Political Orientation and Belief in Science in a U.S. College Sample

Political orientation and belief in factual and nonfactual information

Within society’s desire for scientific discovery, the search for truth becomes an especially powerful motivator in scientists’ endeavors. However, predicated upon this desire for truth is the necessity for objectivity and a rejection of absolute certainty without such truth ultimately being laden with a value judgment (Kuhn, 1977; Popper, 1972). Such rejection of certainty could prove difficult for those whose worldview strongly aligns with an ideology. Skeptical tendencies toward science are generally stronger among conservative rather than liberal individuals (Gauchat, 2012; McCright, 2011). Research on recent social events indicates that “fake news,” which often introduce conspiracy theories into political and social discourse, more often supports conservative political causes rather than liberal agendas, particularly as evidenced during the 2016 U.S. presidential election (Allcott & Gentzkow, 2017), which could subsequently predict more endorsement of nonfactual information among that populace. Indeed, the manner in which certain factions of conservative voters (e.g., those supporting Donald Trump and Ted Cruz in the 2016 primaries) typically evaluate misinformation often leaves them more receptive to misinformation in the form of intrinsically meaningless rhetoric that merely sounds impressive, a type of misinformation designed to foster a good impression without regard for the truth (Pfattheicher & Schindler, 2016).

The consequences, both medical and social, of lay skepticism of science necessitate determining what drives individual differences in support for scientific evidence, particularly factors underlying conservative individuals’ greater skepticism, and dismissal, of scientific truth. These consequences could be exemplified by conservative individuals’ generally greater opposition toward childhood vaccinations relative to liberal individuals (Rabinowitz, Latella, Stern, & Jost, 2016). One explanation for scientific disregard, particularly among conservative individuals, may be different patterns of evidence evaluation and truth-seeking values between liberal and conservative persons (Salvi, Cristofori, Grafman, & Beeman, 2016). Although individuals reporting liberal or moderately conservative ideology generally accept reported scientific facts, research suggests that extremely conservative individuals are likely to disregard certain facts if they are not consonant with their current established belief system (Gauchat, 2012; McCright, 2011; Rodriguez, Moskowitz, Salem, & Ditto, 2017; see Geher & Gambacorta, 2010; Winegard, Clark, Hasty, & Baumeister, 2018, for instances of liberal-specific science denial).

Based on the past research, the current research assessed participants’ relative agreement with factual and nonfactual statements, with the goal of determining whether differences in agreement with these statements among conservative, versus liberal, individuals can be explained by differences in endorsement of truth-seeking values. We hypothesized that, despite individuals’ greater endorsement of facts over untrue statements, individuals reporting stronger liberal political orientation would agree with facts more than participants espousing stronger conservative political beliefs. Additionally, this relation between political orientation and agreement with scientific facts was hypothesized to be mediated by differences in endorsement of truth-seeking values, with those espousing liberal ideologies demonstrating stronger endorsement of truth-promoting values.

Given that conservatism is represented by a constellation of personality and psychological variables, we considered it prudent to examine the possibility that a specific facet of conservatism might be responsible for variations in truth-seeking values and acceptance of science. Specifically, we assessed participants’ self-reported dogmatism. Dogmatism, the tendency to have unwavering, unjustified certainty of beliefs, refers to an individual’s propensity to hold rigid beliefs in spite of contradictory evidence, a characteristic that has been found to define the political attitudes of far-right conservatives. Furthermore, research frequently finds a strong positive relation between political conservatism and dogmatism. In contrast, liberal attitudes have been found to have strongly negative relations with dogmatic attitudes. These relations prompted us to assess both truth-seeking behaviors and dogmatism (Crowson, 2009). By assessing dogmatism in the context of our proposed mediation model, we were able to determine whether liberalism broadly is driving the relation between truth-seeking values and scientific belief or whether liberal individuals’ reduced dogmatism is responsible for our predictions.

