p.27

3

COLLEGE STUDENTS AND LEGALIZED MARIJUANA

Knowledge Gaps and Belief Gaps Regarding the Law and Health Effects

Douglas Blanks Hindman

Communication research has shown a resurgence of interest in the impact of partisan identification in communication processes. For example, self-identification as a Democrat or Republican, or as a liberal or conservative, predicts perceptions of hostility and media bias against in-groups and in favor of out-groups (Gunther, 1992; Gunther, Miller, & Liebhart, 2009; Gunther, Edgerly, Akin, & Broesch, 2012). Partisan identification is associated with motivated reasoning in which conclusions complimentary toward one’s in-group are favored over conclusions favored by nonpartisan experts (Kruglanski & Webster, 1996; Tabor & Lodge, 2012). Selective exposure to media that privileges one partisan group over all others is related to increased polarization between the groups (Dillipane, 2011; Iyengar & Hahn, 2009; Knoblock-Westerwick & Meng, 2011; Stroud, 2008).

The present study is informed by research in which group identity is shown to be a significant predictor of knowledge regarding heavily publicized topics. Drawing from the knowledge gap hypothesis (Tichenor, Donohue, & Olien, 1970) and its extension, the belief gap hypothesis (Hindman, 2009; 2012; Hindman & Yan, 2015), this study derives and tests hypotheses about changes in students’ knowledge about long-term and short-term health effects associated with marijuana use by minors during the first three years following the legalization of recreational marijuana in Washington state.

An underlying question is whether student perceptions of the stigma of marijuana use have changed following the institutionalization of the recreational marijuana industry in 2012. Student perceptions about stigma are addressed in questions about the individual’s feelings about social media portrayals of their own marijuana use. In an evolving media environment, social identity, identity expression, and social media portrayals are increasingly intertwined. In an evolving political environment, truth claims associated with a variety of implied risks—climate change, health care affordability, and sexual health—are increasingly expressions of identity rather than expressions of knowledge (Hindman, 2009; 2012; Hindman & Yan, 2015).

p.28

Marijuana Legalization in Washington State

On Nov. 6, 2012, Washington state voters approved by a margin of 55.7 percent to 43.3 percent Initiative 502 legalizing possession of up to one ounce of marijuana for private recreational use by adults over age 21 (Washington Secretary of State, 2012, November 27). The voter-based initiative was the result of a petition drive that produced over 270,000 certified signatures in 2011 (Ammons, 2012, Jan. 27). The measure went directly on the November general election ballot after the legislature failed to put the issue to a vote.

The campaign for I-502 was supported by $6 million in donations, including $2 million from Progressive Insurance executive Peter B. Lewis (Martin, 2012, November 7). Proponents cited the potential windfall of tax revenues to the state that would result from ending ineffective and expensive law enforcement actions. Indeed, the state has collected over $400 million in excise tax and $137 million in sales tax since June of 2014; total sales topped $1 billion in 2016 (www.502data.com). In July of 2016, the formerly unregulated medical marijuana market was merged with the recreational marijuana regulatory system under the renamed Washington State Liquor and Cannabis Board (Preparations in place . . . n.d.).

Notably absent from the discussion leading to the vote to legalize recreational marijuana for adults was concern regarding health effects associated with marijuana use. Correlational evidence shows significant neuropsychological decline among those who initiated marijuana use in adolescence and who persisted through midlife (Meier et al., 2012). The highest percentage of US marijuana users in 2015 was among college-aged persons (19.8 percent of persons aged 18 to 25). This age group had significantly higher percentages of users in 2011 to 2015 as compared with the percentages from 2002 to 2010 (Center for Behavioral Health Statistics and Quality, 2016).

p.29

Literature Review

Knowledge Gaps

The main contribution of the knowledge gap hypothesis was acknowledging that the effects of mass media are constrained by social structural disparities, such as socioeconomic status, that are associated with the rate at which various groups acquire knowledge. Contrary to popular conceptions about the role of media in democracy, the authors showed that the mass media widened status differentials in knowledge. Because knowledge is tantamount to power, media ultimately serve to reinforce and exacerbate power disparities among the population (Tichenor, Donohue, & Olien, 1970).

The concept of mass media reinforcement of power differentials within society is a key component of the knowledge gap. That component takes on added explanatory power in an era in which facts and science are often disputed, disregarded, or denied when the facts conflict with political agendas. The current polarized political climate is markedly different from that in which the knowledge gap hypothesis was first proposed (Abramowitz & Sanders, 2008; Fiorina, Abrams, & Pope, 2008; Political polarization in the American public, 2015, June 12).

The original knowledge gap hypothesis included assumptions about knowledge that do not square with the polarized political climate. In the knowledge gap hypothesis, knowledge about science and public affairs was assumed to be unproblematic and universally accepted (Tichenor et al., 1970). The only requirement for knowledge to be diffused throughout the population was sufficient time. Knowledge, in this model, was cumulative, so that all groups gained knowledge, albeit at different rates (Tichenor, Donohue, & Olien, 1970). Media publicity was the primary means of diffusing knowledge about science and public affairs throughout social systems.

Missing from the extensive knowledge gap literature were studies about knowledge claims that were politically disputed (Hindman, 2009). Elected officials, political pundits, and spokespersons tend to dispute facts that negatively affect their goals or which pose the perceived risk of burdensome governmental regulations. Acknowledgement of a scientifically established connection between cigarette smoking and cancer, for example, was disputed by researchers representing the tobacco industry in order to forestall governmental regulation (Proctor, 1995). Similarly, the oil companies’ own research first acknowledged the scientific consensus that the earth’s atmosphere was warming and that the changes in the atmosphere were the result of human activity. In 1981, Exxon Mobil used knowledge of the connection between CO2 and fossil fuels to make business decisions that would avoid future governmental regulation. The company funded 30 years of research designed to cast doubt on the connection between fossil fuels and global warming (Goldenberg, 2015, July 8).

p.30

Belief Gaps

The belief gap hypothesis is an extension of Tichenor, Donohue, and Olien’s (1970) knowledge gap hypothesis. The belief gap hypothesis and the knowledge gap hypothesis both are based on issues for which there is a “correct” answer—an answer that can be easily verified or that has reached a level of consensus among nonpartisan experts who conduct evidence-based research. The belief gap hypothesis states that, in an era characterized by political polarization, indicators of partisanship will be better predictors than educational attainment of knowledge about heavily publicized and politically contested issues (Hindman, 2009).

