J. Eric Oliver,
Thomas J. Wood
First published: 5 March 2014Full publication history
DOI: 10.1111/ajps.12084View/save citation
Cited by: 18 articlesCitation tools
This research was supported by a research grant from the University of Chicago. We are grateful to Betsy Sinclair, Kirk Hawkins, and attendees at the 2011 CCES conference for their suggestions on this research. Replication data and code for this article may be found at the AJPS Dataverse website at https://thedata.harvard.edu/dvn/dv/ajps.
Although conspiracy theories have long been a staple of American political culture, no research has systematically examined the nature of their support in the mass public. Using four nationally representative surveys, sampled between 2006 and 2011, we find that half of the American public consistently endorses at least one conspiracy theory and that many popular conspiracy theories are differentiated along ideological and anomic dimensions. In contrast with many theoretical speculations, we do not find conspiracism to be a product of greater authoritarianism, ignorance, or political conservatism. Rather, the likelihood of supporting conspiracy theories is strongly predicted by a willingness to believe in other unseen, intentional forces and an attraction to Manichean narratives. These findings both demonstrate the widespread allure of conspiracy theories as political explanations and offer new perspectives on the forces that shape mass opinion and American political culture.
Throughout their history, Americans have demonstrated high levels of suspicion towards centralized authority and their political elites (Barber 1983; Hart 1978). Often these sentiments go beyond a general distrust of government and encapsulate fears of larger, secretive conspiracies. From the anti-Catholic and anti-Masonic movements of the nineteenth century to the “Red Scares” of the twentieth, Americans periodically have organized themselves around narratives about hidden, malevolent groups secretly perpetuating political plots and social calamities to further their own nefarious goals, what we would define as “conspiracy theory” (Davis 1971). Today, conspiratorial theories exist on subjects ranging from the Kennedy assassination to the 2013 Boston Marathon bombings and appear to have wide circulation in the mass population. For instance, in a recent study by Stempel, Hargrove, and Stempel (2007), nearly a third of American respondents agreed that federal officials either assisted in the attacks of September 11th or did nothing to stop them in order to go to war in the Middle East.
Although scholars have long theorized about the “paranoid style” of American politics (Barkun 2003; Fenster 1999; Hofstadter 1964), none have estimated the pervasiveness of conspiratorial thinking in the general public or empirically demonstrated why Americans do endorse conspiracy theories. Most scholarship about conspiracy theories in America has focused more on interpretive analyses of the theories themselves rather than on empirical research about their support in the mass public (e.g., Clarke 2002; Davis 1971; Marcus 1999; Melley 1999). The few empirical studies are inconclusive because they asked only a few scattered questions either about specific theories (e.g., Stempel, Hargrove, and Stempel 2007) or about conspiratorial reasoning among specific subpopulations (e.g., Barreto et al. 2011; Crocker et al. 1999; Darwin, Neave, and Holmes 2011; Douglas & Sutton 2008; Goertzel 1994; Parsons et al. 1999). To our knowledge, there is no research that systematically examines support for a wide selection of conspiratorial narratives across a representative sample of the entire American population. Given the historical pervasiveness of conspiratorial thinking, this is itself a significant oversight in studies of American politics and public opinion.
More importantly, if such conspiracy theories are as widely accepted as both the historical record and previous research suggest, then it should force us to reconsider our general understanding of both mass opinion formation and American political culture. Most scholarly models prioritize elite discourse and ideological predispositions as the driving engines of public opinion (e.g., Erikson, Luttbeg, and Tedin 2010; Zaller 1992), yet most conspiracy theories directly contradict mainstream explanations for public events and are usually suspicious of political elites. This presents something of a puzzle: if public opinion is so determined by elite discourse, then how can a set of beliefs that openly question the sincerity of political elites and the dominant narratives for political events be embraced by the mass public? Widespread belief in conspiracy theories (what henceforth will be referred to as “conspiracism”) would suggest that a set of unrecognized factors shapes the way most Americans understand politics.
This article examines the extent and determinants of conspiracism in the United States. We theorize that conspiracism is much like conventional forms of public opinion in that it is motivated by specific political messages and individual predispositions; what differentiates conspiracism is the content of its motivating narratives and the types of predispositions it evokes. Specifically, conspiracism is animated less by misinformation, paranoia, or political mistrust, and more by attributional proclivities that are commonly expressed in supernatural and paranormal beliefs. Conspiracism is also motivated by the compelling narrative structures of most conspiracy theories themselves, particularly in their Manichean worldview.1 Four nationally representative survey samples collected in 2006, 2010, and 2011 indicate that over half of the American population consistently endorse some kind of conspiratorial narrative about a current political event or phenomenon and that these attitudes are predicted by supernatural, paranormal, and Manichean sentiments. These findings suggest that conspiracism is not only an important element in American political culture, but also is expressive of some latent and powerful organizing principles behind American mass opinion.
