Introduction
Political scientist E.E. Schattschneider once said, in a fight “the spectators are an integral part of the situation, for, as likely as not, the audience determines the outcome of the fight” (Reference Schattschneider1975, 2). On January 6, 2021, a fight broke out at the nation’s capital between protestors challenging the outcome of the 2020 presidential election, and those set to certify the results. “We fight like hell. And if you don’t fight like hell, you’re not going to have a country anymore,” President Donald Trump said only hours before Joe Biden was to be certified the next President (Jamieson, Levendusky, and Pasek Reference Jamieson, Levendusky, Pasek, Jamieson, Levendusky, Pasek, Lance Holbert, Renninger, Ophir and Walter2023). That fight, over the 2020 election results, resulted in the breach of one of America’s most well-recognized norms—the peaceful transition of power. While public opinion data on January 6th from Pew Research and Data for Progress shows the American public to hold overwhelmingly negative views of the insurrection (Gramlich Reference Gramlich2022; Springs and Blank Reference Springs and Blank2024), a substantial 10–11 percent of participants in both surveys declined to share their feelings. These nonresponders—the politically silent—represent an important cohort of Americans who remain neglected in conversation and scholarship on democratic norms in America.
Using the 2022 Health of Democracy Survey, we test whether questions about the January 6th insurrectionists provoke high levels of nonresponse and measure the independent effects of demographics, racial attitudes, support for democratic norms, support for Christian Nationalism, or media consumption. If, as Schattschneider’s quip tells us, the opinion of the audience can determine the outcome of the fight, then understanding the politically silent matters to our understanding of Americans’ commitment to democratic norms. That political scientists have been relatively mute with respect to the politically silent is due in part to a methodological norm where scholars simply deal with nonresponses by eliminating the missing values, known as listwise deletion (ex. Converse Reference Converse2006 [1964]), or impute the missing values following techniques such as mean substitution or multiple imputations (Little and Rubin Reference Little and Rubin2002; Rubin Reference Rubin1987). The limitation of listwise deletion is a loss of valuable information and possible selection bias (King et al. Reference King, Homaker, Joseph and Seheve1998). Multiple imputation, or the process of generating multiple simulated datasets where missing data is imputed conditional on the observed data, can similarly introduce bias when missing data is not random. Our analysis of data about the January 6th insurrection provides evidence that ignored non-random missingness constrains our ability to interpret Americans’ feelings toward insurrectionists and limits our understanding of the way the intersections of race, racial attitudes, and gender impact public opinion research.
We argue that the Health of Democracy Survey allows us to explore the extent of bias that may be more prevalent when studying politically sensitive topics. The average item nonresponse (INR) rate—meaning a rate at which respondents decline to respond to any particular question—is extremely low at just below 1 percent overall. However, INR increases substantially to 6 percent for the question related to January 6th insurrectionists. That increase in INR could be consequential. In total, the survey shows that 14 percent of respondents report some level of warm feelings towards the insurrectionists, and around 79 percent held cold feelings. If those silent voices were to skew towards a positive view of insurrectionists, the number of supporters could increase to over one fifth of respondents. This possibility should inspire scholars of public opinion interested in democratic feelings in America to think further about who is left out of the current analyses simply because they chose to express themselves through silence. By taking seriously and examining the politically silent, our analysis demonstrates why scholars need to widen the scope of who matters when measuring opinions about democratic norms.
We also believe this is a particularly important topic for the study of democratic norms. As a growing body of research continues to examine the state of democratic norms in the US (Clayton et al. Reference Clayton, Davis, Brendan Nyhan, Ryan and Wood2021; Hall and Druckman Reference Hall and Druckman2023), and as others search for the root causes of support for norm breaking actions on January 6th (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Rhodes and Daher Reference Rhodes and Nteta2024; Davis and Wilson Reference Davis and Wilson2023), we argue that it is important to consider the extent to which analysis of democratic norms may be missing a key population who purposely remains silent. Non-answers constitute an important element in how the U.S. public will determine the meaning of the fight on January 6th. As we show through analysis of INR in the Health of Democracy Survey, by ignoring the silent, research on the insurrection has missed key insights about the ways that the intersections of race, gender, and racial attitudes matter in providing a full picture of opinions. Therefore, our results highlight the necessity to always ask the question: do the silent voices of America teach us anything about what we think we know?
Race, Gender, and The Politics of Silence
What is left unsaid in politics is just as important as what’s said the loudest. As the work of Adam Berinsky highlights, “to understand public opinion in America, we must carefully consider the political interests and values of the politically silent” (Reference Berinsky2006, 2). Failure to do so leads to opinion polls that often “fail to equally represent the preferences of all Americans with regard to some of the most important issues of our time” (Berinsky Reference Berinsky2006, 2). This insight is particularly true for racial policy in America. As Berinsky’s (Reference Berinsky2006) analysis of data from the National Election Studies (NES) survey between 1990 and 1994, and from the General Social Survey (GSS) in 1986 shows, social pressures caused some Whites to withhold opinions about government intervention in school integration. This finding was the direct result of analyzing the characteristics of non-respondents. By recognizing that the group itself predicts lower support for government intervention in school integration, Berinsky makes a strong case for testing for exclusion bias.