Knowledge gap-era assumptions about the unproblematic and cumulative nature of knowledge are suspended when the knowledge consists of verifiable facts that are politically disputed. In the belief gap hypothesis, as the name implies, beliefs are privileged over knowledge. Whereas knowledge requires evidence to support the claim, beliefs are mere statements of personal opinion that are not burdened with the requirement to provide supporting evidence.

An individual, elected official, political pundit, or representative of a special interest group is free to believe or not believe anything that is contrary to the individual’s self-interest, ideology, or values (Graham, Haidt, & Nosek, 2009). For example, spokespersons for tobacco industries, perceiving the risk of future governmental regulations that would affect company profits, learned that they could cast doubt on the scientific evidence linking smoking with cancer by stating, under oath, personal beliefs that were contrary to the facts (Proctor, 1995).

News reports of tobacco industry official’s testimony at congressional hearings helped disseminate doubts about the risks of smoking. News media, relying on journalistic norms of balance, reported both scientific evidence of cancer risks and the tobacco industry’s self-serving beliefs that cast doubt on the evidence. Industry officials knew that their doubts would be embraced by like-minded individuals throughout the country: individuals who were economically or physiologically dependent on tobacco (Proctor, 1995). This process, in which industry or political elites publicly express beliefs that are contrary to scientific evidence, is designed to exploit irrational impulses among individuals who are predisposed to accept the beliefs, regardless of the facts (Hindman, 2012).

Belief gaps are most obvious in polling about politically contested issues. This is because individuals filling out polls who don’t know the correct answer to a question about facts will instead express their beliefs about the issue at hand. Hence, any statement that appears to favor their predispositions will be marked as true. Statements that subjects believe to be negative toward their predispositions will be scored as false (Zaller, 1992). When confronted with a factual question to which they do not know the correct answer, respondents will often use the opportunity for political expression (Bullock, Gerber, Hill, & Huber, 2013, May). Poll respondents would rather exhibit a political predisposition than ignorance—even though the two appear to be precisely the same when judged against the truth.

p.31

This paper applies the belief gap hypothesis to an analysis of knowledge among young adults. In the current study, the realms of knowledge that are of particular interest are a) knowledge about the Washington state law legalizing the recreational use of marijuana, and b) knowledge about the short-term and long-term health effects of marijuana use.

The first type of knowledge—knowledge about the contents of the marijuana legalization law—represents easily verified facts. A similar study compared the ability of Democratic and Republican partisans to identify which statements represented contents of the Affordable Care Act (Hindman, 2012). In that study, the belief gap was supported by evidence that Democrats knew more about the contents of the law over time. Republicans, however, appeared to know less. Contrary to the knowledge gap hypothesis, knowledge was cumulative only for groups that benefitted politically from the knowledge.

The second type of knowledge under consideration in this study requires the respondent to rely on the consensus of experts who conduct evidence-based research about the health risks of marijuana use. A study that employed this type of knowledge showed widening gaps between conservatives and liberals on acceptance of the value of abstinence-only sex education; a scientifically discredited teaching method that was shown to be ineffective in preventing unwanted pregnancies and sexually transmitted diseases (Hindman & Yan, 2015). Another study that relied on the knowledge claims of nonpartisan experts, in this case, climate scientists’ consensus regarding the link between human activity and global warming, showed that ideology was a better predictor than educational attainment in accepting claims that there was solid evidence of global warming (Hindman, 2009). As with the knowledge gap, gaps in beliefs of liberals and conservatives widened over time.

In the knowledge gap hypothesis, both forms of knowledge—easily verified contents of heavily publicized laws and research about the health effects of marijuana use by minors—would be positively correlated with a measure of social status, operationalized as educational attainment. In the belief gap hypothesis, however, measures of social identity are expected to be better predictors than educational attainment of knowledge about marijuana laws and health effects.

p.32

Predictor Variables in Knowledge Gaps and Belief Gaps

Belief gaps are best conceptualized as special cases of the knowledge gap hypothesis. The knowledge gap hypothesizes a widening gap over time in knowledge between higher and lower status groups regarding heavily publicized scientific and public affairs issues (Tichenor, Donohue & Olien, 1970). On the other hand, the belief gap hypothesis is limited to cases in which the knowledge claim is politically disputed (Hindman, 2009). Knowledge as a criterion variable is replaced with beliefs, which accounts for cases in which individual beliefs are contrary to verifiable facts. In the case of facts that are not easy to verify, such as the cause of global warming or the effectiveness of abstinence-only sex education, a consensus of nonpartisan experts provide verification.

The main predictor variable in knowledge gaps is socioeconomic status, measured as educational attainment. Those with the highest levels of educational attainment are hypothesized to obtain knowledge at a faster rate than those with lower levels of education. The originators of the knowledge gap (Donohue, Tichenor, & Olien, 1975) provided an extension in the conceptualization of knowledge gaps. They argued that various indicators of social identity and social organization could serve as predictor of knowledge gaps. Community type and community economic structure were two indicators provided as examples of social organization affecting knowledge distribution among a social system. For example, studies showed knowledge gaps narrowing in communities in conflict. Similarly, communities engaged in conflict with outside groups showed considerable gaps in knowledge between local and non-local groups (Tichenor, Donohue, & Olien, 1980). More recent work has argued that indicators of identity are significant predictors of holding beliefs contrary to scientific consensus (Veenstra, Hossan, & Lyons, 2014).

The belief gap hypothesis also relies on indicators of social identity as predictors of beliefs about verifiable facts. Identification as a conservative or liberal, and political party identification, were both shown to be the best predictors of beliefs regarding politically contested but verifiable facts (Veenstra et al., 2014).

As Table 3.1 shows, knowledge gap and belief gap hypotheses share similarities and significant differences regarding the issues considered, the predictor and criterion variables measured, and the presumed impacts.

