racial diversity in the workforce. The causal variable in this study is the “slideshow” (producing changes in “creativity”). In other words, this study is about the efficacy of slideshows and says nothing about the business-related benefits of people from different racial or ethnic backgrounds working together. More importantly, if creativity can be increased among an all-white group simply by exposing them to a slideshow about multiple “cultures,” there is no need to hire based on diversity at all. Indeed, apparently with the right kind of multicultural, multimedia presentation, companies can boost creativity in a way that is (presumably) easier to implement than engineering their workforces to achieve a particular racial or ethnic composition.

Given that this study suggests that companies do not need any actual racial or ethnic diversity in their workforce, there is no need to allow universities to take race or ethnicity into account in their admissions process. Remember, the companies are arguing that universities need to employ affirmative action in order to produce the diverse workforce that helps the companies succeed. If companies do not actually need diversity to succeed, there are no grounds for allowing a divisive and illiberal program to continue.

Study #4: “Ethnic Diversity Deflates Price Bubbles”

Problem: The Study Emphasizes Diversity’s Costs as Well as Its Benefits (Again)

The brief then cites one of the most widely referenced studies in public-facing arguments about diversity: Levine et al.’s “Ethnic Diversity Deflates Price Bubbles.” Levine et al. conduct an experiment that compares fake stock trades in four ethnically homogenous “markets” of 24 college students in Kingsville, Texas and five ethnically “diverse markets” of 30 college students in Kingsville, Texas. According to the brief:

The study found that traders in ethnically homogenous markets were ‘significantly less accurate’ in making pricing decisions ‘and thus more likely to cause price bubbles.’ Traders in homogenous markets [were] more likely to accept offers that are above true value.

Falling into the same trap that they did in their discussion of McLeod et al.’s study, the authors of the brief never mention the potential costs of diversity. Their silence on diversity’s costs is particularly strange in the context of referencing Levine et al.’s work because Levine et al. are incredibly explicit about the potentially high costs of diversity. In fact, the entire theoretical justification for their work is that observable racial and ethnic diversity enhances skepticism, distrust, and competition. As they write:

"Ethnic diversity facilitates friction. This friction can increase conflict in some group settings, whether a work team, a community or a region. Conversely, ethnic homogeneity may induce confidence, or instrumental trust, in others' decisions (confidence not necessarily in their benevolence or morality, but in the reasonableness of their decisions, as captured in such everyday statements as “I trust his judgment”).

According to Levine et al., this “friction” can produce positive consequences in some contexts and incredibly damaging consequences in others. Specifically:

In modern markets, vigilant skepticism is beneficial; overreliance on others’ decisions is risky…Such friction, however, can cause conflict and complicate collective decisions. The challenge, then, is in establishing rules and institutions to address ethnic diversity and its effects. Without them, conflict can be destructive; with them, diversity can benefit the collective."

In other words, previous research indicates that racial and ethnic diversity might make things much worse in team and non-market settings (like the ones found in tech companies). If diversity has significant drawbacks, affirmative action becomes a less appealing policy option.

Problem: The Study Compares “All Hispanic” Groups to “Diverse” Groups

Levine et al.’s study “created homogeneous markets by including only participants that were Latinos.” In “diverse markets,” Levine et al. explain, “at least one participant was of a numerical minority ethnicity.” The “minority ethnicity” might have been African American, Asian American, or white. The authors, however, never say and they are never explicit about how many “minority ethnicity” members were included in their five “diverse” trading markets.

If we assume that each “diverse market” in Levine et al.’s experiment had only one “minority ethnicity” participant, the study may have had a total of 49 Hispanic and 5 non-Hispanic college students in their sample. It is even possible that the study did not include a single white, Asian American, or African American participant. Put differently, the study created a highly artificial situation in which a national demographic minority (Hispanics) constituted a huge experimental majority.

Obviously, Levine et al.’s findings cannot tell us anything about the amount or ratio of diversity required to begin accruing the “accuracy” benefits described in the brief. Although specific details of market composition are never described, it is possible that only one “minority” member was included in each of the “diverse” markets. Would accuracy in pricing assessments increase or decrease with more “diversity” (e.g. if two or three “minority” members were included in the market instead of one)? Would the presence of more than one kind of racial or ethnic diversity (e.g. one African American, one white, one Asian American instead of just one African American) lead to more or less accuracy in perceptions? The study does not say because the study cannot say. In other words, Levine et al. cannot advance Amici’s claim that they need to “increase” the amount of racial diversity in their workforces.

