Abstract
Are so-called “birthers” best thought of as true conspiracy theorists, or are they merely partisans expressing a sharp dislike of Barack Obama? Recent work on conspiracy beliefs finds that “birthers” are the product of partisan and ideological motivated reasoning. In this manuscript, we explore how the measurement strategies we employ on public opinion surveys may influence the substantive conclusions we draw about conspiratorial beliefs, rumors, and misinformation. We find that partisan stimuli influence reported beliefs in several different conspiracy theories, and, subsequently, the relationships between individual stated beliefs in those conspiracy theories. The implications of these findings are discussed.
Are conspiracy theorists partisans? Are they merely liberals or conservatives? Recent work on conspiracy theory beliefs appears to be split on the answers to these questions. On the one hand, Miller et al. (2016) and Pasek et al. (2015) find that conspiracy beliefs are largely the product of partisan or ideological motivated reasoning. On the other hand, work in the social psychological tradition emphasizes individual differences such as psychological biases, authoritarianism, and alienation as sources of conspiracy beliefs. Furthermore, Uscinski and colleagues (2014, 2016), find that conspiratorial thinking – the general predisposition to see the world through a conspiratorial lens – cuts across partisan and ideological lines.
One possible explanation for the disparity in recent findings, and the focus of this paper, is in the empirical strategy employed to measure conspiracy beliefs. The scholars who find that conspiracism is the product of partisan or ideologically motivated reasoning all employ survey questions about beliefs in specific, partisan conspiracy theories (e.g. Obama “birther,” “death panels”), while the teams who find no such relationship tend to employ survey questions about non-partisan conspiracy theories (e.g. death of Princess Diana, “chem trails”) or general conspiratorial sentiments (e.g. “much of our lives are being controlled by plots hatched in secret places”). The main difference between these “types” of conspiracy theories is the object at the center of them. For partisan conspiracy theories, salient partisan groups and figures such as the major parties, major party candidates, and high-level partisan political figures are the targets of conspiracy theories. Other specific conspiracy questions are devoid of partisan political objects, but speak more generally of “the government” or other non-partisan elite groups such as the Illuminati or Freemasons.
Since political scientists are interested in individual beliefs in specific conspiracy theories that implicate major political actors and groups, this measurement strategy makes sense. Alternatively, stated beliefs in these types of partisan conspiracy theories may be indicative of partisan and ideological attachments more so than a predisposition to view the world through the lens of conspiracy. Indeed, the robust literatures on motivated reasoning (e.g. Lodge and Taber, 2013) and misinformation (e.g. Nyhan and Reifler, 2010) indicate that the mere presence of these recognizable partisan stimuli may substantially influence respondent reactions.
Data and analytical strategy
We take an experimental approach to testing the impact of partisan/ideological stimuli embedded in questions about belief in conspiracy theories. We employ three different questions, adapted from Saunders et al. (2016), about three different conspiracy theories, each of which has a different partisan figure or group as the main conspirator. The specific questions appear below:
Some people believe that
Some people believe that the
Some people believe that Jade Helm 15, a U.S. military training exercise which was ordered by President
In each case, respondents are provided with four response options ranging from the conspiracy definitely having occurred to definitely not having occurred, with probabilistic responses occupying the middle categories. The full response options are provided in the Supplemental Appendix.
For the first two questions, respondents were randomly assigned to either receive the partisan stimuli (Democratically-controlled and Bush administration, respectively) or a more general, non-partisan government stimuli (Congress and government, respectively). For the third conspiracy theory about Jade Helm 15, respondents were randomly assigned to receive a question where the conspiratorial activity is attributed to either Obama (the “true” Jade Helm 15 orchestrator) or Bush. Although there is no true control when it comes to the Jade Helm 15 question, our strategy allows us to test whether partisanship is capable of overriding reality since Jade Helm 15 is a conspiracy created in response to the Obama administration. Respondents answered one of two versions of each of the three questions. The order was randomized.
We fielded this experiment through Amazon’s Mechanical Turk crowdsourcing platform on November 2, 2016. Respondents were paid $0.70 upon completion of the survey. Information about the demographic composition of the sample appears in Table 1. While this sample is not representative of the national population, it is more representative than a convenience sample of college students (Berinsky et al., 2012). Furthermore, we have no reason to expect that the psychological mechanism prompted by exposure to partisan stimuli operates differently across demographic groups.
Demographic composition of MTurk sample.
Total n = 933.
Empirical Results
We test the effects of partisan stimuli on stated conspiracy beliefs by regressing responses to each of the three conspiracy belief questions on a dummy variable denoting whether the respondent received the partisan treatment or not, a dummy variable denoting whether the respondent identified as a Republican or not, a control for conspiratorial predispositions, and an interaction between the treatment and Republican dummy variables. Details about our measure of conspiratorial predispositions, which is very similar to that employed by Uscinski et al. (2016) and Nyhan et al. (2016), can be found in the Supplemental Appendix. The responses to all conspiracy questions are coded 1–4, where a 4 represents the most conspiratorial response. Table 2 presents the results of both additive models, where no heterogenous treatment effects are assumed, and interactive models, where we assume treatment effects vary by partisanship.
Heterogenous treatment effects by partisanship, controlling for conspiratorial predispositions.
OLS estimates with standard errors in parentheses.
Significant at p<0.05 level with respect to two-tailed test.