Social Identities and Risk Perceptions

Social identity is related to perceptual bias when the group is at risk for a negative health outcome. In a pioneering study, intravenous drug users and homosexuals significantly underestimated risk behaviors associated with contracting AIDS when compared with college students who were not members of the at-risk groups (Campbell & Stewart, 1992). The minimization of known risk behaviors was consistent with the need to preserve one’s positive identity as a member of a group (Campbell & Stewart).

p.33

A more contemporary study (McCright & Dunlap, 2013) demonstrated that even poorly defined social groups could have an impact on risk perceptions. It their study, conservative white males were significantly less concerned than the rest of the population about environmental risks such as the quality of the environment, pollution, and global warming. The researchers cited two cognitive mechanisms that help explain the rejection of environmental risks: identity-protective cognition and system justification (McCright & Dunlap, 2013, p. 212–214). Identity-protective cognition that appears as fearlessness among white males dismissing environmental risks is really defensiveness in protecting identity-central activities such as employment in energy-dependent industries (Kahan et al., 2007). Conservative white male group identification was enhanced through conservative media outlets promoting the system justification arguments of conservative white male elites occupying leadership roles in conservative politics and energy-dependent industries (McCright & Dunlap, 2013).

p.34

Social Identities Among College Students

The case of college student beliefs about the legalization of marijuana provides a number of potential extensions to the belief gap hypothesis. First, it offers an opportunity to observe belief gaps in the absence of media coverage of polarized debates among Republicans and Democrats. This is because legalization of recreational use of marijuana was not politically contested in the state. In the absence of Republican-Democrat polarization on the issue, citizens were not exposed to pundits and elected officials making public claims for or against the issue. Without news reports of elite contestation of evidence-supported claims, one would not expect citizens to associate issues surrounding marijuana legalization with their own political identities.

Another challenge and opportunity provided by the study is college students’ indifference to partisan politics (Kiley & Dimock, 2014, September 25). Because college students’ political identities are often not fully developed, partisan identity would not be expected to be a significant predictor of beliefs about the law or about the health risks of marijuana use by minors. Instead, other indicators of social group identity and self-interest in the norms of the group may come into play. The study provides the opportunity to consider aspects of social identity that might be relevant to legalization of recreational use of marijuana.

Religious Conservative Identities and Legalization of Marijuana

Although students may lack political identities, there is no shortage of opportunities for students to build an identity that they can share with others. Students can develop identities based on majors, living arrangements, and social group affiliations.

Of particular interest here is the degree of student identification as a conservative and religious person. Conservative groups in general tend to oppose legalization of recreational use of marijuana in favor of criminal approaches to control. Nationally, 47 percent of all Americans supported legalization of marijuana, up from 27 percent in 1979, and a larger percentage of Democrats than Republicans (51 percent vs. 27 percent) supported the legalization of marijuana in late 2012 (Backus & Condon, 2012, November 29). Federal law classifies marijuana as a Schedule 1 drug similar to heroin or cocaine because of its potential for abuse, threat to safety, and medical insignificance. However, the US Department of Justice has pledged to not interfere with states such as Colorado and Washington that have approved legalization of medicinal and recreational use of marijuana (Southhall & Healy, 2013, August 29).

p.35

Individuals affiliated with conservative community churches or with national conservative Christian organizations tend to publicly announce a formal set of beliefs, adhere to behavioral norms, and frequently attend social events and worship services with church members (Evangelical Protestants, 2015). Chemical dependency in general is anathema to conservative religious groups, which instead stress self-denial and dependence on the group’s religious dogma (DeWall et al., 2014; Stewart, 2001).

Historically, religious groups have opposed marijuana use because of its tendency to intensify the user’s focus on the present, as opposed to the spiritually transcendent, aspects of existence. The drug’s effects are contrary to conservative religious emphases on self-denial, sacrifice, and the afterlife (Pollan, 2002). Religious identity would be expected to be a key predictor of differences with non-religious and non-conservatives in the perception of verifiable facts about the contents of the law and the health risks of marijuana use.

Frequency of Marijuana Use as a Social Identity

Another social group that would be relevant to questions about legalized marijuana would be the subculture surrounding marijuana use (Johnson, Bardhi, Sifaneck, & Dunlap, 2006). Status as a marijuana user would also likely become woven into a student’s social identity. Students who frequently use marijuana would likely surround themselves with other users in order to share the experience of using and to participate in other social events while under the influence.

Students who are users of marijuana would be expected to view knowledge claims about health risks and the contents of the law through the perspective of self-interest and in support of the norms of the group. Among users, one might assume that statements about the contents of the law would be viewed positively, because of their group’s support for the law, and statements about negative health effects of marijuana use would be viewed with skepticism or disbelief because the statements do not reflect well on the users’ daily habits.

p.36

Acceptance of Social Media Portrayals of Marijuana Use and Social Identity

Social media has become an important means of communicating individual and group social identities in an evolving media environment (Dominick, 1999; Livingstone, 2008; 2014; Taddicken, 2014). Among some individuals, the social media post about an activity is more important than the activity itself. If being a Christian conservative or marijuana user is a key part of one’s identity, then attitudes toward social media portrayals of one’s use would be closely tied to that identity. Hence, positive attitudes toward social media portrayal of one’s own marijuana use would be related to rejection of knowledge claims regarding the negative health effects of marijuana use.

Hypotheses

In keeping with the belief gap hypothesis, measures of educational attainment would be less predictive of knowledge than would indicators of social identity. The issues chosen to test the hypotheses in this study are: easily verifiable facts about the contents of the law (H1a, H1b, H1c, RQ1) and statements representing evidence-based scientific research about the short-term and long-term health effects of marijuana use (H2a, H2b, H2c, RQ2).

Based on the previous discussion, the following hypotheses can be stated for testing:

H1a: Measures of social identity will be better predictors than educational attainment of accurate beliefs about the contents of the law that legalized the recreational use of marijuana.

Individuals who tend to oppose an issue may express that opposition in polls. One way is to deny the truth of verifiable facts or expert knowledge claims that appear complimentary to the issue. Another way is to express belief in verifiable facts or knowledge claims that appear to put the issue in negative light. This is consistent with earlier findings by Jamieson and Cappella (2008, pp. 230–235) in which partisans were shown to agree with statements that portray their side favorably or portray opponents unfavorably, and to disagree with the converse: statements that portray their side unfavorably or that portray opponents favorably. It is also consistent with Bullock et al. (2013, May) who argue that partisans use knowledge questions in polls to express opinions consistent with their party’s stance.