Conspiracism as a Form of Public Opinion
Given the fantastical and implausible assertions of many conspiracy theories, it is understandable that they are often dismissed as manifestations of a latent psychopathology (Clarke 2002; Robins and Post 1997), a product of gross misinformation (Berinsky 2011), or a “crippled epistemology” (Sunstein and Vermeule 2009). For this research, we remain decidedly agnostic about the truth claims, accuracy, or epistemological integrity of common conspiracy theories. Our interest is simply in explaining why some people endorse them. We start with the assertion that conspiracy theories are simply another type of political discourse that provides a frame of interpretation for public events. We also consider conspiracism as simply a particular form of public opinion and, as such, subject to the same defining influences of conventional mass belief (Erikson, Luttbeg, and Tedin 2010). Like ordinary public opinion, conspiracist opinion is highly influenced by encounters with elite discourse, in this case the conspiracy narratives.2 Whether citizens accept these narratives depends on their prior predispositions.3 When members of the public are asked about their conspiracist beliefs, they effectively “sample” from the mix of prior information, public signals, and predispositions to generate a survey response, the same as they do with conventional political questions (Zaller 1992).
What distinguishes conspiracism from conventional opinion is the nature of its animating political narratives and the latent predispositions that it activates. Although the sheer number and variety of conspiratorial narratives defy any easy categorization, previous scholarship suggests that most conspiracy theories have three common characteristics (Barkun 2003). First, they locate the source of unusual social and political phenomena in unseen, intentional, and malevolent forces. Second, they typically interpret political events in terms of a Manichean struggle between good and evil. As Hofstadter (1964, p. 29) famously described, “the distinguishing thing about the paranoid style is [to think] that a vast or gigantic conspiracy is the motive force in historical events. History is a conspiracy, set in motion by demonic forces of almost transcendent power” Finally, most conspiracy theories suggest that mainstream accounts of political events are a ruse or an attempt to distract the public from a hidden source of power (Fenster 2008).
The question remains about what types of predispositions these conspiracy narratives will activate. Most studies of mass opinion focus on conventional sources of attitudinal difference, such as partisanship, ideology, or racial identity. Although some scholars argue that conspiracism is a defining feature of the political right (Barreto et al. 2011; Sunstein and Vermeule 2009), it is not clear why conspiracy theories would be appealing to only one end of the political spectrum. Other researchers suggest that conspiracism originates in psychological predispositions such as anomie, authoritarianism, self-esteem, cynicism, and agreeableness (Abalakina-Paap et al. 1999; Darwin et al. 2011; Douglas and Sutton 2008; Goertzel 1999; Swami et al., 2011), but these studies have severe data limitations and their findings remain inconclusive.4 After testing many of these predispositions in studies of British and Austrian subjects, Swami et al. (2011) arrive at the seemingly tautological conclusion that “the strongest predictor of whether or not an individual will ultimately accept a conspiracy theory is the presence of earlier conspiracist ideation” (459).
We suggest this “conspiracist ideation” is not simply the fact of believing in another conspiracy theory but is actually derived from two innate psychological predispositions. The first is a propensity to attribute the source of unexplained or extraordinary events to unseen, intentional forces (Shermer 1997). In psychological studies, this tendency is often found in supernatural, paranormal, or religious beliefs (Boyer 2001; Norenzayan and Hansen 2006; Tobacyk and Milford 1983). We suggest this predisposition originates in a highly adaptive and unconscious cognitive bias to draw causal connections between seemingly related phenomena (Cottrell, Winer, and Smith 1996; Michotte 1963) and to presume predators are behind unknown or novel stimuli (Barrett 2004; Kassin, Fein, and Markus 2007). A common example is when someone presumes that a malevolent force is behind a strange noise in a dark house at night. As with many religious or superstitious beliefs, these types of causal attributions project feelings of control in uncertain situations (Guthrie 2001; Keinan 1994; Newheiser, Farias, and Tausch 2011).
The second is a natural attraction toward melodramatic narratives as explanations for prominent events, particularly those that interpret history relative to universal struggles between good and evil. This type of Manichean narrative is common in both religious and populist political rhetoric (Hawkins 2003) and is consistent with a fundamental pattern in all narratives to set up core oppositions between contradictory elements (Frye 1957; Greimas 1983). These patterns, in turn, seem to arise from basic processes in human cognition, particularly in the retrieval and storage of information (Boyd 2009). Narrative structuring of information is central to human cognition and facilitates the organization of memory and reasoning (Patterson and Monroe 1998). We hypothesize that conspiracy theories are attractive precisely because their Manichean narrative structures better comport with how some people process political information and because they provide compelling explanations for otherwise confusing or ambiguous events (Fenster 1999).