In none of the additive models is the treatment effect statistically significant, though each of the conspiracy beliefs is significantly related to partisanship and conspiratorial predispositions. Importantly, however, all interaction term coefficients are statistically significant. In other words, we observe heterogenous treatment effects by partisanship. Each of these effects persist in the face of a control for the predisposition to view the world in conspiratorial terms. 1
More specifically, Republicans who were exposed to the “Democratically-controlled Congress” treatment as part of the financial crisis conspiracy belief question were significantly more likely to express beliefs in the conspiracy than Republicans assigned to the control condition where no partisan stimulus was present. Similarly, Republicans were less likely to express beliefs in the employment statistics conspiracy when they were exposed to the treatment asserting that the Bush administration played a part in carrying out the conspiracy. Finally, Republicans were less likely than Democrats to express beliefs in the Jade Helm 15 conspiracy theory when Bush was implicated, and more likely than Democrats to express belief when Obama was implicated.
The results of the experimental manipulations in the Jade Helm 15 question are important because they suggest that partisanship can cut through “truth” even when it comes to conspiracy theories. Indeed, even though Jade Helm 15 was a scheme supposedly ordered by Barack Obama, this fact is either unknown or irrelevant to Democrats and Republicans assigned to the “Bush” experimental condition. If, on the one hand, the partisan figure “truly” at the center of a given conspiracy is irrelevant, we have more evidence for the dominmant effect of partisan motivated reasoning. If, on the other hand, the partisan figure “truly” at the center of a given conspiracy is simply unknown, we have evidence that researchers are capturing only partisan tensions with certain specific conspiracy belief questions. In this case, the conspiratorial element of the question is nonexistent. Regardless, it is clear that the choices regarding which conspiracy theories are queried and how they are queried – particularly when it comes to the presence (absence) of partisan stimuli – should be carefully considered before data is collected and interpreted.
Perhaps even more troublesome is the ability of the conceptual “error” induced by the presence of partisan stimuli to aggregate across conspiracy belief questions. Consider, for instance, the correlation matrix presented in Table 3. Each cell value is a Pearson product-moment correlation between responses to pairs of questions across all respondents. Note that the empty cells correspond to cases which do not exist in the data – no individuals received both the non-partisan and partisan stimuli when it comes to any of the three conspiracy belief questions, hence correlations cannot be computed for those question pairs.
Pairwise correlations between conspiracy belief questions.
Cell entries are Pearson correlation coefficients. All significant at p<0.05 level with respect to a two-tailed test.
The correlations between pairs of conspiracy questions which include conflicting partisan stimuli (i.e. one Democratic and one Republican stimulus) are substantially smaller than the correlations between question pairs with only one or no partisan stimuli. The correlation between the partisan financial crisis responses and the non-partisan employment statistics responses is much larger than the correlation between the partisan financial crisis responses and the partisan (Bush) employment statistics responses. We might expect this given the intense partisan divisions in responses to these conspiracy questions – individuals providing the more conspiratorial response to the partisan financial crisis question (Democrats) are not the same individuals who provide more conspiratorial responses to the partisan employment statistics question (Republicans), hence the low correlation. The same relationship holds across all pairs of question responses.
Thus, partisan stimuli can systematically alter the structure of the relationships between conspiracy beliefs. This would explain a latent factor structure where the two major parties or ideological camps serve as the two distinct sources of variance underwriting responses to questions about specific conspiracy beliefs, as Miller et al. (2016) find. While the factor analytic procedure – which is specifically designed to account for intercorrelations between item responses with latent factors – does produce multiple item scales that reduce random measurement error, it also simultaneously ensures that the meaningful estimated sources of variance in question responses is partisanship or ideology, rather than conspiracism. Absent partisan stimuli, the intercorrelations between general conspiracy beliefs may prove more indicative of a general predisposition toward conspiratorial beliefs (e.g. Wood, 2017) than partisanship.
Conclusion
Political scientists and social psychologists have focused their attention on different psychological antecedents of conspiracy beliefs. Positing that this divide was due to differing measurement strategies, we conducted a survey experiment showing that inclusion of even a single partisan stimulus has the power to substantially decrease conspiracy beliefs for members of one party, while increasing such beliefs among members of the other party. Indeed, partisan stimuli have the power to activate motivated reasoning, altering the motives behind responses to survey questions about conspiracy beliefs.
As such, researchers must carefully consider the potential influence of partisan and ideological factors in their research design. If partisan conspiracy beliefs are the construct of interest, survey questions including related stimuli are appropriate. However, when attempting to learn about the general predisposition toward conspiratorial thinking, for instance, other strategies – such as those employed by Uscinski and Parent (2014) or Wood (2017) – may be more valid.
Supplemental Material
Appendix,_Wording – Supplemental material for On the measurement of conspiracy beliefs
Supplemental material, Appendix,_Wording for On the measurement of conspiracy beliefs by Adam M. Enders and Steven M. Smallpage in Research and Politics
Footnotes
Correction (June 2025):
Declaration of conflicting interest
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplementary material
The supplementary material is available at: http://journals.sagepub.com/doi/suppl/10.1177/2053168018763596 The replication
files can be found at 
Notes
Carnegie Corporation of New York Grant
This publication was made possible (in part) by a grant from Carnegie Corporation of New York. The statements made and views expressed are solely the responsibility of the authors.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