For example, Republicans appeared to know less over time about easily verifiable facts regarding the contents of Obamacare (Hindman, 2012). This was likely because the Republican participants tended to deny the aspects of the Affordable Care Act that had widespread acceptance, such as allowing parents to keep dependents on their insurance until age 26 and requiring insurance companies to accept clients regardless of preexisting health conditions (Hindman, 2012).

p.37

In the current study, one would expect that various indicators of social identity will affect beliefs about the contents of the law, regardless of the truth. First, the expected negative predispositions toward the law legalizing recreational use of marijuana among conservatives, Republicans, and self-identified religious people would be reflected in their knowledge level about the easily verified facts regarding the contents of the law.

H1b: Identification as a religious conservative will be negatively associated with accurate beliefs about the contents of the law legalizing recreational use of marijuana.

Second, frequent users of marijuana and those who embrace social media portrayals of their own use would show higher levels of knowledge about the law. This is because the respondents’ self-interest is served by knowing the details of the law that legitimizes their own use of marijuana.

H1c: Frequency of marijuana use, and approval of social media portrayals of one’s own marijuana use, will be positively associated with accurate beliefs about the contents of the law.

The belief gap hypothesis predicts that social identity will grow over time in influence on individual beliefs. This is an extension of the knowledge gap hypothesis, which predicts a divergent interaction between educational status and time as predictors of knowledge. Previous studies have shown significant interactions between partisan and conservative social identities and with time. The current study provides an extension of the belief gap hypothesis into issues that are not politically contested, and among a population that is not overtly partisan. Hence, the relationships among social identity, time, and beliefs about the contents of the law that legalized marijuana will be stated as a research question.

RQ1: Is there a relationship between the interaction of Time x Social Identity on the level of accurate beliefs about the contents of the law that legalized marijuana?

Whereas one would expect that opponents of the law would have lower levels of knowledge or beliefs about its contents than would proponents, the opposite would be expected for information that states facts about the negative health effects of marijuana use. Individuals who are opposed to marijuana use would be more likely to accept as true statements about negative health effects. Proponents of marijuana use would have the opposite reaction; they would express a disbelief in statements about negative health effects. In either case, indicators of social identity would be a better predictor of beliefs about the negative health effects of marijuana use than would educational attainment.

p.38

H2a: Measures of social identity will be better predictors than educational attainment of beliefs about the negative health risks of marijuana use.

Those who are inclined to oppose marijuana use would tend to believe statements about the negative health risks of marijuana use. In the current study, beliefs about the negative health effects of marijuana use would be associated with identification as a religious conservative.

H2b: Identification as a religious conservative will be positively associated with beliefs about the negative health effects of marijuana use.

Similarly, those who are predisposed to favor marijuana use because of their identities as frequent users, or as those who approve of social media portrayals of their use of marijuana, would not likely believe negative statements about the health effects of marijuana use.

H2c: Frequency of marijuana use, and approval of social media portrayals of one’s own marijuana use, will be negatively associated with beliefs about the negative health effects of marijuana use.

The traditional test of belief gaps over time is based on issues in which political elites contest verifiable facts based on partisan or ideological beliefs. Intense media coverage of the controversy widens over time the gaps between partisans in beliefs about the issue. As was stated above, the current issue of legalization of recreational use of marijuana was not politically contested in the state. Further, college students tend to be less politically polarized than older generations and elected officials. Hence, the nature of the relationships among social identity and beliefs about the health effects of marijuana use over time is best stated as a research question.

RQ2: Is there a relationship between the interaction of time and social identity on beliefs about the negative health effects of marijuana use?

p.39

Methods

Overview

Data for testing the hypotheses came from three online surveys of students of 18 years of age or older in communication classes at a public university in Washington state. Each survey was reviewed by the university IRB and received exempt status. Students were recruited using an online subject pool management system, and received extra credit for their participation. Cases were deleted in which the respondent completed the survey in fewer than five minutes, left the majority of items blank, or showed evidence of straight-lining—giving the same response to a series of items.

The first survey was fielded following the November 2012 election and directly before the new Washington state marijuana law went into effect, Nov. 27–Dec. 7, 2012. The survey produced 312 usable cases. The second survey was fielded April 16–May 1, 2014, before legal retail sales began, and produced 402 useable cases. The third survey was fielded Sept. 22–Nov. 7, 2015, and produced 295 usable cases.

Predictor and Criterion Variables

Table 3.2 shows the means for each of the main predictor and criterion variables (operationally defined below) that were used in the study. The variable “month” was used to indicate the months between the first survey and the subsequent two. Hence, for the 2012 survey, “month” was zero; for 2014, “month” was 15; and for the 2015 survey, “month” was 34.

Measures

Female indicates the self-reported gender identity of the respondents with 0 representing male and 1 representing female. Table 3.2 also shows that 61percent of the respondents identified as female, which reflects the population of the classes from which the students were self-selected. The percentage of females was significantly lower in 2014 than in the other two years.

Educational achievement was operationalized as educational aspiration. If using traditional operationalizations of educational attainment, the entire sample for the present study would fit into the same category: “some college, no degree.” Obviously, a sample of college students is relatively homogenous in terms of educational attainment. A more straightforward operationalization, “year in college,” was not equally distributed across the years of the study. Instead, respondents were asked, “How far do you think you will go in school?” with options for “some college, but less than four years,” “graduate with my Bachelor’s degree,” “Graduate or professional school,” and “I am not sure.” The 12 cases in the final category were assigned to the variable mean of 2.2. This operationalization of educational attainment as educational aspiration was the best option, given the nature of the dataset.

p.40

p.41

Religious Conservative (Chronbach’s α =.72) was formed as the mean of standardized measures of four items: self-placement on a liberal-conservative scale (very liberal, liberal, moderate or not political, conservative, very conservative); self-placement Democrat-Republican scale (1= strong Democrat or Democrat, 2 = Independent; 3 = Republican or Strong Republican; Religiosity (1 = not at all religious, 2 = not very religious, 3 = religious, 4 = somewhat religious, 5 = very religious); frequency of attending religious worship services (never, on major holidays only, several times a year, a couple of times a month, once a week, several times a week). Table 3.2 shows that the 2014 sample scored significantly higher on the religious conservative scale than did the respondents in the 2012 sample.