Together, these two factors — the proclivity to make causal attributions of salient phenomena to unseen forces and an attraction to Manichean political narratives — will explain why many otherwise ordinary people may embrace conspiracy theories. Several features of these predispositions are important to note. First, they are distinct from psychopathologies like paranoid personality disorder (Zonis and Joseph 1994) or attitudinal concepts like political trust. Although people with unusual levels of anxiety, paranoia, or personal mistrust are also likely to be attracted to conspiracy narratives, believing in unseen forces or liking Manichean narratives is not irregular and would not otherwise impair “normal” functioning in society. Indeed, our supposition is that these predispositions originate in cognitive tendencies that would appear normal or even appropriate in other circumstances, such as knocking on wood for good luck. And unlike political trust (Hetherington 1998; Stoker and Weir 2001), these predispositions are not always oriented toward specific regimes, although some conspiracy theories are highly partisan in nature. Instead, these predispositions represent more of a general orientation towards understanding public events.
Second, these predispositions also differ from paranoia in that they will crystallize into a specific set of attitudes only after someone encounters a particular conspiracy theory. Whereas the paranoid may see enemies everywhere, there is no evidence that people who endorse one conspiracy theory also see conspiracies behind all political events; this is partly why Hofstadter (1964) famously differentiated the “paranoid style” of conspiracy theories from paranoia. We suggest that most political phenomena or social events are simply not salient enough to motivate a suspicion of conspiracy. For example, we find no evidence that Americans see secret conspiracies behind farm subsidies, food stamps, or tax deductions for home mortgages because, we suspect, these policies are not extraordinary or contentious enough political issues.5 In addition, few people may “conspiracize” all political events as a paranoid might because fabricating one’s own conspiracy theories is so cognitively taxing, particularly in the face of widely disseminated counter-explanations. Instead, we hypothesize that most people will only express conspiracist beliefs after they encounter a conspiratorial narrative that gives “voice” to their underlying predispositions, assuming the particular incident was unusual or salient enough to invoke these feelings in the first place. Take the example of “Birther” conspiracies about President Barack Obama. There was almost no discussion of Obama’s citizenship after he addressed the Democratic Party convention in 2004; these views only became widespread after Birther narratives began circulating in 2008, largely in accordance with Obama’s likelihood of getting the Democratic nomination for president.6
Third, the likelihood that a person subscribes to any one particular conspiratorial narrative will be contingent on his or her other beliefs and his or her exposure to further political messages.7 Preexisting ideologies or religious beliefs will not only shape the type of conspiracy that is endorsed (e.g., Republicans are more likely to believe Birther conspiracies than Democrats) but also influence the particular elements of the conspiratorial narrative (e.g., evangelicals are more likely to believe in supernatural sources of conspiracy than atheists). Consequently, even though the predispositions that motivate conspiracism should be consistent across a wide range of conspiracy theories, we should not expect a single “style” of conspiracism, a uniform embrace of all conspiracy theories, or for conspiracism to be limited to one side of the ideological spectrum. Rather, a myriad of conspiratorial expressions should be endorsed by the public that are distinguishable from, and sometimes incommensurate with, each other.8 In other words, conspiracy theories that are informed by a particular ideology, such as “Birthers” or “Truthers,” will be different from those more uniformly suspicious of prominent figures or political institutions (e.g., conspiracy theories about Queen Elizabeth, the Masons, and the Trilateral Commission).
Together, these points are important for anticipating who is likely to embrace conspiracy theories. Past research suggests that supernatural and paranormal attributions are more common amongst the less educated and the most marginalized segments of the population (Vyse 1997). However, because supernatural, paranormal, and populist beliefs are so widespread in the American populace (Humphrey 2007), we hypothesize that the relative differences across demographic indicators will be small. Most importantly, we hypothesize that these predispositions will be far more important in identifying supporters of conspiracy theories than many of the commonly cited psychological or political variables, such as authoritarianism, interpersonal trust, or conservatism, when examined across a representative survey sample.
Data and Methods
To examine the nature of conspiracism in the American public, four nationally representative surveys were fielded in 2006, 2010, and 2011 as modules in the Cooperative Congressional Election Studies (CCES).9 The supporting information lists the wordings of the various conspiracy theories and the distribution of responses to Likert-scaled questions about them across the 4 different surveys.10 Table 1 lists the question wordings and the distributions from the combined 2011 samples on seven conspiracy theories.
Table 1. Percentage of Americans Agreeing with Various Conspiracy Theories, 2011
The surveys show that Americans have a high degree of familiarity with conspiracy narratives and exhibit high levels of agreement with them. For instance, almost the entire sample in 2011 said they had heard of at least one of the conspiratorial narratives they were asked about, and over 55% of respondents in 2011 agreed with at least one of them. The most widely endorsed conspiracy theory was the Financial Crisis conspiracy, endorsed by 25% of respondents and rejected by only 37%. Next in popularity was the Obama Birther conspiracy theory, endorsed by 24%, followed close behind by the Truther, Iraq War, and Soros conspiracy theories, which each found agreement from approximately 20% of the sample. Of these last four, the Birther, Truther, and Iraq War conspiracy theories elicited disagreement from roughly 50% of the sample as well, indicating that most people had strong opinions about them one way or another. The ChemTrails conspiracy theory about vapor exhaust from high- flying jets, a longtime staple in conspiracy theory circles, was endorsed by only 9% of the sample and was rejected by over 60%.