Problem: The Study Assess Diversity Among Strangers and not Among Teams

Participants in Levine et al.’s study were strangers. They were alerted to the racial and ethnic homogeneity or diversity of their market by being asked to sit in a waiting room for a few minutes with other study participants. As the study’s authors explain:

“While waiting, participants were allowed to speak with one another and had many opportunities to perceive others’ ethnicity, a categorization process that is typically automatic and highly accurate. Once six participants arrived, an experimenter assigned them to separate cubicles, each of which was equipped with a networked computer.

The study makes no attempt to distinguish the effects that diversity might exert on the decision-making of complete strangers who share nothing in common other than their ethnicity and the effects that diversity might exert on the decision-making of people who share some common, non-ethnic set of identities, experiences, or responsibilities. This distinction is important because, according to Levine et al., the benefits of diversity are a direct and exclusive consequence of the distrust that diversity produces among strangers. In their words:

“Ethnic diversity was valuable not necessarily because minority traders contributed unique information or skills, but their mere presence changed the tenor of decision making among all traders…Diversity facilitates friction. In markets, this friction can disrupt conformity, interrupt taken-for-granted routines, and prevent herding. The presence of more than one ethnicity fosters greater scrutiny and more deliberate thinking, which can lead to better outcomes.”

There is a considerable body of research suggesting that shared experiences, mutual dependence, and the formation of superordinate identities can reduce friction between members of distinct groups. My strong suspicion is that Amici invest considerable time, energy, and resources in team-building and training exercises designed to reduce friction between employees. Presumably, this reduced friction would also reduce the benefits of diversity indicated by the study. If diversity has fewer benefits than indicated by the study, the case for affirmative action in university admissions is considerably diminished.

Problem: The Study Looks at Diversity’s Effects in the Context of Individual Competition and not in the Context of Collective Cooperation

After having a chance to observe the racial and ethnic composition of the group, participants in Levine et al.’s experiment were isolated from each other. While “trading” during the experiment, “participants could not see each other or communicate directly.” The experimental context, in other words, was highly individualized and decisions about how to act were made alone and in the absence of any interpersonal communication.

What’s more Levine et al.’s experiment sets up zero-sum competition between its individual participants by creating a simulated stock market (i.e. an individual participant earns more money only by taking money from another participant in the form of relatively beneficial asset “trades”).

The authors of the brief ignore these crucial features of Levine et al.’s experimental design and, instead, include the results under the heading “Diverse Teams Make Better Decisions.” It should go without saying that group-based, cooperative information processing, and decision-making operate according to fundamentally different dynamics than individual, competitive information processing and decision-making.

Consider Solomon Asch’s famous conformity experiments (pictured above). When left alone to determine which lines matched in length, experimental subjects reported the correct answer in more than 99% of cases. When asked to make the same identification in the context of a group giving incorrect answers, however, 37% of respondents gave an incorrect answer in order to conform to the rest of the group. Taking a finding about how individuals make decisions in artificially sequestered environments and claiming that it reveals something about how “teams” process information in a corporate environment is a stretch. Put differently, Levine et al.’s experiments cannot provide Amici with the evidence they need to make a strong argument in favor of university affirmative action.

Conclusion

I’ve focused here on the use of social science research to justify race-based affirmative action in university admissions. If the social science research was to successfully bolster Amici’s argument that “diverse groups make better decisions,” the unit of analysis in the cited studies (i.e. the entity about which we are making claims) should have been the “work team,” the causal variable should have been the “team’s racial and ethnic composition” and the outcome variable should have been “decision quality” (with attention to the costs and benefits resulting from diversity). Further, the study’s findings should have identified exactly how much and what kind of racial diversity is needed to produce the benefits Amici seeks. Finally, the cited research should have indicated that allowing affirmative action in university admissions is the only path to achieving these benefits. We should have expected, in other words, for the brief to discuss research that meets a very specific set of criteria.

As I describe above, however, this is not what the brief chooses to do. Instead, it offers up a number of studies about units of analysis (e.g. individuals, corporations), causal variables (e.g. watching movies, company-level diversity policies), and outcomes (e.g. innovation, benefits but not costs) that have nothing to do with the claim that “racially diverse teams make better decisions.” More importantly, the cited research does not suggest the quantity or quality of diversity needed or indicate that affirmative action is required to achieve more diversity. In fact, many of the studies suggest that the companies may already have sufficient levels of racial diversity. In other words, the brief cites studies that fail to make a compelling case for diversity’s business benefits. As a result, there is no compelling case for affirmative action in university admissions.

