Civil Liberties — Free Speech in a Polarized Era

Six categories, six different stories

The lead figure of this chapter is the simplest one. Plotted on the same axes, the six "should this person be allowed to speak in your community?" series do not move together.

The macro-story of the figure is that Americans have, on average, grown more tolerant. The "Cold War tolerance bargain" of the 1970s — about half the country willing to extend speech rights to communists and homosexuals — is no longer how Americans answer these questions. A solid majority now extends those rights, in some cases approaching consensus. On the categories the 1970s left tolerated least (communists, homosexuals, atheists), the country has caught up.

The micro-story is that the recent leg of the series, post-2018, is not a continuation of the long upward trend. On racist speech specifically, tolerance has fallen by about 15 points in six years. On Muslim-cleric speech, tolerance has been essentially flat at a low level for nearly two decades. And the partisan structure underneath the topline movement is category-specific — each party tolerates speech it finds politically friendly and restricts speech it finds politically threatening, and the sorting of which speech each party owns has become sharper with every decade.

The partisan-selectivity table

The clearest summary of the chapter's central finding is a simple six-row table: for each speaker category, the contemporary Democrat–Republican gap and which party is the more tolerant one.

The six gaps sort themselves into two clean groups. Republicans are more tolerant of racist and militarist speech — speech whose substantive politics overlaps with elements of the contemporary Republican coalition's cultural register, or at minimum is not threatening to it. Democrats are more tolerant of atheist, communist, homosexual, and anti-American Muslim speech — speech aligned with secular, internationalist, or LGBTQ political commitments familiar to the contemporary Democratic coalition.

This is the chapter's central finding. There is no longer a "free speech" party in the American electorate, in the sense of a coalition that defends all unpopular speech by principle. There are two coalitions, each defending the unpopular speech it can imagine itself making, and each more willing than the other to restrict the unpopular speech it cannot.

Racists: the lopsided sorting

The most dramatic of these stories — and the one that most clearly is a recent change rather than a long-stable feature — is racist speech.

Two things in Figure 3 are worth slowing down on.

The first is that Republicans have always been slightly more tolerant of racist speech than Democrats — by 3 to 7 points, in nearly every wave from 1976 through 2018. The contemporary 14-point gap is not the emergence of a partisan pattern that did not exist; it is the widening of a pattern that has been there throughout. The classic civil-libertarian framing of "Republicans were the speech-tolerant party, Democrats discovered hate speech" is too clean. Democrats were the slightly less speech-tolerant party on this specific question from the beginning.

The second is that the widening since 2018 is driven almost entirely by Democratic movement. In 2018, 56% of Democrats said racists should be allowed to speak; in 2024, 39% do. A 17-point drop on the Democratic line, in six years. The Republican line dropped too — from 62% in 2018 to 53% in 2024 — but by half as much. The 2018 → 2024 widening of the gap from 6 to 14 points is the arithmetic of a slow Republican decline meeting a sharp Democratic one.

A college-education view of the same question contains an unexpected finding.

On most attitudinal questions covered in this book, the college-educated population is the more tolerant, more cosmopolitan, more pro-rights group. On racist speech specifically, the college-educated population has moved most sharply against the classic civil-libertarian position. This is consistent with what one would expect if the social and cultural milieus through which "hate speech is not free speech" became a salient frame were disproportionately college-educated — campus debates, urban liberal political culture, the editorial pages of college-aligned newspapers. The non-college population, less exposed to those frames, has held closer to the position it took in the 1970s, even as it has come down somewhat too.

Homosexuals: the consensus that arrived

The mirror image of the racist-speech story is the homosexual-speech story. Both are large; both involve cohort change and within-cohort movement; both have settled into different equilibria.

The age-convergence on homosexual speech is one of the cleanest examples in this book of an attitude reaching ceiling. The 1973 18–34 cohort started at 78%; in 2021 the 65+ cohort sits at 92%. Some of this is generational replacement (the people who were 65+ in 1973 are not the people who are 65+ in 2021), but cohort tracking suggests substantial within-cohort movement as well. The Americans who came of age in the 1950s and 1960s — and who in 1973 expressed the lowest tolerance for homosexual speech in the entire battery — have, on average, ended up close to a uniform pro-tolerance position. The transformation is large, complete, and not visibly partisan.

The ideology view confirms what the age view suggests.