One immediate concern with these findings is how much respondents are really endorsing these conspiracy theories as opposed to either offering “non-attitudes” or because they are exhibiting acquiescence bias toward the Likert scales. Evidence in support of the non-attitudes hypothesis appears to be found in a 2011 survey item about the statement “The U.S. government is mandating the switch to compact fluorescent light bulbs because such lights make people more obedient and easier to control” (i.e., the CFLB conspiracy). This particular conspiracy narrative was made up by the researchers and has not been visible in public discourse. Yet, 17% of respondents said they had heard of this conspiracy, and 10% said they agreed with it. Such findings raise concerns that all of the conspiracy items are not capturing authentic sentiments.
Although measurement error and acquiescence bias are undoubtedly a part of these results, there are several reasons for believing that these survey items mostly capture “authentic” sentiments. First, the responses to the CFLB conspiracy are not necessarily evidence of a non-attitude; respondents may simply be reporting a familiarity with the very real government mandates about compact fluorescent lights, and their agreement with the statement may be triggered by a conspiratorial predisposition, issues that will be discussed below.11 Second, across the other survey items, there is a remarkably high level of consistency in the distributions of the survey items across the 5 years these questions were asked and across the four survey modules. If responses to these items were the result of non-attitudes or acquiescence bias, there would be much greater variance in the distribution of responses across categories (Saris and Sniderman 2004).12
Third, the most common response among respondents who had not heard of a conspiracy theory was to say they neither agreed nor disagreed with the statement; if survey respondents were giving non-attitudes as answers, such differences would not exist. Fourth, there is relatively little overlap in agreement among the conspiracy items. Among the respondents in 2011 who agreed with at least one conspiracy theory, about half endorsed just one and about 27% endorsed only two. Looking at the entire sample, we find that only 12% of respondents endorsed three or more conspiracies. Similarly low levels of conspiracy congruence also occur in the 2006 and 2010 samples. If the results were a manifestation of serious acquiescence bias, then one would find higher levels of multiple agreements.
Finally, those respondents who do endorse multiple conspiracies generally do so in ways that are ideologically consistent. Table 2 depicts loadings from an unrotated factor analysis and a pair of bifactor analysis of the conspiracy items from the two 2011 CCES studies.13 The unrotated factor analysis reveals two major dimensions. The first is what we label a “general” conspiratorial dimension. All of the items load highly on the general dimension, but especially the more anomic Financial Crisis, CFLB, and ChemTrails conspiracies. This is in contrast with items that load highly on a second “ideological” dimension, including the Iraq War and Truther items (which have large negative scores) and the Birther and Soros conspiracies (which have high factor scores).14 The ideological characteristics of these four conspiracy theories are also evident in the bifactor analysis that tests for a secondary ideological conspiracy theory and separate liberal and conservative conspiracy theories. These results are remarkably robust to model specification. All models indicate a general dimension loading positively on all conspiracy items, and then a separate ideological dimension, loading in a variegated fashion on particular conspiracies. Whether we estimate a single ideological dimension in the bifactor design (the second model), or adopt the canonical bifactor approach of allowing a separate factor for each ideological pole (the third model), the same ideological pattern emerges.
Table 2. Unrotated Exploratory Factor Loadings and Confirmatory Factor Estimators (Bifactor) of Conspiratorial Dimensions
The dimensionality of conspiratorial thinking is also evident when comparing average scores across the general and ideological conspiracy scales by individual self-reported ideology.15 As illustrated in Figure 1, there are few differences and no linear trends in the average scores on the general conspiracy scale by individual self-reported ideology. Self-described liberals or moderates are not significantly more likely to agree or disagree with conspiratorial statements about vapor trails or compact fluorescent light bulbs than self-described conservatives. This is in sharp contrast with the average scores on the conservative-scored ideological conspiracy scale, where a strong linear shift occurs. Self-described liberals score over one point lower, on average, on the ideological conspiracy scale than self-described conservatives.
Figure 1. Open FigureDownload Powerpoint slide
Mean Scores and Ideological and General Conspiracies by Self-Reported Ideology. Source: 2011 CCES; N = 1,985
In sum, these simple statistics indicate three important characteristics of conspiracism in the American public. First, conspiracism is a widespread and stable aspect of public opinion, with most Americans being familiar with a wide range of conspiracy narratives and roughly half agreeing with at least one; as illustrated in the supporting information, the level of agreement with these conspiracy theories stays remarkably constant over the 5 years we have asked about these theories and across different samples in 2011. Second, conspiracy narratives are differentiated somewhat between those that are uniformly suspicious of power and those that are ideologically oriented. Third, conspiratorial reasoning is not simply a style of one political group but is evident across the ideological spectrum and manifests itself in a variety of distinguishable forms.