It’s important to point out here, however, that the weaknesses in the studies cited in this section of the brief (“Part A: Diverse Teams Make Better Decisions”) do not undermine arguments or evidence presented in other parts of the brief. Most importantly, it does not undermine the claim that universities need to be diverse because it educates people in a particular way. That claim, which is explained in the brief, needs to be examined separately.

Subscribe now

On April 9, 2021, U.S. Secretary of Defense Lloyd Austin directed “immediate actions to counter extremism in the military.” According to the New York Times, not all extremism was equally concerning to the Secretary of Defense. “Right-wing extremism” would be the focus of the Department of Defense’s counterextremism efforts.

A key part of the counterextremism campaign was to establish the “Countering Extremist Activity Within the Department of Defense Working Group” (CEAWG). In December 2021, the working group issued its findings. According to the report,

“The CEAWG incorporated quantitative and qualitative data from internal and external experts to help inform this report, including: A briefing from the University of Maryland National Consortium for the Study of Terrorism and Responses to Terrorism (START).”

…

“START data indicated recent spikes in the instances of domestic violent extremism, and an uptick in veteran participation in these cases.”

In conjunction with the release of the CEAWG report, START released a research brief based on an extension of their Profiles of Individual Radicalization in the United States (PIRUS) dataset. In the brief ( “Extremism in the Ranks and After”), the research team explicitly identified their desire to assist with Secretary Austin’s goal of pursuing a “department-wide stand-down to address the problem of extremism in the ranks.” As the authors write, the brief was

“intended to help in this effort by providing information on the military service backgrounds of individuals who committed extremist crimes in the U.S. from 1990 through the first eleven months of 2021.”

Upon its publication, the START brief was covered in a number of media outlets (e.g. Defense One’s article entitled “At Least 458 U.S. Crimes Tied to Extremism Involved Veterans, Active Duty Troops”. ):

On July 20 of this year, the Senate Armed Services Committee called on the Defense Department to halt its programs to prevent and root out extremism in the ranks. In a report accompanying the Senate’s National Defense Authorization Act, the committee states that “spending additional time and resources to combat exceptionally rare instances of extremism in the military is an inappropriate use of taxpayer funds, and should be discontinued by the Department of Defense immediately.” Shortly thereafter, the START team released a new research brief (still titled “Extremism in the Ranks and After”). This new brief increased the estimated number of criminal military extremists from 458 to 545 between 1990 and July 2022.

The Plan for this Post

A central goal of this newsletter is to encourage skepticism about the growing use of social science research in the policymaking process. My motivating concern is that we might be implementing costly, inefficient and, possibly, counterproductive policies on the basis of bad or misappropriated social science research. When I see policymakers like the Secretary of Defense citing research from places like START, I am not confident that good decisions will follow.

I’m also concerned about second-order effects of using imprecise social science research in our policymaking process. Specifically, I’m worried that headlines like the ones above might needlessly erode confidence in America’s military. The military is one of the only institutions in the country that a majority of the public still trusts. This trust will not last long, however, if Americans come to see the military as a haven for criminal extremists (right-wing or otherwise). Before leveling charges of widespread extremism, we need to be sure to have good evidence to back these accusations up. Sadly, we rarely do.

I will start by saying that I’m actually agnostic about whether military extremism is increasing or decreasing. I’m also agnostic about whether military extremism is, at the moment, a sufficiently large problem to warrant aggressive actions to root it out. What I am not agnostic about, however, is whether the data used to justify the Secretary of Defense’s efforts (and media reports about them) provide a sufficiently solid empirical foundation for deciding any of these issues one way or the other. To state the matter simply, there are too many unanswered questions about the data at the center of these internally-focused counter-extremism debates. We should be wary of letting it guide us.

In this post, I am going to walk through five questions raised by the START’s PIRUS project data on extremism and point out how we might be drawing bad conclusions about both the overall level of and over time changes to military extremism. My hope is that this discussion will help us improve our data and improve the way we incorporate that data in military decisionmaking.

Question #1: How much extremism is too much extremism?