By 2021, even Conservatives were at 86% saying homosexuals should be allowed to speak in the community. The 2021 partisan gap on the by-party version is 4 points (D=93%, R=89%). Of all six categories in this chapter, homosexual speech is the only one where the contemporary partisan and ideological gaps are smaller than they were in the 1970s. The sorting on this issue has run in reverse — both parties moved upward, with the more skeptical party moving farther, and the consensus is now nearly complete.

Muslim clergy: the only category that has not risen

The "speak" battery's most recent addition — anti-American Muslim clergy, first asked in 2008 — is the one category where tolerance has neither risen nor reached a clear partisan equilibrium. The series sits in the low 40s and has stayed there.

The Muslim-clergy series is a useful counterweight to the chapter's broader story. Tolerance for the speech of an explicitly anti-American religious figure is low in both coalitions — only one in three Republicans and slightly under half of Democrats extend it. This is a question where the substantive content of the speech — opposition to the United States, in the post-9/11 context — overrides the abstract free-speech principle for most respondents in both parties. The partisan order, on a question where neither party reaches majority tolerance, is volatile and not obviously meaningful. Whatever else one wants to say about American free-speech attitudes, "anti-American religious speech is restricted by both parties' base voters" is the cleanest summary of the 2008–2024 picture.

The institutional gradient: speak, teach, library

The GSS asks the same six categories about three different settings: should they be allowed to speak in the community, should they be allowed to teach in colleges, and should their books be allowed in public libraries. The three settings produce different levels of tolerance — usually but not always in the same order.

For the categories with the largest pro-tolerance majorities (homosexuals, atheists), the rank is speaking highest, then library books (an anti-censorship framing), then teaching (an institutional-gatekeeping framing). Teaching is consistently the highest bar: Americans are more willing to let someone speak in public than to let them teach in a college classroom.

For racist speech, the rank is reversed. Anti-censorship norms — libraries should not remove books — produce 56% support for keeping racist books on the shelves. Public speaking comes in at 46%. Teaching comes in at 35%. The framing of the question matters: "remove a book" sounds like censorship; "should this person teach" sounds like an institutional endorsement. The American public answers those two framings about the same speaker very differently.

What Figures 8 and 9 together show is that the framing of a civil-liberties question matters at least as much as its content. The same respondents who would not let a racist teach (35% support) will, on average, let a racist's book stay in the library (56% support). The same respondents who would not let an anti-American Muslim cleric speak in the community (40% support) would, by similar margins, not let him teach. Civil-liberties opinion in 2024 is not a uniform principle being applied across cases. It is a series of context-specific judgments, each of which the parties make slightly differently.

What this chapter changes

Three things to take from this chapter into the rest of Part II.

First, the "free speech" coalition does not exist anymore — and may never have. The proposal that became this book described the trajectory as moving from a classic civil-libertarian position (defend all speech) toward partisan selectivity (defend our speech). The data shows partisan selectivity has always been there, in the form of small consistent gaps on the four 1970s-era categories. What has changed is the magnitude. The contemporary 14-point gap on racist speech and 13-point gap on Muslim-cleric speech are several times the size of the 1970s-era partisan gaps. The pattern is older; the amplitude is recent.

Second, the diploma divide is not the master variable on every cultural question. On racist speech specifically, college-educated tolerance has fallen more than non-college tolerance — the opposite direction from the standard diploma-divide story used elsewhere in this book. Whatever moved on this question between 2018 and 2024, it moved through the educated, liberal, urban segment of the country first. That is a reminder that "the more educated are the more tolerant" is a useful generalization across many domains but not a universal law; on specific items, the educated segment can move sharply against a previously consensus position.

Third, the institutional framing of the question matters. "Should this person be allowed to speak" produces one answer. "Should this person be allowed to teach" produces a substantially lower answer. "Should their book be allowed on the library shelf" produces something in between, with an anti-censorship overlay. The American public does not hold one civil-liberties position that gets applied uniformly across institutions. It holds a series of context-specific positions whose ordering is itself a feature of the data — and whose partisan signatures differ depending on which framing is being asked.

The next chapter turns from civil liberties to sexual morality, where the same underlying generational and ideological dynamics — older cohorts liberalizing, younger cohorts at ceiling, the partisan gap widening even as both parties move — produce one of the most-cited "great sorting" charts in American social science.