Predictors of Conspiracism
Earlier, we suggested that these patterns of conspiracism arise from a predisposition toward making attributions to unseen forces and an attraction toward Manichean narratives. These ideas were tested with a number of indicators in one of the 2011 surveys. Two items were designed to directly measure each of these sentiments: a measure we created called Secret Cabal, which asks respondents how much they agree with the statement “Much of what happens in the world today is decided by a small and secretive group of individuals,” and a measure called Manichean, which is drawn from survey work on populism (Hawkins 2007) and questions agreement with the statement, “Politics is ultimately a struggle between good and evil.” We also use other measures to gauge the willingness to make attributions to unseen phenomena, including questions of belief in supernatural phenomena (e.g., the Devil and angels) and paranormal phenomena (e.g., ghosts and extrasensory perception [ESP]).16 Finally, we further tried to gauge the attraction of Manichean narratives with agreement to the statement “We are currently living in End Times as foretold by Biblical prophecy” (End Times). The distributions of these items are listed in Table 3.
Table 3. Americans’ Attitudes on Predispositions Related to Conspiracism
Americans express a broad willingness to believe in paranormal or supernatural forces and exhibit a strong agreement with Manichean narratives about politics. Roughly a third of the sample says they either believe in ghosts or agree that some people have ESP; these two items are highly correlated (.536) and can be combined into a Paranormal belief scale (alpha = .78). Even higher percentages of Americans believe in angels (66%) or “the Devil” (57%); once again, these two beliefs are highly correlated (.785) and are combined into a Supernatural belief scale (alpha = .83). Over half of the sample agreed with the Secret Cabal statement, over a third agreed with the Manichean worldview statement, and nearly a quarter agreed with the end times statement. Of course, some of these items also tap into distinct belief systems—end times and supernatural phenomena are religious sentiments, whereas paranormal beliefs are not. Consequently, we do not expect all of these items and scales to be highly correlated, even if we are using them to measure the same underlying predispositions. And this is what the data show. Among these items, the only high correlation (.497) occurs between the supernatural belief scale and agreement with the End Times statement, a consequence of both ideas being rooted in similar religious doctrines; none of the other items have correlations above .28. A full list of the correlation coefficients is in the supporting information.
Although these predispositions are common in the population, their support varies by demographic and attitudinal traits. Table 4 lists coefficients from ordinary least squares regressions estimating the effects of education, sex, age, race, ideology, and partisanship on the five measures of conspiracist predisposition. To differentiate the conspiratorial predispositions from related attitudes, measures were also included in the equation for interpersonal trust, external and internal political efficacy, religiosity, and authoritarianism.17
Table 4. Demographic and Attitudinal Predictors of Conspiratorial Predispositions
Outside of education, there are few consistent predictors of the predisposition measures. Less educated respondents routinely score higher on all the predisposition scales, but, beyond this, the predictors vary more according to individual traits and the particular predisposition in question. For example, women are more likely to score higher on the Secret Cabal and Paranormal measures and Democrats lower on the End Times measure. Blacks, conservatives, and the less politically knowledgeable are morel likely to agree with the End Times statement, even when controlling for their greater religiosity. Liberals are less likely to believe in supernatural phenomena, Republicans and authoritarians more likely. Not surprisingly, people with high religiosity scores are more likely to believe in supernatural phenomena and End Times, but also are more likely to agree with the Manichean measure. Together, these findings suggest that both types of predispositions are widely distributed across the population and vary by a wide variety of political and psychological variables. But most importantly, they suggest these predispositions are not a uniform expression of any one demographic or psychological characteristic (beyond being less educated).
We now turn our attention to models that predict support for conspiracy theories. In the supporting information, we list a table showing the results of ordinary least squares regressions with the responses to the 2011 conspiracy questions regressed on the predisposition measures (i.e., the 5-point Supernatural and Paranormal scales and the 5-point Likert responses for the Secret Cabal, Manichean, and End Times statements) and other measures, including education, race, sex, ideology, partisanship, political interest, and factual political knowledge. To differentiate our predisposition measures from religious sentiments, we also include a measure of religiosity.18
Although the results of the equations vary somewhat depending on the particular conspiracy theory in question, some overall patterns are clearly evident. Less educated respondents exhibit higher levels of conspiracism, as do African Americans. These results are consistent with prior research that suggests conspiratorial cognition is more common among the socially disempowered (Crocker et al. 1999). “Ideological” conspiracies find differential support by individual-level partisanship and ideology. For example, self-identified Republicans and conservatives are more likely to agree with the Soros and Birther conspiracies than independents or moderates, respectively. Self-identified liberals were no more likely than moderates to identify with any conspiracy theory, but Democrats were far less likely to identify with the Birther conspiracy than Independents.
Most interestingly, attitudes like interpersonal trust, political efficacy, and authoritarianism have no consistent relationship with conspiracism. Although less trusting and more authoritarian respondents were slightly more likely to agree with the financial crisis conspiracy, these effects are relatively small and are not seen in any other equation. Neither are conspiracy adherents uniformly less politically informed either—political knowledge is only a significant predictor among the ideological Truther, Birther, and Iraq War conspiracy theories, with the more knowledgeable being somewhat less supportive. Conspiracism appears only to be indicative of greater ignorance when it coincides with ideological priors.19
The most robust predictors of conspiracism are the predisposition measures. We present a depiction of the coefficients and standard errors of these measures in Figure 2. Not only are respondents who score highly on the Manichean, End Times, Secret Cabal, and Paranormal scales uniformly more likely to agree with all of the conspiracy theories, but the magnitude of these effects are also far greater than any other variable. The strongest predictor of conspiracism is agreement with the End Times statement. The consistency of this predictor is remarkable considering that end times belief is much more prevalent among political conservatives, yet end times adherents are also more likely to subscribe to all of the conspiracies, including the more liberal Truther and Iraq War statements. The Manichean variable is the second strongest predictor of conspiracism and, in the instance of the Financial Crisis conspiracy, actually exceeds the predictive power of the End Times variable. Following behind these variables are the Paranormal and Secret Cabal measures, which are strongly predictive of all but the Birther conspiracies. Indeed, the only predisposition measure that does not positively relate to conspiracism is the supernatural belief in angels and the Devil. Indeed, holders of these supernatural beliefs are significantly less likely to support the liberal Truther and Iraq War conspiracy theories. Partly, the low effects of the Supernatural scale is the function of also having the measure of End Times belief in the same equation, but the negative relationship to the liberal conspiracy theories is contrary to our hypothesis.
Figure 2. Open FigureDownload Powerpoint slide
Regression Coefficients of Predisposition Measures on Conspiracy Beliefs
Note: S = Supernatural; M = Manichean; E = End Times; C = Secret Cabal; P = Paranormal. Insignificant effects are indicated by hollow points. Full equations are in the supporting information.
Despite their robustness, it is reasonable to wonder whether these predispositions represent a distinct source of variation in conspiratorial attitudes, particularly in relation to the other independent variables; in other words, are respondents more likely to support conspiracy theories because of these innate predispositions or because these innate predispositions represent latent differences in their political interest, education, sex, or ideology? To test the second concern, we conducted a series of mediational analyses, adopting Baron and Kenny’s (1986) method to estimate whether any of the independent variables are mediated by the five conspiratorial dispositions (see also Imai, Keele, and Tingley 2010). The purpose of these mediation models is not to establish prior causality but to determine whether the effects of the predisposition measures are large and significant in their own right and not because they capture latent differences from the other independent variables.20 The fully estimated set of causal mediation effects for the five predisposition measures, the remaining independent variables, and the seven conspiracy theory items are depicted in Figure 3. The full model specification and an example of how to interpret the figures are provided in the supporting information.21
Figure 3. Open FigureDownload Powerpoint slide
Results of Causal Mediation Analysis
Note: Point size is the product of the proportion of a variable’s absolute effect that is mediated, and the ratio of the size of this effect to that variable’s largest effect on any conspiracy.
Source: 2011 CCES.
In nearly all instances, the effects of the five predispositions occur independently of the other explanatory variables, as indicated by the proximity of the hollow points to the dashed lines. For instance, the impact of End Times, Manichean, or Paranormal belief on support for conspiracism is rarely a consequence of differences in education, race, partisanship, ideology, or political interest, knowledge, or efficacy. The only exception to this is with the interpersonal trust item: although interpersonal trust rarely has a direct relationship with conspiracism, in several instances, the Manichean and Secret Cabal measures are mediating the effects of trust. In other words, one reason why people who have a Manichean worldview or believe that a small, secret group is behind much of what happens in the world are also likely to agree with conspiracy narratives about 9/11, the Iraq War, fluorescent lights, or Obama’s birth certificate is because they are less trusting of people in general. Nevertheless, these significant mediated effects are uncommon; instead, the causal mediation models demonstrate that conspiracist predispositions are largely independent of the other measured factors in predicting support for conspiracy theories.
In both the popular media and scholarly community, it is quite common to disparage conspiracy theories as an expression of either deluded and dangerous cranks (e.g., Sunstein and Vermeule 2009), right-wing zealots (e.g., Barreto et al. 2011), or the grossly misinformed (Berinsky 2011). Nationally representative survey data provide a much more complex picture. Although we do not have data on the active propagators of conspiracy theories, we do see that both the willingness to agree with conspiracy theories or see them as valid explanations for political phenomena are quite commonplace in the American public. Not only does half of the American population agree with at least one conspiracy from a short list of conspiracy theories offered, but also large portions of the population exhibit a strong dispositional inclination toward believing that unseen, intentional forces exist and that history is driven by a Manichean struggle between good and evil, particularly in the high proportion of Americans who believe we are living in biblical “end times.” Interestingly, conspiracism does not seem to be an expression of political ignorance. With the exception of those adherents of ideological conspiracy theories such as Birthers or Truthers, respondents who endorse conspiracy theories are not less informed about basic political facts than average citizens. Far from being an aberrant expression of some political extreme or a product of gross misinformation, a conspiratorial view of politics is a widespread tendency across the entire ideological spectrum.