Participants

A sample of 270 students (227 women, 40 men, and 3 identifying as “Other;” M_Age = 20.44 years, standard deviation (SD) = 5.40; 63.2% White) from a mid-sized public university in the Southeastern United States participated for course credit. A power analysis indicated that 191 participants would sufficiently detect small–medium effects (ρ = 0.20, β = 0.80). We sampled until the end of the university semester to ensure adequate sample size in the event we would have had to exclude participants from the final analyses. No participants warranted exclusion.

Materials and procedure

Scientific facts and nonfactual statements

Participants were asked to rate their agreement (“Please rate the extent to which you agree with the following statements”) with 20 scientific statements: 10 empirically (i.e., “true”) supported statements (M = 3.51, SD = 0.78) and 10 empirically unsubstantiated (i.e., “false”) statements (M = 3.35, SD = 0.85; see Table 1 for items) using a seven-point Likert-type scale (1 = not at All; 7 = completely). Items were gathered by the authors from various sources based on the public availability of the information. ^{1 2}

OPEN IN VIEWER

Table 1.

Factual and nonfactual statements.

Category	Items
Factual	Humans are responsible for climate change.
	Sharks existed before trees.
	Humans evolved through natural selection.
	The brain regenerates brain cells.
	A typical cumulus cloud weighs about 1.1 million pounds.
	Humans and dinosaurs did not coexist.
	Nintendo was founded in 1889.
	Human children don’t get boney knee-caps until they’re around three years old.
	One 18-inch pizza is more pizza than two 12-inch pizzas.
	There are more trees on earth than there are stars in the galaxy.
Nonfactual	The Great Wall of China is visible from space.
	Vaccines cause autism.
	Horoscopes are accurate.
	Bigfoot is real.
	Alcohol makes the body warmer.
	Shaving actually thickens hair.
	Humans evolved from chimpanzees.
	Sugar makes people hyperactive.
	MSG causes cancer.
	Humans only use about 10% of their brain.

MSG: monosodium glutamate.

Preliminary analyses

We first determined participants’ overall agreement with factual versus nonfactual statements. A paired samples t test indicated that participants agreed with factual (M = 3.51, SD = 0.78) more than nonfactual statements (M = 3.35, SD = 0.85), t(269)=2.54, p = 0.012, d = 0.19. We additionally conducted a pair of one-sample t tests comparing responses against the scalar midpoint (i.e., 4) to identify whether the aforementioned means were above or below. This afforded us the opportunity to determine whether participants demonstrated an overall categorical belief (mean significantly above 4) or disbelief (below 4) for facts and nonfacts. Somewhat surprisingly, participants displayed overall categorical disbelief for both facts and untrue statements (i.e., both means were significantly below the scale midpoint; factual statements, t(269) = −10.366, p = 0.001, d = 0.40, nonfactual statements, t(269) = −12.54, p = 0.001, d = 0.40), suggesting individuals were overall more skeptical than accepting of statements.

Primary analyses

See Table 2 for correlation coefficients. Individuals espousing stronger agreement with pledge items agreed more with true facts, whereas no relation emerged between pledge agreement and untrue statement endorsement. Thus, those with a greater truth-promotion propensity were agreed more with empirically supported facts but not with untrue statements. Additionally, a positive relation emerged between liberal political ideology and agreement with both facts and nonfactual statements, indicating that liberal persons were more accepting of both types of statements. Finally, a correlation emerged between liberal political ideology and truth-promotion tendencies, such that more liberal individuals reported stronger truth-promoting values.

OPEN IN VIEWER

Table 2

Correlations for individual differences.

	Dogmatism	Facts	Nonfacts	Pledge
Politics	−0.465**	0.154*	0.132*	0.146*
Dogmatism		−0.113†	−0.003	−0.264**
Facts			0.264**	0.164**
Non-facts				−0.046

Notes. †p<0.07, *p<0.05, **p<0.01.

Given the strong association between political orientation and dogmatism as well as both of their negative associations with fact endorsement, we found it prudent to determine the full extent of both predictors’ effects while accounting for the other. While controlling for dogmatism, partial correlations indicated a marginal positive association between political orientation and fact endorsement, such that more liberal individuals remain marginally more endorsing of facts after covarying out dogmatism, r(267) = 0.115, p = 0.059. Conversely, while controlling for political orientation, no association emerged between dogmatism and fact endorsement, r(267) = 0.048, p = 0.436. Taken together, these findings indicate that the effects of dogmatism and political orientation are independent of each other, with political orientation specifically predicting endorsement of scientific facts, independent of any association with dogmatism.