Frequency of marijuana use was computed first as the sum of four standardized indicators of frequency of marijuana use (Chronbach’s α = .86). The first frequency indicator measured how much marijuana the respondents have used in their life (seven items ranging from 0 times to 100 or more times). The next item measured how much marijuana the respondents have used in past thirty days (I didn’t use marijuana, 1 or 2 days, 3 to 5 days, 6 to 9 days, 10 to 19 days, 20 to 29 days and all 30 days). The third question measured typical marijuana use on one occasion (I don’t use marijuana; 1–2 small hits on a marijuana cigarette, pipe, or bong, or 1–2 small bites of a marijuana food product; 3–5 hits on a marijuana cigarette, pipe, or bong, or 3–5 bites of a marijuana food product; smoke a whole joint, pipe, or bong, or eat an entire marijuana product; or smoke more than one joint, more than one pipe or bong, or consume a lot of marijuana product). Finally, the last question measured self-perception of marijuana use status with six items (non-user, very light user, light user, moderate user, heavy user, and very heavy user). The scale was positively skewed because 28 percent of respondents chose the non-user option for each item. The scale was recoded into a four-level measure with 0 representing non-users, .25 representing those respondents who identified as a user on one of the four indicators, .5 for those choosing two indicators of use, .75 for three indicators, and 1.0 for those who identified as a user on all four indicators. Table 3.2 shows that there were no significant differences across years on the frequency of marijuana use scale.

Approval of social media portrayals of one’s own marijuana use (Chronbach’s α .82) is a measure of five items, each preceded with, “How would you feel if friends or acquaintances of yours posted on social media pictures of you using marijuana?” The choices were “proud;” “angry” (reversed); “happy,” “ashamed” (reversed); “embarrassed” (reversed). The five responses were “very (4), somewhat, a little, not at all (1).” The index was computed as the mean of the five measures. An examination of the distribution of the scale indicated a positive skew resulting from 35 percent of respondents selecting the lowest value on each scale item. To control for skew, the scale was trichotomized so that zero represented the 35 percent of respondents who chose the lowest value on each item in the scale, 1 represented the 30 percent of respondents that scored between 1 and 2 on the scale, and 2 represented the 35 percent who scored higher than 2 on the scale.

p.42

Beliefs about the contents of the law legalizing recreational use of marijuana was computed as the number of statements correctly identified as true about the Washington state marijuana law: It will be legal for those over age 21 to possess up to two ounces of marijuana (false); all adults 21 and over will be allowed to have up to six marijuana plants in their homes (false); It will be legal to purchase marijuana only from state-licensed providers (true); it will be illegal to use marijuana in public (true); individuals under age 21 will be allowed to distribute marijuana (false); persons under age 21 will not be allowed on the premises of licensed marijuana retailers (true); driving under the influence of marijuana will be prohibited (true); advertising of marijuana sales will be restricted (true); taxes will be applied to the sale of marijuana by producers, processors, and retailers (true). In 2012 respondents scored significantly lower on the index measuring knowledge of the state law legalizing recreational use of marijuana.

Beliefs about the short-term health effects was computed as the number of items correctly identified as true about the “immediate health consequences of using marijuana” including: distorted perceptions (true); impaired coordination (true); problems with sleeping (false); difficulty with thinking or problem solving (true); sexual dysfunction (false); problems with learning and memory (true); numbness in the feet (false); increased heart related problems (palpitations, arrhythmias and heart attack) (true); ringing in the ears (false); there are no immediate health effects (false) (Ammerman et al., 2015; Drug Facts, 2017, February; Meier et al., 2012).

Beliefs about the long term health effects was also computed as the number of items correctly identified as true about the long-term health effects of marijuana use: compulsive drug seeking (true); tremors in hands (false); quitting withdrawal symptoms (true); anxiety (true); brain cancer (false); vision problems (false); depression (true); schizophrenia (true); sexual dysfunction (false); bone degeneration (false); increased respiratory problems (true); stomach cancer (false); irritable bowel syndrome (false); problems with short-term memory (true); decreased attention spans (true); learning impairments (true); lack of motivation (true) (Ammerman et al., 2015; Drug Facts, 2017, February; Meier et al., 2012).

In 2012, respondents scored significantly lower on the index measuring both short-term and long-term health effects of marijuana use than in 2014 and 2015. This provides some evidence of higher levels of awareness of health effects over time.

p.43

Results

To test the hypotheses, the three dependent measures—knowledge of the contents of the law legalizing recreational use of marijuana, knowledge of short-term health effects of marijuana use, and knowledge of long-term health effects of marijuana use—were regressed on the predictor variables in a hierarchical multiple regression. Table 3.3 shows the results.

The first hypothesis was stated as:

H1a: Identification as a religious conservative will be a better predictor than educational aspiration of beliefs about the contents of the law that legalized recreational use of marijuana.

p.44

As shown in Table 3.3, both the coefficient for educational aspiration (β = .02, n.s.) and the F testing the change in R² were not statistically significant. In contrast, the three indicators of social identity, entered as a block in the Hierarchical Multiple Regression showed a significant F test of the change in R² (6.4, p<.001). All three coefficients representing social identity were statistically significant: frequency of marijuana use (β = .07), religious conservative (β = -.08), and acceptance of social media portrayals of one’s own marijuana use (β = -.09). Together the findings support H1a.

The second hypothesis was stated as

H1b: Identification as a religious conservative will be negatively associated with beliefs about the contents of the law.

The test for this hypothesis is shown in Table 3.3 as the statistically significant negative coefficients for religious conservative (β = -.08, p < .05). This indicates that those identifying as religious conservatives were less able than non-religious conservatives to identify true statements about the contents of the law legalizing recreational use of marijuana. The hypothesis was supported.

The third hypothesis was stated as:

H1c: Frequency of marijuana use, and approval of social media portrayals of one’s own marijuana use, will be positively associated with beliefs about the contents of the law.