The prevalence of conspiracism offers new possibilities for the study of political opinion. Ever since Converse’s (1964) seminal description of American belief systems, political scientists have struggled to identify the central organizing principles behind public beliefs. Given the near random quality to survey responses among half of the population, scholars have tended to focus their attention on variables like ideology and race, even if this yields only partial explanations for how ordinary citizens comprehend political life. Conspiracism illuminates some alternative mechanisms that organize public opinion. For many Americans, complicated or nuanced explanations for political events are both cognitively taxing and have limited appeal. A conspiracy narrative may provide a more accessible and convincing account of political events, especially because it may coincide with their ordinary cognitive tendencies.
The power of conspiracy theories is not limited, however, to the politically naïve. Even highly engaged or ideological segments of the population can be swayed by the power of these narratives, particularly when they coincide with their other political views. Just because someone adopts a political ideology or consumes a great deal of political information, it does not mean he or she will cease making attributions to unseen forces or will find Manichean narratives less intuitively compelling. In fact, many predominant belief systems in the United States, be they Christian narratives about God and Satan (LaHaye and Jenkins 2004) or left-wing narratives about neoliberalism (e.g., Klein 2007), draw heavily upon the idea of unseen, intentional forces shaping contemporary events. The fact that a Birther conspiracy persists largely among knowledgeable and engaged Republicans or that large numbers of liberals still maintain that “9/11 was an inside job” illustrates how powerful these elements can be for sustaining a particular set of views. Conspiracy theories can thus reveal not only how some people come to interpret specific events, but also the psychological tendencies of many people for understanding their political world.
This term is borrowed from early Persian religion, which placed particular emphasis on a contest between forces of light and darkness. In this context, a Manichean worldview is adopted when a person believes that political events are the consequence of a contest between good people and malevolent people, rather than between self-interested actors possessed of different perspectives and priorities.
For example, Zaller (1992) defines elite political discourse as providing “a depiction of reality that is sufficiently simple and vivid that ordinary people can grasp it. … [I]t is unavoidably selective and unavoidably enmeshed in stereotypical frames of reference that highlight only a portion of what is going on” (13). This would certainly characterize most conspiracy narratives.
Zaller (1992) defines predispositions as “stable, individual-level traits that regulate the acceptance or non-acceptance of the political communication the person receives” (22).
This is because such studies are either based on undergraduate populations (e.g., Douglas and Sutton 2008) or, if based on surveys (Goertzel 1999; Swami et al. 2011), they are not representative samples or do not control for a host of important factors such as age, education, race, political interest, and knowledge.
A Google search on terms like “Food Stamp Conspiracy Theory” or “Farm Subsidies Conspiracy Theory” fails to reveal any indication of such conspiracy theories actively propagated on the Internet.
The Obama Birther conspiracy narrative began during the Democratic primary in the spring of 2008 with anonymous email chains. These were picked up by online commentators during the general election. However, in the fall of 2008, only 60% of Americans reported hearing about this conspiracy and only 10% agreed with it. By April 2010, a Harris poll reported that over 80% of respondents had heard of this theory and that roughly a quarter of them (largely Republicans or ideological conservatives) believed that Obama had not really been born in the United States(Harris Polls 2010).
A conspiratorial predisposition is also different from concepts like alienation and political trust. Scholars have offered a wide range of conceptualizations of these terms (see Levi and Stoker 2000), and while conspiracism will undoubtedly share some elements with these concepts, it differs considerably in its targets. Political trust typically entails citizens’ evaluations of institutions and leaders relative to their governing performance at particular periods of time (Hetherington 1998). Although some conspiracy theories may emerge as mechanisms to discredit particular regimes (e.g., Obama as a secret Muslim or George W. Bush perpetuating massive voter fraud in Ohio in 2004), conspiracism is not always contingent on partisan political circumstance.
Hofstadter (1964) famously described the paranoid style as “heated exaggeration, suspiciousness, and conspiratorial fantasy” (***): with a focus primarily on conspiratorial narratives emanating from the political right. We suggest the existence of conspiracism in the mass public hinges more on the predispositions outlined above and thus allows for a greater ideological diversity.
The CCES sample was drawn from the YouGov/Polimetrix online panel, which recruits a large pool of respondents who make themselves available to complete surveys in exchange for points that can be redeemed for various gift items. YouGov/Polimetrix uses sample matching for each module so that the panel is generally representative in terms of various demographic and attitudinal traits of a national random sample (for a full description and comparison with other survey methods, see Ansolabehere and Shaffner 2011). In 2006 and 2010, these questions were asked in the second part of a panel survey design after the November elections, hence the smaller sample sizes due to the attrition in the survey research pool. In 2011, these same questions were asked in October and November on two different survey modules administered by teams from REDACTED and REDACTED.