The START PIRUS team claims, “This project includes all known cases of individuals who served in the U.S. military and committed extremist crimes in the United States from 1990 through 2021.” The July 2022 brief includes the following chart summarizing the yearly number of “US extremists with military backgrounds” between 1990 and July 2022. As Graph #1 shows, there were 545 criminal military extremists identified by their data collection efforts over the last 32 years:

Graph #1: US Extremists with Military Backgrounds (1990-2022)

There are nearly 19 million veterans currently living in the United States. In 2021, there were 2.1 million active-duty and reserve U.S. Department of Defense members, including officers and enlisted personnel. In other words, in 2021 (which had nearly three times more criminal military extremists than any other year), the “extremism” rate was .0007% (or 7 in one million). In more typical years (e.g. 2010, 2015, 2016, 2018, 2019), the yearly extremism rate among current and former members of the military was .0001% (or 1 in one million).

While most “one in a million” problems are not worth our serious attention, some are (e.g. nuclear plant meltdowns, nuclear weapons attacks, plane crashes, etc.). Is criminal military extremism one of these issues? There are certainly high profile instances throughout American history were military extremism has proven catastrophic. In 2009, for example, Nidal Malik Hasan killed 13 people and injured more than 30 others in the Fort Hood mass shooting. In 1995, Timothy McVeigh, a former army soldier, killed 168 people (including 19 children) when he and his accomplice Terry Nichols bombed the Alfred P. Murrah Federal Building in Oklahoma City.

These are powerful examples of the dangers of extremism from individuals with a military background. But how typical are these extreme examples of extremism? The PIRUS data from 1990 to 2018 can help provide some context (as I will discuss below, comparable data is not available from the project after 2018).

Between 1990 and 2018, the PIRUS data includes information on 148 criminal military extremists. 15 of the 148 individuals with a military background were “charged, arrested, or indicted” for misdemeanor crimes (e.g. trespassing, vandalism, etc.). A majority (75 out of 148) of the criminal military extremists identified in the PIRUS data were not “violent” (i.e. did not “actively participate in ideologically motivated operations/actions that resulted in causalities/injuries or clearly intended to result in causalities/injuries (but failed), or charged with conspiracy to kill or injure but were interdicted in the plotting phase”). In other words, a significant portion of criminal military extremists appear to avoid engaging in the most extreme forms of criminal activity.

Another way of contextualizing the number of criminal military extremists over the last 32 years is to compare extremism by members of the military to extremism from other demographic groups. The most obvious comparision here is with Islamic extremism. According to the PIRUS data, there were 251 criminal Muslim extremists between 1990 and 2018. Given that the overall number of individuals with a military background during this period was at least six times greater than the number of Muslims living in the United States (roughly 20,000,000 compared to roughly 3,000,000), we get much higher rates of criminal extremism in the Muslim community than among those with a military background (see Graph #2):

Graph #2: Extremism Rates among Muslims and Military Members (1990-2018)

Even higher rates of criminal extremism can be found among some first generation immigrants. The PIRS data show 30 first generation Pakistani immigrants arrested, charged, or indicted for criminal extremism between 1990 and 2018. In 2006, when there were only 300,000 Pakistani immigrants living in the country, six were arrested for criminal extremism (a rate of .002% or 20 in one million). The PIRUS data indicate there were 27 first generation Somali immigrants arrested, charged, or indicted for criminal extremism between 2007 and 2018, producing a criminal extremism rate as high as .003% (30 in one million) in 2015.

What about the problem highlighted by the New York Times and the Secertary of Defense: “right-wing” extremism among current and former military members? Of the 148 criminal military extremists included in the PIRUS data between 1990 and 2018, 87 (58.7%) were radicalized by “right-wing” movements. Despite the fact that there were only 5,895 Muslim Americans serving in the military during this period, 35 of the 148 criminal military extremists in the PIRUS data (23.6%) were radicalized by “Islamist or jihadist” movements.

I am NOT arguing here that these differential rates justify more scrutiny of Muslim Americans, first generation Pakistani immigrants or first generation Somali immigrants. I am also NOT pointing this out to suggest the need for “counter extremism” policies targeting any of these groups. On the contrary, the rates discussed here are to suggest how tiny ALL of these problems are and, further, how strange it is to treat them so differently in our thinking about extremism. Any discussion of extremism (and efforts to combat it) must start with a clear eyed look at how rare the problem of military extremism truly is.

Question #2: Who is an Extremist?

Who is an “extremist”? I will start here by saying that defining extremism has proven itself to be a nearly impossible task. Consider the tautological hopelessness of the standard dictionary definition of “extremism”: