In the same year that Converse (Reference Converse and Apter1964) convinced American political behavior scholars that few voters conceived of politics through an intricate and coherent ideological prism, McClosky (Reference McClosky1964) demonstrated that the mass public’s lack of constraint was applicable to opinions about the tenets of liberal democracy. Specifically, McClosky’s 1950s data showed that despite political elites expressing decidedly more system support than the general public, both segments of the polity typically fell far short of registering “consensus” on democratic principles. Indeed, echoing the conclusion of Converse, McClosky offered a similar and hardly uplifting pronouncement: as long as political elites maintained a considerably higher degree of system support in the midst of a generally disinterested and ideologically naïve electorate, it was perhaps enough to keep the republic.Footnote 1
Similar to Gibson (Reference Gibson2008), who compared levels of political intolerance in the American public drawing on survey data gathered a half-century apart, we revisited McClosky’s classic work to assess the extent to which the mass public exhibits continuity and change in its responses to most of McClosky’s original questions. We also took another step by conducting an exploratory factor analysis (EFA) on McClosky’s battery of survey items. We discovered that these questions clearly load onto three political dimensions, and two of the factors—elitism and racial supremacy—influenced preferences and behavior in the 2016 presidential election. Our findings speak to the contemporary political relevance of McClosky’s path-breaking research. In an age of sharper, although not always deeper, partisan divisions, McClosky’s line of inquiry continues to inform our understanding of public opinion about American democracy, particularly in light of recent mass demonstrations over police killings, the Black Lives Matter movement, and the broader concerns about both endemic racism and the rise of authoritarianism in the United States today.
IDEOLOGICAL INNOCENCE AND SUPPORT FOR AMERICAN DEMOCRACY
Despite the incontrovertible evidence that the mass public has become more cognizant (Hetherington Reference Hetherington2001) of the increasing ideological polarization occurring among officeholders (Fleisher and Bond Reference Fleisher and Bond2004; Theriault Reference Theriault2008), it remains true that only the more politically engaged (Abramowitz and Saunders Reference Abramowitz and Saunders2008) and ideologically involved (Jewitt and Goren Reference Jewitt and Goren2016) albeit growing (Abramowitz Reference Abramowitz2011) minority of the electorate has come to more closely resemble the increasingly polarized mindset of modern-day political elites. Specifically, this subset of the mass public possessing the type of ideological constraint that Converse (Reference Converse and Apter1964) spoke of, which would place it in the highest level of political conceptualization (i.e., “ideologues”), is still no more than around 5% (Lupton, Myers, and Thornton Reference Lupton, Myers and Thornton2015).
Most of the evidence strongly suggests that the contemporary partisan sort of American voters into the proper ideological camp stems most heavily from party cues as opposed to a detailed comprehension of ideological principles (Levendusky Reference Levendusky2009).Footnote 2 The frequent use of the term “tribalism” to distinguish opposing partisans speaks to a social-identity–based attachment to one’s political party (Green, Palmquist, and Schickler Reference Green, Palmquist and Schickler2002) rather than an ideologically driven, policy-based division between Democrats and Republicans. The rise of affective polarization (Webster and Abramowitz Reference Webster and Abramowitz2017) and negative partisanship (Abramowitz and Webster Reference Abramowitz and Webster2016; Reference Abramowitz and Webster2018) rests primarily on a loathing of the opposing partisan outgroup and not because of an intimate knowledge of ideologically grounded differences (Mason Reference Mason2018).Footnote 3
Because most of recent scholarship identifies a mass public that is overwhelmingly ideologically innocent (e.g., Kinder and Kalmoe Reference Kinder and Kalmoe2017) despite an increasingly polarized environment, we have no expectation that contemporary American views of liberal democracy have undergone a sea change since the publication of McClosky’s seminal work. Furthermore, because the more pronounced partisan polarization of the electorate is not a result of growing ideological sophistication, we suspect that some of the movement in modern opinions about liberal democracy is tied to negative partisanship and/or affective polarization.
DATA AND METHODS
To explore contemporary Americans’ attitudes toward liberal democracy, we produced a module included in the 2016 Cooperative Congressional Election Study (CCES). The module contains most of the items originally created by McClosky (Reference McClosky1964) for a national survey of 1,484 adults conducted by Gallup in January 1958. We compared only our national survey to his national survey, leaving to future researchers to survey a comparable sample of contemporary elites.
We replicated McClosky’s items with three objectives in mind. First, how have ordinary Americans’ attitudes toward liberal democracy changed over time? Second, what exactly do McClosky’s questions measure; that is, do his questions reveal any unique dimensions to this somewhat vague concept? In pursuit of an empirically driven answer, we subjected McClosky’s questions to an EFA. Third, do Americans’ attitudes toward liberal democracy shape other political preferences and behaviors? Using the results of our EFA, we discovered how attitudes toward liberal democracy shaped preferences and votes for presidential candidates in 2016. Our results indicate that attitudes toward liberal democracy, first assessed in the late 1950s, continue to have important implications for Americans’ political behavior.Footnote 4
THE 1950S VERSUS TODAY: HAVE AMERICAN ATTITUDES TOWARD LIBERAL DEMOCRACY CHANGED?
Figures 1–4 report the mass public’s responses to McClosky’s (Reference McClosky1964) survey items alongside those of the 2016 CCES respondents.Footnote 5 We are interested in how the responses of the mass electorate in 1958 compare to those in 2016. These charts report proportions of citizens who agree, alongside confidence intervals. McClosky’s raw data, unfortunately, are unavailable. Therefore, we generated our own confidence intervals around the proportions he reported. McClosky’s article included all of the necessary information because he reported the proportion of respondents who agreed with each question and the corresponding sample size.Footnote 6
Attitudes toward Political Equality, 1950s and 2016
Attitudes toward Racial and Ethnic Equality, 1950s and 2016
Attitudes toward Rules of the Game, 1950s and 2016
Attitudes toward Free Speech and Procedural Rights, 1950s and 2016
These figures establish both continuity and change in mass attitudes toward democracy. Continuity is arguably most apparent among questions regarding political equality, as shown in figure 1. The figure reveals that although Americans were marginally more conservative about political equality in the 1950s, Americans today are not significantly different. In fact, figure 1 reveals only one statistically significant difference: in the 1950s, Americans were more likely to agree that “‘Issues’ and ‘arguments’ are beyond the understanding of most voters.” On every other question, Americans in 2016 were statistically indistinguishable from Americans in the 1950s.
Change, however, is evident with respect to racial and ethnic equality, as shown in figure 2. In the 1950s, majorities agreed with several of these items, revealing more conservative attitudes concerning race and ethnicity. In 2016, far less than a majority expressed racially conservative attitudes. Still, there were significant numbers who possessed racially conservative attitudes: more than one of five Americans in 2016 believed that all races are not equal; roughly one of three agreed that “…there are certain races in the world that just won’t mix with Americans”; and more than one of three agreed that “we have to teach children that all people are created equal but almost everyone knows that some are better than others.”
Figure 3 presents a more complicated picture. Four of nine questions concerning the rules of the game revealed statistically indistinguishable proportions between the 1950s and 2016. Americans in 2016, for example, were no more likely to agree with the statement, “I don’t mind a politician’s methods if he/she manages to get the right things done.” Americans were also just as likely to believe that “Almost any unfairness or brutality may have to be justified when some great purpose is being carried out.”
However, Americans’ attitudes have changed on other questions regarding rules of the game. In 2016, Americans were significantly more likely to agree that “There are times when it almost seems better for the people to take the law into their own hands rather than wait for the machinery of government to act.” In fact, the difference in Americans’ attitudes between 1958 and 2016 was almost 20 percentage points on this question. Americans in 2016 also were more likely to agree that “The true American way of life is disappearing so fast that we may have to use force to save it.”
Finally, figure 3 also shows an American public becoming more supportive of some tenets of liberal democracy. For example, compared to the 1950s, in 2016 fewer Americans agreed that “We might as well make up our minds that in order to make the world better a lot of innocent people will have to suffer.” Moreover, significantly fewer Americans agreed in 2016 that “To bring about great changes for the benefit of humankind often requires cruelty and even ruthlessness.” Furthermore, Americans in 2016 were less likely to agree that “Politicians have to cut a few corners if they are going to get anywhere.”
Regarding free speech and procedural rights, Americans in the 1950s were mostly more conservative (see figure 4). The first two questions revealed that the American public was more supportive of censorship in the 1950s than in 2016. Substantially more Americans in the 1950s also agreed that “Any person who hides behind the laws when he/she is questioned about their activities doesn’t deserve much consideration.” That said, the mass public today is more conservative on one question: “In dealing with dangerous enemies like ISIS, we can’t afford to depend on the courts, the laws, and their slow unreliable methods.” To be fair, this is the only question we needed to change—McClosky asked the same question but about “Communists” instead of “ISIS.”
Figures 1–4 reveal a public somewhat more aligned with elements of liberal democracy on several issues. The most notable of these relate to racial and ethnic equality and to free speech and procedural rights. This change, however, is not uniform. Regarding procedural rules of the game, Americans have become more antagonistic toward this aspect of liberal democracy. Moreover, where change moves in favor of liberal democracy, these proportions also reveal significant numbers of people in 2016 who harbored anti-democratic sentiments.
LIBERAL DEMOCRATIC IDEOLOGY AND THE 2016 PRESIDENTIAL ELECTION
Our EFA (see the online supplementary materials) shows that respondents’ answers to these survey items load strongly on three dimensions. The first dimension, labeled racial supremacy, overwhelmingly consists of respondents’ answers to the survey items in figure 2. The second dimension, authoritarianism, comprises several questions in figure 3. The third dimension, elitism, consists of each question (except the last) regarding political equality presented in figure 1. We created three summary scales from only those items loading strongly on these three dimensions. For example, the elitism scale represents the summary score for respondents’ answers to questions that loaded well on this dimension, divided by the number of questions. Respondents’ scores vary from 0 to 1, where 1 indicates that a respondent agrees with each elitism item.
We wanted to determine if and how these summary scales influenced the following behavior in 2016: (1) vote preference, (2) vote choice, and (3) a combination of feeling thermometers for Hillary Clinton and Donald Trump. Vote preference is based on a survey question asking all respondents about their candidate preference, even if they abstained. First, we separated respondents on this variable into three mutually exclusive and exhaustive categories: prefer “other,” prefer Clinton, and prefer Trump. Second, we assessed citizens’ voting behavior in somewhat similar groupings: abstain, vote for Clinton, and vote for Trump.Footnote 7 Third, we gauged the difference in candidate affect (Trump versus Clinton), subtracting respondents’ feeling toward Clinton on a thermometer (0 to 100) from feeling toward Trump on a thermometer (0 to 100).
We used multinomial logits for the first two outcomes: vote preference and vote choice. This approach determined how our scales shaped the preference and choice of Trump versus Clinton (as well as Trump versus “other” or abstain). We fit an ordinary least squares (OLS) model to our feeling thermometer comparison. Reporting the results of only one outcome ran the risk of misstating the power of anti-democratic attitudes in the 2016 election.
We included several crucial control variables, such as an individual’s partisanship, ideology, and policy positions. Our policy positions scale combined respondents’ attitudes in five areas: gun control, abortion, environmental regulation, criminal justice, and immigration. (Survey questions are included in the online supplementary materials.) The policy positions scale was built by first creating a summary scale of our policy items and then normalizing it to vary from 0 to 1 in a conservative direction. A respondent whose score was equal to 1 took the most conservative position on each policy question. Other factors being equal, Republicans relative to Democrats should be decidedly more favorable toward Trump, and likewise in the case of conservatives relative to liberals as well as respondents whose policy views are more conservative than liberal.
We also controlled for race, gender, and education. There were four dummy variables for race: Black, Latino, other, and white (non-Latino) as the baseline category. Gender consists of women (1) and men (0). Finally, we used an ordinal variable to capture education: from no high school (0) to postgraduate (5). We expected whites to favor Trump more than any other racial group and men to favor Trump more than women. We were agnostic about the effect of education, even though on the campaign trail Trump expressed his deep affection for the “poorly educated” (Hafner Reference Hafner2016).Footnote 8
We also included a political sophistication variable for only the vote-choice model. This variable was important in differentiating abstainers from either Trump or Clinton voters. When included in our other models, it was neither significant nor changed our conclusions; therefore, we excluded it because there was insufficient theoretical justification. Citizens’ amount of political sophistication, however, drives turnout. This variable is based on three knowledge questions: (1) Which party is more conservative?; (2) Which controls the US House?; and (3) Which controls the US Senate? Values of this variable equaled the proportion 0 to 1 for these three questions that a respondent answered correctly.
Finally, we controlled for respondents’ economic perceptions. First, we used a variable that captured how citizens feel about the national economy, based on a question asking respondents if they believe the nation’s economy has gotten much worse (0) to gotten much better (5) during the past year. Our second measure captured individuals’ perceptions of their own economic circumstances, based on a question that asked respondents if their household income decreased a lot (0) to increased a lot (5) during the past four years.
Table 1 presents the results. Of particular importance is how the three scales performed. Two of the three scales reached conventional levels of significance, which supports the idea that citizens’ attitudes toward liberal democracy shape their political behavior. Elitism was significant across all three models. Citizens who harbor stronger elitist attitudes felt warmer about Clinton, preferred Clinton to Trump, and were more likely to vote for her. In contrast, respondents expressing sentiments of racial supremacy felt warmer about Trump than Clinton and preferred Trump to Clinton. In terms of actual voting behavior, however, racially conservative respondents were more likely to abstain than vote for either candidate.
Liberal Democratic Values and Electoral Preferences and Behavior in 2016
Notes: First column derived from OLS fit to Trump thermometer minus Clinton thermometer. Second and third columns derived from multinomial logit fit to vote preference, with “prefer Clinton” omitted. Fourth and fifth columns represent multinomial logit fit to voter preference with “vote Clinton” omitted. + p<0.1, * p<0.5 (two-tailed).
Figure 5 clarifies how elitism and racial supremacy shaped the preference for Trump over Clinton based on the second and third coefficients in table 1. For purposes of the plot, we set the respondent as an independent, moderate, white woman. We replicated this figure for partisans in the online supplementary materials. We set other variables at their average values.Footnote 9 The figure demonstrates that moving this hypothetical citizen’s level of racial supremacy from its minimum to maximum flips her preference from slightly Clinton to slightly Trump. It also shows that shifting her level of elitism from its minimum to maximum flips her preference from slightly Trump to slightly Clinton.
Racial Supremacy, Elitism, and the Preference for Trump over Clinton
The thermometer model tells a similar story. Compared to a respondent at the minimum value for elitism, an individual at the maximum value for elitism was more than 12 degrees warmer toward Clinton than Trump. However, relative to a respondent who was at the minimum for racial supremacy, an individual at the maximum for racial supremacy was almost 15 degrees warmer toward Trump. As expected, our findings also demonstrate that partisanship, ideology, policy preferences, race, and political sophistication shaped respondents’ electoral preferences. The policy preferences scale has a notably large effect in all of the models. Net of other crucial variables, citizens who prefer conservative policies to liberal policies strongly preferred Trump to Clinton, voted for Trump, and felt warmer toward Trump.
CONCLUSION
More than half a century since McClosky’s (Reference McClosky1964) seminal study, his battery of liberal democracy questions continue to help us understand contemporary political behavior. We found that elitism and racial supremacy affected attitudes and behavior in the 2016 presidential election. Elitist respondents aligned with Clinton regarding affect, presidential preference, and vote choice. Furthermore, based on the tenor of the McClosky questions tapping elitism, Clinton’s statement that many of Trump’s supporters were a “basket of deplorable[s]” (Reilly Reference Reilly2016) is a valid characterization of elitism. These sentiments go a long way to help us comprehend what underpins our current discontents—from mass demonstrations against racism and police killings, to the toppling of Confederate monuments and other symbols of white supremacy, to the militant defiance of social distancing and mask-wearing mandates of state and local governments during the COVID-19 pandemic.
Public opinion on racial and ethnic equality—since McClosky’s survey—is the one domain in which Americans have become uniformly more democratically liberal. Nonetheless, respondents scoring higher on the racial-supremacy dimension expressed greater affect for Trump and preferred him over Clinton—even though these individuals were not more likely to vote for him. Trump’s racially antagonistic rhetoric (Valentino, Neuner, and Vandenbroek Reference Valentino, Neuner and Vandenbroek2018) comports with our evidence that racial supremacy was positively linked to his candidacy and reflects “white Americans leveraging political anger toward electoral behavior more effectively than racial minority groups” (Phoenix Reference Phoenix2019, 23). The racial animosity toward African Americans by a non-inconsequential number of whites, as investigated by Banks (Reference Banks2014), continues to amplify political differences.
Almost 60 years after McClosky (Reference McClosky1964), we found that the mass public has not become noticeably more politically sophisticated or welcoming of liberal democracy. Americans, however, have become more adept at recognizing the increasingly polarized partisan disagreements fueled by political elites, and these divisions have prohibited a one-sided movement in favor of a consensus on the many values constituting liberal democracy. According to the American National Election Studies (ANES), in 1956, only 8% of eligible American voters had graduated from college; in 2016, fully 31% had at least a college degree.Footnote 10 Despite a more-educated populous, there is no evidence that the mass public has become more learned in the domain of civics and thereby more appreciative of liberal democracy. Indeed, we found that the mass public remains woefully deficient in reaching a consensus levelFootnote 11 on the majority of questions measuring support for liberal democracy.
Our evidence reveals at least two avenues that future researchers should pursue in an effort to learn more about public opinion on liberal democracy. First, this study shows that liberal democratic attitudes shape voter choices and preferences net of crucial control variables, but why do some people harbor more democratic attitudes than others? Which covariates explain the range of attitudes that we observed on the three dimensions that our EFA revealed: racial supremacy, authoritarianism, and elitism? Stated simply, whereas our study demonstrates the consequences of these attitudes, we encourage future scholars to investigate their sources.
Second, we encourage future researchers to develop new survey items in an effort to validate these dimensions. We were motivated to replicate McClosky’s (Reference McClosky1964) survey items to learn how much (or how little) Americans have changed on many of these ideas, and this effort required that we borrow his specific language in drafting survey items. Admittedly, however, some items used unusual language by today’s standards. Developing and administering new items would reveal how much—or whether at all—that language matters.
DATA AVAILABILITY STATEMENT
Replication materials are available on Dataverse at https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/TL3BLS
SUPPLEMENTARY MATERIALS
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S1049096520000918.