Mediation analyses

To determine whether pledge endorsement mediated the association between political ideology and belief in facts and nonfactual statements, data were submitted to Model 4 of PROCESS (Hayes, 2013) using 5000 bootstraps. In the first model, political orientation was the predictor, true statement agreement was the outcome, and pledge endorsement was the mediator. The second model was identical, except untrue statement endorsement was the outcome. Of these two models, only the first, with true statement endorsement as the outcome, was significant. Liberal political ideology predicted greater pledge endorsement (b = 0.08, standard error (SE) = 0.03, p = 0.016), with pledge endorsement predicting higher agreement with true statements (b = 0.13, SE = 0.05, p = 0.017). An indirect test for the effect of liberal ideology on fact endorsement mediated the link, 95% confidence interval (CI) (0.0002, 0.0272). A significant direct effect also emerged, 95% CI (0.0072, 0.1306). Thus, greater support for truth-promoting values among more liberal persons was associated with greater agreement with facts.

An alternative mediational model with pledge endorsement as the predictor, agreement with true facts as the outcome, and political orientation as the mediator was significant. Pledge endorsement predicting liberal ideology (b = 0.24, SE = 0.10, p = 0.016), and liberal ideology predicting agreement with true facts (b = 0.07, SE = 0.03, p = 0.029). Because both models with true statement agreement as the outcome were significant, the causal pathway may be different than our proposed model. One concern emanating from the significance of both the proposed and alternative mediation models stems from potential collinearity of the predictor variable and mediator. However, the correlation between political orientation and pledge endorsement was small (r = 0.146), suggesting collinearity was not a statistical concern. Therefore, the proposed model was chosen as our a priori theoretical model due to stronger theoretical and empirical support in previous literature; that is, evidence suggests that the espousal of a particular political ideology, which is often guided by parental exposure in early life, may precede individuals’ endorsement of truth-promoting values developmentally (e.g., Hoge, Petrillo, & Smith, 1982; Schwartz, Cote, & Arnett, 2005). ⁴

Exploratory mediation analyses

We found it prudent to conduct mediation models using each PTP subscale as potential mediators given that the Encouraging subscale was noticeably less correlated with the other two subscales, with Honor: r(270) = 0.694, p < 0.001; with Share: r(270) = 0.655, p < 0.001, than the other two subscale were with each other, Honor and Share; r(270) = 0.791, p < 0.001. This suggests that the subscales may be tapping into different facets of truth-promotional values, warranting individual mediation models with each subscale as the mediator were run. The Share subscale model is not discussed, as it was nonsignificant.

Limitations and future directions

Although these findings indicate that liberal individuals were more likely to endorse truth-seeking behaviors and agree with facts, several other potentially influential factors were not examined in this study that could have similarly influenced our results. One potential predictor that could have accounted for greater skepticism or openness includes participants’ social media use. Social media has contributed to a great deal of untrue statement spreading (Allcott & Gentzkow, 2017), given its pervasive audience. The amount of time individuals spend on social media may further influence exposure to false information.

One potential limitation for the current study could be the content of the facts. Despite the veracity of these facts, some facts may not represent “common knowledge” as well as others (e.g., “Sharks existed before trees”). These facts could have resulted in lower base rates of fact endorsement overall. Although we sought to utilize facts that were widely available through a basic Internet search, future research would benefit from specifically identifying true and untrue statements that were pretested as being largely well known in the mental lexicon.