The test for this hypothesis is shown in Table 3.3 as the statistically significant and positive coefficient for the frequency of marijuana use variable (β = .07, p < .05). This shows that those with higher frequencies of marijuana use were more successful at identifying true statements about the law. Contrary to the hypothesis, the other social identity indicator, approval of social media portrayals of one’s own marijuana use, was negative and statistically significant (β = -.09, p < .01). In other words, those who approve of social media posts of their own marijuana use are less likely than those who don’t post to know about the contents of the law legalizing marijuana use. Hence, the hypothesis was partially supported.

The first research question was stated as

RQ1: Is there a relationship between the interaction of Time x Social Identity on beliefs about the contents of the law that legalized marijuana?

This research question is addressed by the interaction terms in Table 3.3. The F test measuring the change in R² was statistically significant (F 2.6, p < .05), which provides an indication of a relationship among the group of variables and knowledge about the contents of the marijuana legalization law. Two of the Time x Social Identity interaction measures were statistically significant, albeit one was only marginally significant (time x religious conservative, β = .06, p < .1) and was in the opposite direction as would be expected based on previous belief gap findings about declining knowledge among conservatives (Republicans) in knowledge about a disliked law (Hindman, 2012). The expectation would be that the more one knows about a disliked law, the more one disputes factual statements about its contents—particularly aspects that are widely viewed as favorable. The other significant coefficient, time x acceptance of social media images (β = .07, p < .05) was in the expected direction. The positive coefficient shows that, over time, acceptance of social media images of one’s own marijuana use is positively associated with knowledge about the contents of the law. In other words, those who approve of social media portrayals of their own use tend to learn more about the law over time than those who don’t want to see themselves using marijuana online. Overall, the research question shows a modest relationship between the interaction of Time x Social Identity on beliefs about the contents of the law that legalized marijuana.

p.45

To further test the extension of the belief gap hypothesis to questions surrounding student beliefs about the legalization of recreational marijuana, two measures of knowledge of the health effects of marijuana were used to test hypotheses relevant to the question of social identity as a predictor of beliefs. The general expectation is that those who are inclined to oppose marijuana use would tend to identify as true those statements that claim negative health effects resulting from marijuana use.

The first hypothesis about health effects was stated as:

H2a: Measures of social identity will be better predictors than educational attainment of negative health effects of marijuana use.

Tables 3.4 and 3.5 show results relevant to the hypothesis. The first indicator is a comparison of the F test of the change in R² for the Educational Aspiration block versus the F test of the change in R² for the Social Identity block (frequency of marijuana use, religious conservative, acceptance of social media portrayals of one’s own marijuana use). In Table 3.4, the hypothesis was supported by a non-significant F (2.6, n.s.) test of the change in the R² in the Educational Aspiration block, and a significant F (22.1, p < .01) test of the change in the R² in the Social Identity block. Similarly, in Table 3.5, the hypothesis is supported by a non-significant F (3.7, n.s.) test of the change in the R² in the Educational Aspiration block, and a significant F (43.7, p < .001) test of the change in the R² in the Social Identity block. The hypothesis was supported.

p.46

The next hypothesis was stated as:

H2b: Identification as a religious conservative will be positively associated with beliefs about the negative health effects of marijuana.

p.47

Table 3.4 shows that the coefficient for religious conservative (β = .05, n.s.) was in the predicted direction, but was not statistically significant. Table 3.5 also shows that the coefficient for religious conservative (β = .05, n.s.) was in the predicted direction, but was not statistically significant. The hypothesis was not supported. Religious conservatives were not more likely than anyone else to correctly identify truthful statements about short-term or long-term negative health effects of marijuana use.

The next hypothesis tests beliefs about negative health effects among those predisposed to support recreational use of marijuana. The hypothesis was stated as:

H2c: Frequency of marijuana use, and approval of social media portrayals of one’s own marijuana use, will be negatively associated with beliefs about the negative health effects of marijuana use.

The test for this hypothesis is shown in the coefficients for the two pro-marijuana social identity variables. In Table 3.4, both frequency of marijuana use (β = -.08, p < .05) and acceptance of social media images (β = -.17, p < .001) were in the hypothesized direction and were statistically significant. In Table 3.5, both frequency of marijuana use (β = -.15, p < .001) and acceptance of social media images (β = -.21, p < .001) were also in the hypothesized direction and were statistically significant. The hypothesis was supported.

p.48

The second research question also attempts to determine whether indicators of social identity strengthen over time in predicting beliefs about the negative health effects of marijuana use. It was stated as:

RQ2: Will there be a relationship between the interaction of time and social identity on beliefs about the negative health effects of marijuana?

This research question is addressed by the interaction terms in Tables 3.4 and 3.5. In neither Table 3.4 or Table 3.5 was the F test measuring the change in R² statistically. This is strong evidence of a lack of a relationship. Similarly, none of the coefficients of the Time x Social Identity measures was statistically significant. No evidence was found of a relationship between social identity and time on beliefs about the negative health effects of marijuana use.

Summary and Conclusions

This paper applied the belief gap hypothesis to an analysis of young adults’ knowledge about recreational marijuana laws and health effects. The knowledge indicators represented two types of knowledge: easily verified facts (about the contents of the law legalizing recreational marijuana use) and facts that require expert opinions (about the short-term and long-term health effects of marijuana use).

Findings showed that, in general, knowledge levels grew during the four years of the study. Knowledge of the law was significantly higher in 2014 and 2015 than in 2012. Similarly, knowledge of both short-term and long-term health effects was significantly higher in 2015 than in 2012. This supports the “cumulative knowledge” proposition of the knowledge gap hypothesis. Evidence from this study supports the proposition that uncontested scientific and public affairs knowledge tends to accumulate over time.

Contrary to the knowledge gap hypothesis, but consistent with the belief gap hypothesis, a measure of educational attainment (aspirational) was not a significant predictor of either types of knowledge: the easily verified contents of the law and the short-term and long-term health effects of marijuana use.