The main contention of this paper is that nonresponses in public opinion data on the January 6th insurrection matter, and that the literature on democratic norms, racial attitudes, and gender politics will benefit from considering non-respondents in future analysis. As the public opinion data reveals, the events on January 6th were highly polarized (Gramlich Reference Gramlich2022), and connected with racial attitudes (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Rhodes and Daher Reference Rhodes and Nteta2024; Davis and Wilson Reference Davis and Wilson2023); both of which are relevant to social desirability (Berinsky Reference Berinsky2006; West and Iyengar Reference West and Iyengar2022; White and Laird Reference White and Laird2020). Just as Berinsky (Reference Berinsky2006) demonstrated a necessary intervention in the literature on racial prejudice by Whites (Bobo Reference Bobo2001; Krysan Reference Krysan2011), we show that the literature on the January 6th insurrection requires the same treatment. Despite the implementation of online surveys, which help limit the effect of social desirability bias (Wallace et al. Reference Wallace, Hedberg and Cesar2014), our results show that the effects of social desirability may not be completely eliminated. Respondents who take part in online public opinion surveys may feel social pressures despite the absence of face-to-face interaction when the question is about a sensitive topic such as the January 6th insurrection.
Additionally, as previous research has recognized the presence of a gendered response gap in polling and survey data (Shapiro and Mahajan Reference Shapiro and Mahajan1986), we believe this is an important understudied element in our understanding of public opinion surrounding the insurrection. Explanations about women’s relative silence has focused on their relative levels of political knowledge (Allum et al. Reference Allum, Sturgis, Tabourazi and Brunton-Smith2008; Burns, Schlozman, and Verba Reference Burns, Lehman Schlozman and Verba2001; Carpini and Keeter Reference Carpini and Keeter1996; Dolan and Hansen Reference Dolan and Hansen2020; Ferrín et al. Reference Ferrín, Fraile, García-Albacete and Gómez2020; Frazer and Macdonald Reference Frazer and Macdonald2003; Miller Reference Miller2019; Rae Atkeson and Rapoport Reference Rae Atkeson and Rapoport2003; Wolak and McDevitt Reference Wolak and McDevitt2011), as well as levels of political interest, efficacy, or the unique role of partisanship (Atkeson and Rapoport Reference Rae Atkeson and Rapoport2003). We contend that more attention must be placed on the social pressures involved in gendered INR, particularly with how pressures may involve racial attitudes and racial identity.
Little work has explored the possibility of a gendered response gap in questions pertaining to democratic norms or politically salient events. The work of Goenaga and Hansen (Reference Goenaga and Hansen2022) remains one of the first attempts to examine the gendered gap in response rates related to democratic institutions, finding only a small gender gap regarding questions about basic democratic practices, but a larger gap for questions requiring more specific knowledge about politics and therefore higher cognitive costs (Goenaga and Hansen Reference Goenaga and Hansen2022). Their results therefore follow the previous literature theorizing the importance of political knowledge gaps in predicting gender nonresponse gaps (Allum et al. Reference Allum, Sturgis, Tabourazi and Brunton-Smith2008; Atkeson and Rapoport Reference Rae Atkeson and Rapoport2003; Dolan and Hansen Reference Dolan and Hansen2020; Ferrín et al. Reference Ferrín, Fraile, García-Albacete and Gómez2020), and the literature which shows that women’s evaluations of democracy have historically been more critical than men (Anderson and Guillory Reference Anderson and Guillory1997; Hansen and Goenaga Reference Hansen and Goenaga2021).
We propose a new perspective on the question of the gender gap in political survey research on democratic norms. Like Goenaga and Hansen (Reference Goenaga and Hansen2022), we build upon the work of Berinsky (2004) and highlight the importance of understanding the role of political silence in understanding public opinion of democratic norms. However, we examine how gender may be one part of the whole story of political silence surrounding the January 6th insurrection. We also explore the role of race and racial attitudes as important elements in our understanding of who remained silent when asked about the January 6th insurrection.
Because previous research determining the causes of support for the insurrection has recognized that some support is racially motivated (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Rhodes and Nteta Reference Rhodes and Nteta2024; Davis and Wilson Reference Davis and Wilson2023), we believe the racial attitudes of the politically silent merit greater attention. As Barreto et al. (Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023) find, negative attitudes toward immigrants and the belief in White replacement theory—the idea that there is an active, ongoing and covert effort to replace white populations—were key predictors of support for the January 6th insurrection among White respondents. Employing a survey experiment to distinguish between old and new racial grievances, anti-immigrant beliefs, Black activism, and support for the January 6th insurrection, Barreto et al. (Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023) convincingly argue that support for the insurrection was partially motivated by anti-Black Lives Matter rhetoric spread by Donald Trump and right-wing news sources. Their findings, however, do not account for possible exclusion bias among those who shared no opinions of the insurrection. The 1,340 completed responses from self-identified non-Hispanic whites in their study is reduced to 1,214 in their analysis. That is a 9-percentage point drop in responses with no breakdown of nonresponses.
The work of Davis and Wilson (Reference Davis and Wilson2023) and Rhodes and Nteta (Reference Rhodes and Nteta2024) also highlights the importance of racial attitudes in determining support for the insurrection with little attention to nonresponse. According to Davis and Wilson (Reference Davis and Wilson2023), racial resentment is the dominant explanation for American’s differing opinions on whether or not the January 6th insurrection was justifiable, or whether it required investigation through the creation of the U.S. House Select Committee. Their findings were further strengthened by the fact that many of the districts targeted by “Stop the Steal” allegations were centered on districts with large African American and Latino populations, and that many of the insurrectionists were self-declared white nationalists. Rhodes and Nteta (Reference Rhodes and Nteta2024) argue that racial affect is a key determinant of American attitudes toward the accountability of the insurrection. They find that Trump’s use of racial rhetoric established a racialized identity which connected negative racial attitudes with support for his presidency among members of the mass public and in doing so created the conditions in which those negative attitudes were likely to “spill over” into the not yet racialized attitudes toward the January 6th insurrectionists.
We contend that the limitations within the literature on racial attitudes and the insurrection are not their conclusions, but only where their analysis stops short. First, across all the aforementioned research, little to no attention was given to those survey respondents who refused to answer questions about the insurrection. Within the YouGov survey used by Rhodes and Nteta (Reference Rhodes and Nteta2024), out of the 1,051 respondents who were then matched down to a sample of 1,000, their models only included a maximum of 695 observations, and provided no analysis for the reduced sample in their analysis. Similarly, using the 2022 political unrest study and the 2020 Collaborative Multiracial Post-election Survey (CMPS), Barreto et al. (Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023) only include respondents with full information. We recognize that listwise deletion is a common practice in survey research, but this may introduce some bias in the findings. On the other end of the spectrum, Davis and Wilson (Reference Davis and Wilson2023) compelling results based on nationally representative 2021 Cooperative Election Study (CES) forced respondents to respond to the question on the insurrection, thereby possibly leading some respondents to provide a neutral or socially desirable answers. While it is possible that forced answers causes those possible nonrespondents to share important democratic tendencies, we believe this itself devalues the importance of messages conveyed through silence.
It is difficult to draw generalizations based on the results by Barreto et al. (Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023) and Davis and Wilson (Reference Davis and Wilson2023) because both studies limit their analyses to White respondents. While Rhodes and Nteta (Reference Rhodes and Nteta2024) do include Non-Whites in their analysis of racial attitudes and the insurrection, it is again not clear whether forced responses on sensitive items yields fully honest responses. The insufficient attention to patterns of nonresponse across the subsamples neglects the similarities and differences driving Whites and Non-White opinion about the insurrection. Even if racial resentment and White identity politics lead to higher levels of support for the insurrection, closer attention should be paid to whether racial attitudes among Whites are driving levels of support the insurrection in comparison with their Non-White counterparts. As Kam and Burge (Reference Kam and Burge2019) suggest, White and Black survey respondents often share highly similar open-ended interpretations of the items in the racial resentment scale.
We find that while some factors may similarly affect White and Non-White nonresponses, these same factors do not predict views about the insurrection. This matters because it suggests that Whites and Non-Whites are affected by unique factors when deciding whether to share their opinions about an important democratic norm. As we discuss below, this does not mean that race, racial attitudes, gender, or other factors themselves are not important in determining support for the insurrection, but only that we as scholars must be attuned to the ways these factors operate differently across race and gender.
Support for Democratic Norms in a Polarized Era
American political scientists have warned that democracy may be under threat (Finkel et al. Reference Finkel, Bail, Cikara, Ditto, Iyengar, Klar and Mason2020). For instance, Levitsky and Ziblatt (Reference Levitsky and Ziblatt2018) state that by breaking the democratic norm of allowing the President to nominate Supreme Court justices, the Republican-led Senate’s refusal to grant a hearing to President Barrack Obama’s Supreme Court nominee Merrick Garland to replace the late Antonin Scalia set a new undemocratic precedent for political elites in Congress. On the other hand, the work of Holliday et al. (Reference Holliday, Iyengar, Lelkes and Westwood2024) demonstrates that while political elites increasingly express anti-democratic positions, the American public across the political spectrum overwhelmingly opposes norm violations and partisan violence. This bodes well for democracy, but we must discern whether America’s commitment to democratic norms may be skewed if certain voices are over-represented.
While we do not stake out a position in the debate over the exact causes of anti-democratic beliefs or the willingness to break norms among the America public, we seek to challenge future work to better recognize the importance of silent voices in this discussion. Additionally, by providing evidence of a persistent pattern of silence around the January 6th insurrection, which broke the longstanding norm of the peaceful transition of power, we show that what we think we know about the levels of anti-democratic behavior in America is incomplete, and that race and gender have important effects on these silent voices. We therefore contend that the initial conclusions about Americans’ beliefs about the insurrection—particularly those related to racial attitudes and gender (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Gramlich Reference Gramlich2022; Rhodes and Nteta Reference Rhodes and Nteta2024; Davis and Wilson Reference Davis and Wilson2023—require a closer examination of those who actively decided to remain silent. By including those silent voices present in the discussion around the insurrection, we can ensure there is no overestimation of democratic or undemocratic attitudes on the January 6th insurrection, and no limitations in the diversity of whose voice matters.
As Pew Research Data collected days after the insurrection found, there was a general consensus among Americans at this time that the insurrection was shocking and bad for the country (Gramlich Reference Gramlich2022). However, the results also showcase that some Americans do hold sympathies for insurrectionists, and some believe the insurrection itself was an attack manufactured by the political left to harm President Trump. When analyzed by party identification, Republicans disagreed with who was to blame, and if protestors deserved to be punished. In fact, nearly 17 percent of Republicans who volunteered a reaction stated that Trump and his supporters were not to blame, but instead groups such as Antifa or Black Lives Matter (Gramlich Reference Gramlich2022). Beyond the positive or negative opinions, the survey shows that 10 percent of respondents made the active decision not to share any opinion. Low response rates do not inherently mean that there is bias (Jennings and Wlezien Reference Jennings and Wlezien2018; Keeter Reference Keeter2018) because a survey’s utility should be judged by the representativeness of sample compared to the intended population of interest (Prosser and Mellon Reference Prosser and Mellon2018). However, ignoring nonresponse patterns may limit our understanding of American public opinion (Berinsky Reference Berinsky2006). We address this limitation by analyzing the data on political silence in relation to the January 6th insurrection.
Hypotheses
Based on the previous literature, we test four hypotheses in our analysis of January 6th INR. First, (H1) we hypothesize that questions about the January 6th insurrection and other racialized issues related to the insurrection will garner more nonresponses than non-racialized questions among White and Non-White respondents. Second (H2), we hypothesize that Whites and Non-Whites will withhold opinions about the January 6th insurrection at the same rates. As Elisabeth Noelle-Neumann’s (1974) theory of the spiral of silence suggests, the overall status of public opinion has an effect on the individual’s opinion. Because of the politicized (Gramlich Reference Gramlich2022), racialized (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Rhodes and Nteta Reference Rhodes and Nteta2024; Davis and Wilson Reference Davis and Wilson2023), and predominantly negative nature of feelings towards the insurrection (Springs and Blank Reference Springs and Blank2024), we expect the social stigma surrounding opinions of the insurrection to be felt among Whites and Non-Whites alike. In contrast, for our third hypothesis, (H3) we posit that because the insurrection was a major violation of norms, support for democratic norms will have a negative and significant relationship with high January 6 INR. Due to the stigma around anti-democratic beliefs, those individuals who hold strong beliefs in democratic norms should not be hindered from sharing their negative beliefs about the insurrection.
And lastly, we believe that racialized feelings will play a key role in determining when an individual chooses to remain silent. While we are unable to assess the direct consequences of nonresponse bias in extant research about the January 6th insurrection, we argue that nonresponse may affect individuals who hold previously identified racial attitudes linked to supporting the insurrection (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Rhodes and Nteta Reference Rhodes and Nteta2024; Davis and Wilson Reference Davis and Wilson2023). However, we believe this finding will extend to Non-White racial attitudes. Therefore, (H4) we hypothesize that high levels of racial resentment, Non-White resentment, and belief in the white replacement theory will be significantly related to those with high January 6 INR. We expect these results because of the high salience of the insurrection at the time of the Health of Democracy Survey, which will have raised the stakes of an issue proven to be racialized (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Rhodes and Nteta Reference Rhodes and Nteta2024; Davis and Wilson Reference Davis and Wilson2023).
We believe that scholars have been too quick to discount the possibility that Non-Whites racial attitudes have any impact on how they share their feelings toward the insurrection, especially given the current literature showing racial attitudes to be fundamental in predicting support for the insurrection among Whites (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Rhodes and Nteta Reference Rhodes and Nteta2024; Davis and Wilson Reference Davis and Wilson2023). If the literature tying racial attitudes to support for the insurrection suggest that for some Whites sharing more sympathetic views becomes less socially desirable, we should then expect a similar effect among Non-Whites. As White and Laird Reference White and Laird2020 show, social desirability is a major force in Black politics. And as research from Kam and Burge (Reference Kam and Burge2019) shows, some Black Americans also express racial views similar to those of White Americans who score high on measures of racial resentment. We cannot ignore this possibility among Latinos. The possibility of Latino support for the insurrection was clearly shown by the fact that one of the most high-profile insurrectionists was Enrique Tarrio, leader of the self-proclaimed “Western chauvinist” group the Proud Boys (Office of Public Affairs).
Data and Measurements
We leverage the nationally representative 2022 Health of Democracy Survey to test the impact and significance of nonresponses in public opinion data on the January 6th insurrection. This survey was conducted from October 20 to 26 in 2022 via the NORC AmeriSpeak Panel and includes a total sample of 1,557 adults aged 18 years and older. The racial and ethnic composition of the sample is 1,025 White respondents, 193 Black respondents, 233 Latino respondents, 47 Asian respondents, and 59 other or mixed-race non-Latino respondents. This allows us to distinguish between the levels of item nonresponse between Whites and Non-Whites. To supplement our analysis, we also leverage the 2023 Health of Democracy Survey to test the longevity of our findings. We chose to limit our main analysis to the 2022 survey in order to capture opinions nearest to the insurrection and the subsequent January 6th Committee, which completed its final report in December of 2022. By doing this, we are ensuring our data captures the opinions during a moment where the insurrection is of high salience. However, by using the 2023 survey, we can show that political silence is not a momentary phenomenon—but instead a persistent trend that deserves more attention.
It is important to note the quality of the 2022 Health of Democracy Survey allows us to analyze the importance of nonresponse rates. The survey has a margin of error that is plus or minus 2.5 percentage points at the 95 percent confidence level, and a mean item-nonresponse rate—or the mean number of questions skipped—of 13.7, less than 1 percent overall within the 95 questions asked of all 1,557 respondents. This high response rate provides us with valuable near perfect data on participants’ demographics, partisanship, ideology, and a wide range of questions related to opinions on democratic norms, undemocratic tendencies, and feelings about the January 6th insurrectionists. The survey therefore allows us to determine whether there are any statistically significant patterns of who most likely to withhold their opinions on January 6th insurrectionists and the importance of nonresponses in predicting racialized and undemocratic attitudes.
The key dependent variable within this study is item nonresponse for feelings toward January 6th insurrectionists—which we will refer to as January 6th INR. In our analysis, we code all refusals (skips) as January 6th INR. We measure this directly by using the Health of Democracy’s feeling thermometer question which asked respondents to range their feelings for insurrectionists from the “Extremely cold” level of 0 to the “Extremely warm” level of 100, or to refuse to answer and skip the question. Overall, 15 percent of Whites and 14 percent of Non-Whites held positive feelings toward January 6th insurrectionists. Demographic characteristics represent the independent variables tested in our first model, including gender, age, education, marital status, homeownership, employment, party identification (party ID), veteran status, urban location, and religious affiliation.
While additional surveys have offered questions relating to the insurrection, the value of the Health of Democracy Survey is that it directly asks respondents about their feelings toward insurrectionists and allows for nonresponses. In other surveys, such as the 2020 Collaborative Multiracial Post-election Survey, the question pertaining to January 6th asks respondents to determine if it was “a coordinated act of insurrection against the United States,” or if it was “mostly a protest that went too far.” While those questions are valuable, we are interested in exploring the feelings about the insurrectionists. Therefore, the Health of Democracy Survey provides us the best opportunity to measure the dependent variable we set out to analyze.
We base our additional independent variables on the current literature explaining support for the insurrection. First, we include a number of variables measuring racial attitudes which were shown to be important in predicting support for the insurrection (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Davis and Wilson Reference Davis and Wilson2023). Following the work of Davis and Wilson (Reference Davis and Wilson2023), we use the three-question measurement of racial resentment within the Health of Democracy, itself modeled on the work Davis and Wilson (Reference Davis and Wilson2021), to measure racial resentment. Then, in order to assess the significance of views about white replacement theory in the work of Barreto et al. (Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023), we use the Health of Democracy Survey’s measurement of the level of threat respondents feel toward the statement that shifting demographics “will make Whites a minority.” However, in addition to these variables, we include an additional variable which emphasizes our belief that racial attitudes must be explored among Whites and Non-Whites: a variable for Non-White resentment. This is measured by using the Health of Democracy Survey’s measurement of the resentment Non-Whites feel toward Whites’ denial of racial discrimination. This will allow for us to determine is racial attitudes held different effects on January 6th INR among Whites and Non-Whites.
We also include support for democratic norms, reliance on Fox News, and support for Christian Nationalism as additional independent variables. Following the January 6th insurrection, a public debate has ensued over the level of support for democratic norms among Americans (Holliday et al. Reference Holliday, Iyengar, Lelkes and Westwood2024). Our measurement of support for democratic norms is a composite of 6 questions asking respondents their level of agreement with democratic norms and is shared in full in Appendix A. This independent variable allows us to examine the relationship between January 6th INR and opinions about democratic norms. Our analysis therefore provides the first examination of the importance of democratic norms among those who chose to remain silent when asked about their feelings toward January 6th insurrectionists. We include a variable that measures reliance on Fox News due to the uniquely cautious and in some instances sympathetic coverage of the insurrection by some Fox News hosts (Dreisbach Reference Dreisbach2021; Mascaro, Amiri, and Jalonick Reference Mascaro, Jalonick and Amiri2023). Lastly, we include the Health of Democracy Survey’s measurement of support for Christian Nationalism due to the movement’s presence on the ground during the attack (Seidel Reference Seidel2022). These variables allow us to examine out hypothesis while ruling our additional explanations for January 6th INR.
We test our hypothesis using logistic regression. This allows us to uncover the significance of key demographic variables in determining January 6th INR among a representative sample of Americans. With this demographic data, we can also test to examine how racial resentment, racial affect, and support for democratic norms on January 6th INR among different subgroups of the population. To do this, we first split up the data between White and Non-White respondents. And second, we split up the data using significant demographic variables that predict January 6 INR.
