Americans consistently express broad levels of support for free speech and free expression. For example, 87% of respondents in a recent survey reported that freedom of speech is “very” or “extremely” personally important (Knight Foundation 2022). Moreover, this support seemingly transcends party lines, with 91% of Republicans and 88% of Democrats endorsing this importance.Footnote 1 Yet, there are reasons to be skeptical that broad levels of support translate to on-the-ground tolerance of free speech. There are numerous historical examples of Americans’ willingness to selectively withdraw First Amendment protections to those deemed undeserving, particularly along racial, ethnic, and ideological dimensions (King and Smith Reference King and Smith2005). Additionally, people are far less likely to tolerate and extend rights to members of their “least-liked” group, especially when threatened (Lambert and Chasteen Reference Lambert and Chasteen1997; Marcus et al. Reference Marcus, Sullivan, Theiss-Morse and Wood1995; McClosky and Chong Reference McClosky and Chong1985; Nelson, Clawson, and Oxley Reference Nelson, Clawson and Oxley1997).
There are reasons to be skeptical that broad levels of support translate to on-the-ground tolerance of free speech.
Therefore, we are left to wonder what to make of these strong endorsements that Americans continue to give regarding First Amendment protections. One reason to be skeptical about these declarations of support for free speech is that these endorsements lack tradeoffs and are socially desirable. These low-stakes features could inflate levels of overall public support for civil rights protections. In practice, however, questions related to rights typically do not ask whether a constitutionally protected right should exist. Instead, debates often center on the scope of those rights and/or the groups to whom those protections extend. Creating an additional complication is that beliefs about freedom of speech—and other related rights—can be politicized along partisan lines. Democrats and Republicans could differ significantly in defining First Amendment rights and the values they attach to them based on the object they seek to defend. For example, the previously cited Knight Foundation (2022) report noted that partisans differ substantially in their belief about whether spreading misinformation or hate speech online should be a protected form of speech.
To test the limits of Americans’ commitment to free speech, we relied on two survey experiments that were designed expressly to assess whether broad commitments to speech change when tradeoffs or costs to that speech are introduced, as well as whether those speech protections extend to partisan groups. The first experiment focused on broad support for free speech; the second concerned views of free speech on college campuses. The studies yielded four broad conclusions: (1) in the absence of tradeoffs, support for free speech was high; (2) Republicans expressed greater support for free speech than Democrats; (3) the introduction of tradeoffs altered support for free speech and did so similarly for Democrats and Republicans; and (4) support for free speech did not depend on whether partisan in-groups or out-groups engaged in the speech.
The survey experiments were included in the Knight Foundation–Ipsos Study from the Knight Free Expression Research Series (Knight Foundation 2022). This was an omnibus project that convened teams of scholars who were interested in studying questions related to Americans’ views of free speech and expression. The survey was nationally representative and also included an oversample of nonwhite adults and college students. The experiments described in this article included approximately 2,500 participants (Carlos, Sheagley, and Taylor Reference Carlos, Sheagley and Taylor2022).Footnote 2
EXPERIMENT 1: GENERAL SUPPORT FOR FREE SPEECH
The first experiment was designed to test two questions: (1) whether support for free speech changes in the presence of tradeoffs, and (2) whether the source is a partisan in-group or out-group member. Respondents were randomly assigned to one of the conditions listed in table 1. In the control condition, participants read a statement about general opposition to restricting free speech rights. In the “Democrats Say” or “Republicans Say” condition, participants read the language used in the control condition as well as the language that highlighted a tradeoff or cost to that speech (i.e., the use of hurtful language) and whether the speech was coming from Democrats or Republicans.Footnote 3 The dependent variable was measured on a 5-point scale ranging from “1–Strongly Disagree” to “5–Strongly Agree.” We coded the dependent variable such that higher values corresponded to greater support for free speech.
Table 1 Experimental Manipulations for Experiment 1
Results
We began by examining views of free speech among respondents who were randomly assigned to the control condition. This allowed us to assess whether there were baseline differences in general support for free speech between Republicans and Democrats in the sample. For this analysis, we used ordinary least squares to regress the outcome on a categorical measure of party identification (i.e., Republican, Democrat, Independent, or Other).Footnote 4
Democrats and Independents in the sample were significantly less likely than Republicans to agree with the statement: “I would never support restricting my or someone else’s freedom of speech.” The difference between Republicans and Democrats was 0.67 point on the 5-point scale, corresponding to slightly more than half of a scale point. As a reference, the standard deviation of the outcome among participants in the control condition was 1.1. Thus, the difference between Republicans and Democrats corresponded to 0.61 standard deviation of the outcome. The difference between independents and Democrats was smaller (0.25) but also statistically significant (p = 0.03).
Next, we analyzed whether support for speech was conditioned by the presence of costs and whether respondents explicitly associated speech protection with members of one of the political parties. For this analysis, we regressed the free-speech outcome on indicators for the experimental condition, the respondent’s party identification, and the interaction of these two variables. The predicted mean levels of the dependent variable with 95% confidence intervals are shown in figure 1.
Figure 1 Mean Level of Support for Free Speech by Experimental Condition and Respondent’s Party Identification
Figure 1 reveals that support for free speech was highest in the control conditions. Among Republicans and Democrats in the sample, support was lower when respondents were presented with tradeoffs—in this case, the ability for certain groups to say hurtful things online. Furthermore, support for free speech did not depend on whether co-partisans or opposition partisans were the sources of hurtful speech online. For example, among Republican participants, the difference in support for free speech when Democrats versus Republicans say hurtful things was 0.03 (p = 0.77).