Consistent with the belief gap hypothesis, indicators of social identity were more significant predictors than was educational attainment (aspiration) of beliefs about the contents of the law and about the health effects of marijuana use. Further, hypothesized differences in social group identity were supported. Evidence showed contrasting beliefs about the law and about negative health effects between those with different social identities. The pattern that appears in the findings is consistent with the conclusion that individuals answer survey questions less to show knowledge than to express opinions consistent with their social identities (Bullock et al., 2013, May).

p.49

The present study advances the belief gap hypothesis in four ways. First, the study finds that belief gaps persist in the absence of political contestation of facts. The majority of belief gap studies concern factual issues that are politically contested. In contrast, the issue of legalization of recreational use of marijuana was not politically contested in the state. Hence, individuals could not rely on cues from political elites when answering questions about the law or health effects. Second, the study detected belief gaps among young adults—a demographic group that is less politically polarized than others (Kiley & Dimock, 2014, September 25). Whereas previous studies of adults show belief gaps developing between partisan groups, the present study showed that the belief gap hypothesis can be extended to include social identities. The study is notable for operationalizing three indicators of identities which were shown to predict knowledge, or, more formally, beliefs about knowledge, as was hypothesized. Finally, the study showed that the belief gap hypothesis can be extended to health risk perceptions. Individuals perceived the truth of negative short-term and long-term health effects in ways that were consistent with their behavior (frequency of use of marijuana) and online identities (accepting of online images associating them with marijuana use). This is consistent with previous studies showing group identification as significant predictors of risks relevant to group membership (Campbell & Stewart, 1992; McCright & Dunlap, 2013).

Three main indicators of social identity were chosen: religious conservatives, marijuana users, and individuals willing to be portrayed on social media as accepting of marijuana use.

Special attention was given to conservative religious groups because a) the significant presence of conservative religious groups on college campuses, b) the potential relevance of this identity to legalized recreational marijuana, and c) the intensity of the commitment required of members in those groups. As expected, religious conservatism was negatively associated with knowledge about the easily verified contents of the law that legalized recreational use of marijuana. This is in keeping with the belief gap hypothesis: respondents opposed to the law would claim to know less about its actual contents than would those predisposed in favor of the law. This effect was likely because religious conservatives disagree, on principle, with both the idea of the law but also with specific aspects of the law. In the absence of knowledge, beliefs take precedence.

Another identity expected to be relevant for the study was frequency of marijuana use. In college settings, the subculture of marijuana users would be expected to affect significantly beliefs about the health effects of marijuana use. As expected, frequency of use was a negative predictor of knowledge of the negative health effects of marijuana use. In the absence of knowledge, respondents use questions about negative health effects to express an opinion that serves their self-interest and group identity. Again, polls are used to represent beliefs about facts, not knowledge.

p.50

The third identity considered relevant to legalization of marijuana in this study was those who are proud, happy, and not at all angry, ashamed, or embarrassed when their friends post pictures of them using marijuana on social media. The central role of social media in creating and maintaining college students’ social identities makes this variable of considerable interest. As expected, approval of social media portrayals of one’s own marijuana use were negatively associated with knowledge of the negative health effects of marijuana use. Social identity was a better predictor than educational aspiration of knowledge about the negative health effects of marijuana use.

The findings of the study are relevant to the link between perceptions of risk and health. It is reasonable to expect that individuals who denied the truth of scientifically supported claims about health effects were, in effect, denying the health risk posed by marijuana use. Future work on beliefs about the health effects of marijuana use will include measures of the perceived degree of risk linking recreational use to specific effects, and expert assessments of the same risks. Measures of individual level of identification with various groups will be included with external indicators of group membership. The hypotheses would, however, be the same. Those individuals with identities predisposed to marijuana use will view risks as lower than those who oppose marijuana use. The “belief gap” in which social identities are more predictive than traditional predictors of knowledge, such as educational attainment, would be hypothesized to hold. Beliefs about facts, and beliefs about risk, are both expected to be supportive of social identities and existing behaviors. Perceptions of risk, particularly risks that are associated with meaningful social identities, are expected to be as prone to irrational and counter-factual interpretations as are political views.

In an evolving media environment, identities and media are more closely linked than ever. Audiences of traditional media expressed identity passively as taste publics. Audiences of social media use the platforms for self-presentation and identity creation. Through Facebook Likes researchers can accurately predict a wide range of personal attributes such as sexual orientation, ethnic origin, gender, and intelligence, as well as behaviors that social media users may not know they are revealing such as political views, religion, relationship status, and substance use (Kosinski, Stillwell, & Graepel, 2012).

Similarly, for many college students, online identities are social identities. How students present themselves, or how others present them through pictures taken at social events, is an indicator of how various forms of behavior evolve from stigmatized to accepted. In the present study, questions about how students would feel if images appeared online of themselves using marijuana or in settings in which marijuana was being used, became one of the strongest predictors of knowledge of the law or knowledge of associated health effects. Social identity, as mediated in social media, can be a key indicator in research on health and risk.

p.51

Limitations of the present study are numerous. First, the study relied on three self-selected samples of students seeking extra credit. The results cannot be generalized beyond the individuals participating in the study. The samples were too small and homogeneous to detect what might have been significant differences. The education measure, for example, was based on educational aspiration and not actual achievement. Clearly, results regarding that variable must be interpreted with caution.

Finally, the explained variance in the regression equations was small, particularly regarding knowledge of health effects. This suggests caution in drawing conclusions from the results described in this study. Clearly, there are other processes that drive college student knowledge and beliefs related to the legalization of marijuana.

References

Abramowitz, A. A., & Saunders, K. K. (2008). Is polarization a myth? Journal of Politics, 70, 542–555.

Ammerman, S., Ryan, S., Adelman, W., The Committee on Substance Abuse, & The Committee on Adolescence. (2015, March). The impact of marijuana policies on youth: Clinical, research, and legal update. Pediatrics, 135(3). Available: http://pediatrics.aappublications.org/content/135/3/e769.

Ammons, D. (2012, Jan. 27). Marijuana Initiative 502 certified to Legislature/ballot. Available: http://blogs.sos.wa.gov/FromOurCorner/index.php/2012/01/marijuana-initiative-502-certified-to-legislatureballot/.

Backus, F., & Condon, S. (2012, November 29). Nearly half support legalization of marijuana. CBS NewsNews Polls. www.cbsnews.com/8301-250_162-57556286/poll-nearly-half-support-legalization-of-marijuana/?pageNum=2#postComments.

Bullock, J., Gerber, A., Hill, S., & Huber, G. (2013, May). Partisan bias in factual beliefs about politics. NBER Working Paper Series, 19080. Available: www.nber.org/papers/w19080.