Results
The results of the 2022 Health of Democracy Survey corroborate previous research showing that support for the January 6th insurrection is low (Gramlich Reference Gramlich2022; Orth Reference Orth2023). Only 14 percent of respondents reported having some warm feelings towards the insurrectionists, while 79 percent of those surveyed expressed some level of cold feelings. But what can this survey show us about those who chose to remain silent? Following our first hypothesis, the data show that the question probing respondents’ feelings toward the January 6th insurrectionists garnered the highest level of INR on the survey—visualized by the highest peak for the green line in Figure 1. While the overall average INR rate was low with a mean of less than 20 nonresponses per question, the question measuring respondents’ feelings toward January 6th insurrectionists garnered 93 nonresponses, representing a 5.89 percent January 6th INR rate for all respondents, and a 75 person increase from all other questions. Again, although this remains low, its shows that nonresponse was a uniquely important problem for the January 6 question.
Item Non-Response by Race and Question 2022.
Following our second hypothesis, we find that January 6th INR patterns do not differ among Whites and Non-Whites. Both groups were more likely to withhold their feelings toward insurrectionist than any other question, signaling a generalizable pattern within the data. Given that there are half as many Non-Whites in the sample as Whites, this result should give scholars of public opinion and democratic norms pause. While previous research has shown the importance of racial attitudes in predicting support for the insurrection, little research has compared the similarities or differences between Whites and Non-Whites feelings toward insurrectionists separately. As later results will show, examining this distinction also matters for exploring the effects of racial attitudes on January 6th INR. In addition to the question on feelings toward insurrectionists, Figure 1 shows that the INR rate was, to a lesser degree, higher for questions related to racial (in)justice and white identity politics. Respondents were less likely to answer all questions probing feelings toward Black Lives Matter protestors, their beliefs in the white replacement theory, and belief in the danger of critical race theory in comparison with non-race related questions. These summary results indicate a descriptive pattern within INR rates that aligns with past findings that racial issues impact INR rates (Berinsky Reference Berinsky2006).
Figure 2 corroborates the pattern we found in 2022, where January 6th INR is higher than all other questions, within the 2023 Health of Democracy Survey. This is represented by the highest peak of the gray line. Additionally, in 2023, we find that Non-Whites were more likely than Whites to remain silent when asked about their feelings toward insurrectionists in 2023. This result again provides evidence of the importance of examining social desirability biases among Whites and Non-Whites. Our results also show that the general trends for INR were similar to 2022. Again, there were higher levels of INR evidenced in racialized questions on issues like diversity equity and inclusion. One additional trend within the data we are able to discover because of the differing questions from the 2023 survey is high INR for questions asking for gender inequality estimates—a trend which is important to note as we highlight and explore the importance of gender in INR.
Item Non-Response by Race and Question 2023.
While the aforementioned results alone do not provide a causal link between silent voices and racialized undemocratic opinions or other trends of nonresponse, high January 6th INR represents an important element of the story which scholars have not given adequate attention. In particular, the presence of high levels of January 6th INR among Whites and Non-Whites shows the importance of examining these groups separately. Precisely because previous research has shown support for the January 6th insurrection to be racialized (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Rhodes and Nteta Reference Rhodes and Nteta2024; Davis and Wilson Reference Davis and Wilson2023), we should not expect the social pressures impacting INR related to the insurrection to operate the same among Whites and Non-Whites. We therefore test to find which, if any, demographic variables predict January 6th INR among all respondents, Whites, and Non-Whites separately. In Table 1, the coefficients then represent the relationship between each demographic variable and January 6th INR. A positive coefficient would therefore indicate that individuals with that demographic characteristic are more likely to withhold their opinion on the question regarding their feelings toward the insurrectionists.
Predicting January 6 item nonresponse by demographic characteristics
* p < .10, **p < .05, ***p < .01; Change in predicted probabilities from minimum to maximum are calculated holding other values at their mean and shown only for statistically significant values.
In Table 1, we find that gender had a positive and significant relationship with January 6th INR among all respondents. Regardless of race, female respondents were significantly more likely than men to withhold feelings toward January 6th insurrectionists. The magnitude of the effect of gender was, however, shown to be stronger among White women. For Non-White women, identifying as a female had a positive and significant coefficient of .068 (p<.04), but for White women the coefficient was .035 (p=.01). When calculating the change in predicted probability, holding other valued at the observed mean, compared to their male counterparts, White women were 3 % more likely to withhold an opinion about January 6th insurrectionists. Among Non-White respondents, the change in predicted probability was only 1%. While both were highly significant, below the p=.05 level, the effect on their probability of withholding and opinion suggests that social and racial dynamics may operate differently among Whites and Non-Whites. Additionally, the results in our base model show that gender was the only significant variable among Non-Whites, while a number of demographic characteristics had a strong impact on January 6th INR among Whites. This again points to the necessity of examining race and ethnicity when studying political silence.