The nature of these facts may have also produced the partisan difference in fact endorsement that are consonant with the ideological conflict hypothesis (Brandt, Reyna, Chambers, Crawford, & Wetherell, 2014; Crawford, 2014). That is, individuals demonstrate a systematic intolerance toward ideologically dissimilar or threatening entities. Among such threatening entities may scientific facts that would ultimately undermine individuals’ ideological stance. Much like conservative individuals, those espousing progressive ideologies are nonetheless susceptible to certain types of misinformation (Nilsson, Erlandsson, & Västfjäll, 2019). For example, although liberal individuals endorse the notion of climate change having an anthropogenic component to it more than conservative individuals, liberal individuals remain largely skeptical over scientific facts, including the safety of GMOs and nuclear energy as well as potential accuracy in stereotypes (e.g., Hamilton, 2015; Stern, West, Jost, & Rule, 2013). Liberal individuals’ greater endorsement of evolution as fact additionally has boundary conditions, particularly as it pertains to existence of biologically sex differences (Geher & Gambacorta, 2010). This skepticism could subsequently be the basis for progressive ideologies frequently undermining objectivity in sex research (Bailey, 2019). This domain-specific science denial necessitates future research that could address whether similar mechanistic processes underpin endorsement and skepticism for ideologically relevant facts.

Given the use of a preselected sample of university students, the current findings should be interpreted with caution when applied to participants from other populations (e.g., age groups, geographical regions), as generalizability is limited. Participants within this sample were all from the Southeastern United States, and therefore it is specifically to this population that the results of this research refer. Further research could benefit from considering how differing educational experiences, on top of other sociodemographic factors, could influence political ideology.

Although political orientation served as a significant predictor of endorsement for factual and nonfactual statements and endorsement of the pledge, the magnitude of these effects may necessitate a level of caution in interpretation. That is, these associations’ absolute effects are below what is typically considered a minimally practical effect size (i.e., r  = 0.20; Ferguson, 2009). However, such small effects may be unsurprising in the current study, as we were assessing especially subtle attitudinal differences of sociopolitical information. In fact, such subtleties in effects may actually make our results impressive in their own right and valuable for future research (Prentice & Miller, 1992). Most individuals ultimately possess a level of general acceptance toward various scientific principles, which could suggest difficulty in detecting differences in science endorsement already. These small findings could demonstrate identifying systematic differences in endorsement is indeed possible and could be the impetus behind future research to identify more robust predictors than general political affiliation that would elicit larger effects and therefore elucidate upon these partisan differences more clearly. For example, if conservatism is itself a predictor of skepticism toward scientific facts, future research could consider which aspects of conservatism (e.g., social dominance orientation, right-wing authoritarianism) would be most predictive of these differences, as simply liberal–conservative dichotomies could be insufficient in understanding scientific literacy (Gauchat, 2015).

Consideration for future interventions

Future research would benefit from examining which interventions could be effective for individuals whose trait characteristics are unconducive to support for truth-promoting values. Research suggests that individuals on each end of the political spectrum endorse different moralities, with conservatives showing greater moral concern for the group as a whole whereas liberals show greater concern for rights of individuals within groups (e.g., Graham, Haidt, & Nosek, 2009; Van Leeuwen & Park, 2009). Therefore, presenting true and false information in ways that prime unique moral sensibilities in conservatives and liberals could heighten belief in empirically supported facts and reduce belief in untrue statements in both groups. Another topic for future exploration could include whether there is a difference in fact belief between religious and nonreligious conservatives. Given religious individuals are more likely to disregard scientific facts, such as evolution, it may be that religious conservatives show greater disbelief in facts than nonreligious conservatives (Clobert & Saroglou, 2015). Much of the current findings are largely rooted in an understanding of the rigidity of attitude formation, which suggests considerable difficulty in producing meaningful attitudinal change. Future work may additionally benefit from employing a theoretical framework inspired by cognitive dissonance theory as means to change attitudes toward accepting long-term scientific facts (Festinger, 1957). Previous work indicates that attitude-disconfirming facts are largely perceived as ultimately being scientifically untestable among individuals, thereby precluding such facts from being accepted from an ostensibly skeptical public (Munro, 2010). Within this dissonance framework, researchers could present attitudinally dissonant information in a context that presents the findings as indubitably testable, which would necessitate individuals reduce their feelings of dissonance toward these facts by accepting them.