Taken together, these results reveal that Republicans, Democrats, and Independents have different baseline levels of support for restricting their own or someone else’s freedom of speech. However, all groups of respondents were less supportive of free speech when they were informed that hurtful things were being said online. Finally, we did not find any evidence that the partisan source altered support for free speech.
EXPERIMENT 2: IDEOLOGICAL SPEECH ON COLLEGE CAMPUSES
The second experiment focused on the more specific debate about freedom of speech and ideological bias on college campuses. This study replicated the same general framework as before: that is, a baseline condition in which free speech was viewed without any tradeoffs and two treatments that included tradeoffs. In this case, those tradeoffs involved whether specific ideological perspectives were given priority by a speaker on campus. Table 2 summarizes the manipulations. The outcome was measured on the same 5-point scale reported in the first experiment and coded so that higher values reflected opposition to censorship.
Table 2 Experimental Manipulations for Experiment 2
Results
We began by examining opposition to censorship among respondents who were randomly assigned to the control condition. This allowed us to assess baseline differences in general support for free speech on college campuses between Republicans and Democrats in the sample. We used the same approach as in the first analysis. Similar to Experiment 1, Democrats and Independents were significantly less likely to endorse the statement: “I do not believe that any form of speech on a college campus should be censored.” The difference between Republicans and Democrats was 0.49 point on the 5-point scale (p < 0.01).
Next, we regressed the censorship outcome on indicators for the experimental condition, the respondent’s party identification, and the interaction of these two variables.Footnote 5 Results are shown in figure 2.
Figure 2 Mean Level of Opposition to Censorship on College Campuses by Experimental Condition and Respondent’s Party Identification
The tradeoff treatments resulted in greater opposition to censorship compared to the baseline/control conditions. Thus, attaching a conservative or liberal source of speech increased opposition to censorship compared to the baseline condition that lacked ideological information. This may suggest that conservative/liberal bias conditions narrow the scope of censorship. Whereas respondents in the control condition could have been thinking of any form of censorship, those in the tradeoff conditions may have been thinking more narrowly about political perspectives.
Consistent with the findings from the first experiment, there were no differences in censorship views within the party of respondents based on whether conservatives or liberals engaged in the speech. Among Republican identifiers, the difference between the liberal versus conservative source was 0.17 (p = 0.13); the difference among Democrats was 0.15 (p = 0.1). Finally, Independents in the sample were more opposed to censorship of liberal perspectives than they were of conservative perspectives (difference = 0.39, p < 0.001).
CONCLUSION
At the onset of the project, we were guided by two general beliefs. First, in the absence of tradeoffs or costs, participants would express high levels of support for free-speech protections. Second, we expected this support to diminish when it was made clear to participants that opposition-party members also would receive these protections while engaging in specific forms of speech. In summary, we expected that Americans’ high levels of support for free speech would come with a crucial qualifier about the type of that speech and who engaged in it.
Instead, we found a complicated set of results that highlight reasons for optimism as well as concern for advocates of free speech. The experimental results demonstrate that Americans are willing to limit First Amendment protections for everyone (themselves included) when they deem the speech to be hurtful. However, they appear less ready to limit First Amendment protections for strictly partisan or ideological purposes. For example, as Experiment 2 shows, respondents were relatively more opposed to censorship on college campuses when they heard different ideological viewpoints.
Mirroring recent research that shows that Democrats are less supportive of First Amendment protections than Republicans (Alvarez and Kemmelmeier Reference Alvarez and Kemmelmeier2018; Chong Reference Chong1993; Chong, Citrin, and Levy Reference Chong, Citrin and Levy2021; Crawford and Pilanski Reference Crawford and Pilanski2014), we observed baseline differences in the levels of support that Democrats and Republicans have for speech generally and specifically on college campuses. The results of these experiments are surprising in that these views do not appear to center on the specific partisan source of speech but rather on baseline differences among different partisans in the survey.
However, important caveats are needed. Despite baseline differences in support for First Amendment protections, Democrats and Republicans expressed high overall levels of support for free speech. Partisans on both sides of the aisle also were significantly more likely to oppose speech when it was framed as hurtful. Respondents from both parties also were less supportive of censorship on college campuses when it was strictly about ideological disagreements rather than a baseline of general censorship.
To summarize, whereas Democrats and Republicans may begin in different places in terms of their overall support for free speech, partisan tropes about Democrats’ willingness to engage in censorship and Republicans’ willingness to support a person’s right to say whatever they want are simplistic. Both groups reduced their support when speech could be hurtful, and both groups were opposed to censorship that is couched as simply ideological. Moreover, neither group was more supportive of speech when it came from their side rather than the other side.
This study highlights the importance of expanding our understanding of how different groups define free speech. For example, were Democrats and Republicans thinking of the same types of actions when they evaluated speech that is hurtful? We suspect not, given the history of racial discrimination in the United States, which is rooted—at least in part—in violent and threatening speech and the racial composition of both political parties (Hutchings and Valentino Reference Hutchings and Valentino2004; Omi and Winant Reference Omi and Winant2014; Saavedra Cisneros et al. Reference Saavedra Cisneros, Carey, Rogers and Johnson2022). Thus, it is important that future research contributes more nuance to and understanding of how different groups along partisan and racial dimensions view free speech.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the PS: Political Science & Politics Harvard Dataverse at https://doi.org/10.7910/DVN/EP63RI.
CONFLICTS OF INTEREST
The authors declare that there are no ethical issues or conflicts of interest in this research.