Among Whites we found that the relationship between an increase in age and January 6th INR was negative, represented by the coefficient −.016 (p=.02) and a 4 % change in predicted probability for respondents in the oldest age category, compared to those in the youngest age category. This means that older White respondents were significantly more likely to share their feelings about the insurrectionists. Being married also had a significant positive relationship with January 6th INR with a coefficient of .049 (p<.01)—signaling that the topic of the insurrection may be controversial within the dynamics of married White couples. Higher income was also negatively associated with January 6th INR with a coefficient of −.015 (p=.04), meaning that those with higher incomes were less likely to remain silent. And lastly, identifying as a Democrat had a significant positive relationship with January 6th INR among Whites, represented by a .007 coefficient (p=.03). Given the insurrectionists’ attempts to overturn the Democrat Joe Biden’s electoral victory, the perceived social cost among Democrats for sharing their opinions about insurrectionists may be higher.
We then build upon our demographic model by including our variables measuring racialized attitudes, support for democratic norms, support for Christian Nationalism, and Fox News consumption in Table 2, again examining each variable’s relationship with January 6th INR among Whites and Non-Whites. Because questions measuring racial resentment were only asked of White respondents, our racial resentment variable was only included within our White model. And because questions regarding resentment toward Whites were only asked of Non-Whites, this variable was only included in the Non-White model. The variables representing belief in the white replacement theory, support for Christian Nationalism, and reliance on Fox News were included in all models because they were asked of all respondents. We chose to include the measure of support for Christian Nationalism due to the many connections that insurrectionists had with Christian nationalist movements (Seidel Reference Seidel2022), and the 6 point measure of reliance on Fox News due to the networks’ more cautious and sympathetic reporting of the insurrection (Mascaro, Amiri, and Jalonick Reference Mascaro, Jalonick and Amiri2023). Within Table 2, coefficients therefore represent the relationship between racial attitudes, beliefs in democratic norms, and demographics with January 6th INR among White and Non-White respondents.
Predicting January 6 item nonresponse by demographic characteristics and norms and racial attitudes
* p < .10, **p < .05, ***p < .01; change in predicted probabilities from minimum to maximum are calculated holding other values at their mean and shown only for statistically significant values.
Among Whites in Table 2, Fox News consumption was the only additional variables found to have a significant relationship with January 6th INR, with a negative coefficient of −.011 (p=.02). This means that White respondents who rely more on Fox News to follow politics were more likely than others have shared their feelings toward insurrectionists. This result is not surprising given the way Fox News reported on the insurrection. On the night of the attack, Fox News hosts spread rumors that antifa may be partially responsible for some of the violence, and tried to distance President Trump’s supporters from the violent actions of insurrectionists (Dreisbach and Folkenflik 2021), and after the insurrection Fox News host Tucker Carlson portrayed the insurrection as a peaceful gathering which has been misconstrued by the January 6 Committee tasked with its investigation (Mascaro, Amiri, and Jalonick Reference Mascaro, Jalonick and Amiri2023). There is no empirical support for our third hypothesis, as we find there is no relationship between support for democratic norms and January 6th INR among any group.
In Table 2, we find significant differences in the way racial attitudes impact January 6th INR among Non-Whites and Whites. Our results show that racial attitudes are important predictors for January 6th INR only among Non-Whites. As shown in Table 2, there is no statistically significant relationship between racial resentment or belief in the white replacement theory with January 6 INR. However, Christian Nationalism and White resentment among Non-Whites are significantly associated with January 6th INR among Non-Whites. The negative coefficient −.087 (p=.08) suggests that Non-White respondents who support Christian Nationalism were more likely to share their feelings about insurrectionists, while the .005 (p=.02) coefficient suggests that Non-Whites who hold higher levels of resentment toward Whites, capturing the extent to which they perceive White denial of racial discrimination, were more likely to withhold their feelings about insurrectionists. These findings provided mixed support for our fourth hypothesis. Although racial resentment and belief in the White replacement theory were insignificant, Non-White resentment was significantly associated with higher January 6th INR. These results highlight precisely why scholars interested in racial attitudes should always consider including Whites and Non-Whites in their analysis. Racial attitudes exist among all racial and ethnic groups and were a significant factor in political silence surrounding feelings toward the insurrectionists among Non-Whites.
Additionally, we find that in Table 2 the relationship between gender and January 6th INR lost its significance among Non-Whites. This suggests that racial attitudes held stronger effects on January 6th INR among Non-Whites than gender, even though the base model indicated that gender was statistically significant. This finding highlights the significance of the intersection of race and gender in conversations about democratic norms, racial attitudes, and withholding views about highly politicized events. Racialized attitudes and gender impact our conclusions about what we think we know and should be given proper consideration in the study of public opinion among all racial and ethnic groups. This is true for research on the opinions individuals choose to share and the messages they communicate through silence.
Conclusion
After the January 6th attack on the Capitol, political scientists have worked to examine what the insurrection means for American democracy. Using the 2022 Health of Democracy Survey we demonstrate that the literature examining American’s feeling about the insurrection should reconsider the importance of selective nonresponses and what this silence may communicate. We show that the silent voices of White and Nonwhite women are being under-examined in public opinion research on the insurrection and that the intersection of race and gender is important to consider in conversations around effects of racial attitudes on political silence. These results additionally suggest that that the literature showing women less likely to share opinions about democratic institutions and norms due to higher information costs should not discount the effects of social pressures and require further exploration into the ways women do engage with political information. As we show, women who rely heavily on Fox News were more likely to share their opinions. But most importantly, more work needs to explore how the interception of race and gender impact differences in nonresponses among Whites and Non-Whites.