The conspiratorial narratives were drawn from a number of sources, including a variety of Internet searches on search terms like “conspiracy theory” along with other contemporary phenomena. Some of these, like the Truther and Birther conspiracy theories, are widely disseminated in public discourse. Others come from more esoteric sources, yet we still found wide circulation of such ideas on the Web. For example, as of February 2012, a Google search on the terms “chemtrails conspiracy” and “George Soros conspiracy” yields over 3.1 million hits and over 8 million hits, respectively. In the 2011 surveys, respondents were first presented with a one-sentence conspiratorial narrative and asked whether they had heard it before; afterward, they were asked about how much they agreed or disagreed with it. In 2006 and 2010, they were asked only about their agreement with the statement.
There are other conspiracy theories that such lights contribute to greater fatigue or may serve as a weapon to induce mercury poisoning through a massive electromagnetic pulse.
Another check against acquiescence bias comes from research by Berinsky (2011), who finds remarkably similar distributions to the Truther and Birther items even though they were asked in an entirely different format.
The exploratory factor analysis is in Table 2.
These dimensions are also apparent when comparing polychoric correlations among the seven indicators of conspiratorial reasoning in the 2011 survey. The highest correlations are among those items that only load on the general conspiratorial factor (i.e., the CFLB, ChemTrails, and Financial Crisis conspiracies). The lowest correlations occur among items that load at the two ends of the ideological conspiracy dimension. The correlation between liberal conspiracy indicators (i.e., Truther and Iraq War) and the conservative conspiracy indicators (i.e., Soros and Birther) is less than .17 in all cases.
To construct the general scale, the three general conspiracy items from the 2011 surveys (ChemTrails, CFLB, and Financial Crisis) were all coded from strongly disagree (−2) to neither agree nor disagree (0) to strongly agree (2) and averaged. The ideological scales were calculated from averaging the two liberal items (Truther, Iraq War), which were reverse coded (i.e., strongly disagree = 2, strongly agree = −2), and the two conservative items (Soros, Birther). Thus, respondents with high scores on the ideological conspiracy scale are both strongly disagreeing with the liberal conspiracies and strongly agreeing with the conservative conspiracies.
For three of the items (ghosts, angels, and the Devil), respondents were asked, “For each of the following items, please indicate whether it is something you believe in, are not sure about, or don’t believe in.” They were then given a series of boxes to check. These question wordings were derived from a Harris poll (http://www.harrisinteractive.com/vault/Harris-Interactive-Poll-Research-Religious-Beliefs-2007-11.pdf). The distribution of our responses is roughly consistent with Harris Poll findings.
Interpersonal trust was measured by two questions: “Do you think most people would try to take advantage of you if they got a chance or would they try to be fair?” and “Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?” External efficacy was measured by two Likert scale items: “Public officials don’t care much about what people like me think” and “People like me don’t have much say in what government does.” Internal efficacy was measured by two Likert items: “I consider myself well qualified to participate in politics” and “Sometimes politics and government seem so complicated that a person like me can’t really understand what’s going on.” Authoritarianism was measured by a battery of three forced-choice items. Respondents were told, “Although there are a number of qualities that people feel children should have, every person thinks that some are more important than others. Listed below are pairs of desirable qualities. For each pair, please mark which one you think is more important for a child to have: Independence versus Respect of Elders, Obedience versus Self-Reliance, and Curiosity versus Good Manners.”
The measure of a respondent’s political knowledge was constructed from 11 binary indicators. Respondents were asked to name the party in control of the U.S. Senate and House, and their state’s upper and lower legislative chambers. They were then asked to name their current governor, their U.S. representative, and both U.S. senators. They were then asked to recall the vote of their U.S. representative and both senators in the debt ceiling negotiation. A simple Rasch item response model was used to estimate each respondent’s political knowledge from these observed binary responses. Religiosity was a scale of three items on church attendance, prayer, and the importance of religion.
To test whether the effects of the dispositions fundamentally differed between the informed, we ran a series of interaction terms between the measures of political interest and knowledge with all the predisposition terms. The results are depicted in the supporting information. In not a single case could we find a sgignificant interaction between the predisposition and the respondents’ interest or knowledge of politics.
Structural equation models (SEM) are an alternative way to specify these relationships, but they proved impractical here principally because of the small number of indicator variables, and the high correlation among the latent variables. This meant an SEM could not be identified.
Figure 3 presents over 400 separate mediational tests. Accordingly, if our theory predicted positive results (large mediation effects), a multiple correction procedure would be used to militate against type I error. However, since we predict null results (indicating our predispositions are a distinct source of variation in conspiratorial subscription), uncorrected results present a more conservative test of our claim. We are indebted to an anonymous reviewer for emphasizing this point.
- Eric Oliver is Professor of Political Science, University of Chicago, Pick Hall 518, 5828 South University Avenue, Chicago, IL 60637 (firstname.lastname@example.org).
Thomas J. Wood is a Ph.D. candidate in Political Science, University of Chicago, Pick Hall 518A, 5828 South University, Chicago, IL 60637 (email@example.com).