The complex relationship between race, gender, and intentional silence when asked about the January 6th insurrection highlights a number of methodological issues for future research. As it has already been shown, women tend to be more critical of democratic institutions (Anderson and Guillory Reference Anderson and Guillory1997; Goenaga and Hansen Reference Goenaga and Hansen2022; Karp and Banducci Reference Karp and Banducci2008). By disproportionately losing female voices in public opinion research on democratic norms, scholars may be losing critiques of democracy from a group which has historically faced challenges to their political inclusion (Wolbrecht Reference Wolbrecht2000). Similarly, by opting to only include Whites within research on public opinion and racial attitudes, scholars may be missing important response patterns across racial and ethnic groups. While racial attitudes may have fueled White’s support for the insurrection, racial attitudes were fuel for Non-White’s political silence. If democratic norms are to remain a topic of inquiry, more must be done to ameliorate the racial and gender gap in political silence. If little is done, researchers will face possible biases—where the silent voices of women and Non-White respondents may be limiting our understanding of the public’s beliefs in democratic institutions and norms.
A first step to ameliorate possible bias in the analyses is to adequately sample Non-whites and to recognize that political attitudes and non-attitudes among women and Non-White should not simply be assumed to happen at random and are therefore consequential. In particular, scholars need to take gendered experiences and socialization into account within the study of Americans’ feelings towards democracy and always include analysis or distinct racial groups based on what’s available in the data. Gender blindness refers to the failure, or possible unawareness, to give attention to the differences in socialization, roles, needs, opportunities, and interactions between women and men (Bacchi Reference Bacchi, Essed, Goldberg and Kobayashi2009; Forman-Rabinovici and Mandel Reference Forman-Rabinovici and Mandel2023). As Forman-Rabinovici and Mandel (Reference Forman-Rabinovici and Mandel2023) show in an exploration of all quantitative articles between 2018 and 2019, researchers who account for gender are more likely to obtain more accurate result and a better understanding of political phenomena. So, in the study of democratic norms, we believe more work needs to be done to theoretically and empirically contend with how gender is affecting political beliefs and behavior.
In addition to our findings surrounding the importance of gender, we also find that Americans with high levels of racial resentment and those who believe strongly in the white replacement theory are not afraid to voice their opinions on the insurrection—possibly hinting at the fact that these groups do not feel any social stigma around the insurrection. Because Americans with those racial attitudes are more likely to support the insurrection (Barreto et al. Reference Barreto, Alegre, Isaiah Bailey, Davis, Ferrer, Nguy, Palmisano and Robertson2023; Rhodes and Nteta Reference Rhodes and Nteta2024; Davis and Wilson Reference Davis and Wilson2023), this result raises questions about how the election of 2024 will embolden further breaches of norms. With President Donald Trump reentering the White House for a second term in 2025, it has become more imperative now to continue to examine how public opinion surrounding the January 6th insurrection may change. It is possible the prestige and power of the U.S. Presidency will embolden a new narrative. Additionally, more work needs to explore the possibility that opinions may change due to Americans experiencing political fatigue, simply overwhelmed by the frequent encroachment of socially sensitive topics into everyday life.
While the empirical importance of democratic norms is not the direct subject of our analysis, the inclusion of all Americans in our understanding of support for democratic norms is. Because we found that women and Non-Whites who hold certain racial attitudes represented a significant portion of the silent voices on the January 6th insurrection, we have found an important missing element within our understanding of the health of American democracy. The analysis provided is therefore a catalyst for future public opinion research on the strengths and weaknesses of democratic norms in America. As Berinsky (Reference Berinsky2006) already made clear, the interests and values of the politically silent in America are just as important as those we can hear. Without adequate representation for all voices, our understanding of public opinion in America is left lacking. Going forward, it is crucial for scholars to widen their attention to include all voices, even the silent ones.
Acknowledgments
An earlier version of this article was presented at the 2024 Keeping the Republic Conference at the University of Notre Dame and the CMPS January 6 Paper Symposium at the University of California, Los Angeles. We would like to thank all those who provided comments on those earlier versions.
Funding statement
This research did not receive any specific financial support.
Competing interests
The authors declare none.
Appendix A: Questions Representing Support for Democratic Norms
Each questions was asked for respondents to rank 1 (Strongly Disagree) – 7 (Strongly Agree).
-
1. [Everyone should be allowed to vote] Indicate the extent you agree or disagree with the following statements about government:
-
2. [Everyone should be allowed to express any idea, even potentially dangerous ideas] Indicate the extent you agree or disagree with the following statements about government:
-
3. [The government should never shut down media outlets, even if they spread disinformation] Indicate the extent you agree or disagree with the following statements about government:
-
4. [The president should not be above the law] Indicate the extent you agree or disagree with the following statements about government:
-
5. [The law should treat everyone the same, regardless of wealth or power] Indicate the extent you agree or disagree with the following statements about government:
-
6. [In order for a leader’s actions to be legitimate, they need to follow the rules] Indicate the extent you agree or disagree with the following statements about government:
Open access