Campbell, L., & Stewart, A. (1992). Effects of group membership on perception of risk for AIDS. Psychological Reports, 70(3, supplement), 1075–1092.

DeWall, C. N., Pond, R. S., Jr., Carter, E. C., McCullough, M. E., Lambert, N. M., Fincham, F. D., & Nezlek, J. B. (2014). Explaining the relationship between religiousness and substance use: Self-control matters. Journal of Personality and Social Psychology,107(2), 339–351. doi:http://dx.doi.org/10.1037/a0036853.

Dominick, J. R. (1999). Who do you think you are? Personal home pages and self-presentation on the world wide web. Journalism & Mass Communication Quarterly, 76(4), 646–658.

p.52

Donohue, G., Tichenor, P., & Olien, C. (1975). Mass media and the knowledge gap: A hypothesis reconsidered. Communication Research, 2(1), 3–23.

Drug Facts. (2017, February). National Institute on Drug Abuse: Marijuana. Available: www.drugabuse.gov/publications/drugfacts/marijuana.

Evangelical Protestants. (2015). Religious landscape study. Pew Research Center for Religion and Public Life. Available: www.pewforum.org/religious-landscape-study/religious-tradition/evangelical-protestant/.

Fiorina, M., Abrams, S., & Pope. (2008). Polarization in the American public: Misconceptions and misreadings. The Journal of Politics, 70, 556–560.

Goldenberg, S. (2015, July 8). Exxon knew of climate change in 1981, email says—but it funded deniers for 27 more years. The Guardian. Available: www.theguardian.com/environment/2015/jul/08/ exxon-climate-change-1981-climate-denier-funding.

Graham, J., Haidt, J., & Nosek, B. (2009). Liberals and conservatives rely on different sets of moral foundations. Journal of Personality and Social Psychology, 96, 1029–1046.

Hindman, D. (2009). Mass media flow and the differential distribution of politically disputed beliefs: The belief gap hypothesis. Journalism and Mass Communication Quarterly, 86, 790–808.

Hindman, D. (2012). Knowledge gaps, belief gaps, and public opinion about health care reform. Journalism and Mass Communication Quarterly, 89, 585–605.

Hindman, D., & Yan, C. (2015). The knowledge gap vs. the belief gap and abstinence-only sex education. Journal of Health Communication: International Perspectives, 20(8), 949–957.

Iyengar, S., Sood, G., & Lelkes, Y. (2012). Affect, not ideology: A social identity perspective on polarization. Public Opinion Quarterly, 76, 405–431.

Jamieson, K. G. H., & Cappella, J. N. (2008). Echo chamber: Rush Limbaugh and the conservative media establishment. Oxford: Oxford University Press.

Johnson, G. (2015, July 4). Washington state has brought in $70 million in tax revenue from legal marijuana sales. Business Insider. Available: www.businessinsider.com/recreational-marijuana-washington-state-tax-revenue-2015-7.

Johnson, E., Bardhi, F., Sifaneck, N., & Dunlap, E. (2006). Marijuana argot as subculture threads: Social constructions by users in New York City. The British Journal of Criminology, 46(1), 46–77 Available: http://ssrn.com/abstract=905679 or http://dx.doi.org/10.1093/bjc/azi053.

Kahan, D., Braman, D., Gastil, J., Slovic, P., & Mertz, C. (2007). Culture and identity-protective cognition: Explaining the white-male effect in risk perception. Journal of Empirical Legal Studies, 4(3), 465–505.

Kiley, J., & Dimock, M. (2014, September 25). The GOP’s millennial problem runs deep. Pew Reseach Center. Available: www.pewresearch.org/fact-tank/2014/09/25/the-gops-millennial-problem-runs-deep/.

Kosinski, M., Stillwell, D., & Graepel, T. (2013). Private traits and attributes are predictable from digital records of human behavior. PNAS: Proceedings of the National Academy of Sciences of the United States of America. Available: www.pnas.org/content/110/15/5802.full.pdf.

p.53

Livingstone, S. (2008). Taking risky opportunities in youthful content creation: Teenagers' use of social networking sites for intimacy, privacy and self-expression. New Media & Society, 10(3), 393–411.

Livingstone, S. (2014). Developing social media literacy: How children learn to interpret risky opportunities on social network sites. Communications: The European Journal of Communication Research, 39(3), 283–303. doi: 10.1515/commun-2014-0113.

Major, J. (2015, July 4). State presents new medical pot rules. The Daily Record. Avalable: www.dailyrecordnews.com/members/state-presents-new-medical-pot-rules/article_e8bd9c18-83e4-11e5-aa99-b737c89a0c35.html.

Martin, J. (2012, November 7). Voters approve I-502 legalizing marijuana. The Seattle Times, http://seattletimes.com/html/localnews/2019621894_elexmarijuana07m.html.

McCright, A. M., & Dunlap, R. E. (2013). Bringing ideology in: The conservative white male effect on worry about environmental problems in the USA. Journal of Risk Research, 16(2), 211–226. doi: 10.1080/13669877.2012.726242.

Meiera, M. Caspia, A., Ambler, A., Harrington, H., Houts, R., Keefe, R., McDonald, K. Ward, A., Poulton, R., & Moffitt, T. (2012). Persistent cannabis users show neuropsychological decline from childhood to midlife. PNAS: Proceedings of the National Academy of Sciences of the United States of America. Available: www.pnas.org/content/109/40/E2657.

Pollan, M. (2002). Botany of desire: A plant’s eye view of the world. New York: Random House.

Political polarization in the American public. (2015, June 12). Pew Research Center. Available: www.people-press.org/2014/06/12/political-polarization-in-the-american-public/.

Proctor, R. (1995). Cancer Wars: How politics shapes what we know & don’t know about cancer. New York: Basic Books.

Preparations in place for July 1 alignment of medical and recreational marijuana systems. (n.d.). Washington State Liquor and Cannabis Board: 2016 Press Releases. Available: http://lcb.wa.gov/pressreleases/alignment-med-and-rec-mj-systems.

Southhall, A. & Healy, J. (2013, August 29). U.S. won’t sue to reverse states’ legalization of marijuana. The New York Times. Available: www.nytimes.com/2013/08/30/us/politics/us-says-it-wont-sue-to-undo-state-marijuana-laws.html?_r=0.

Stewart, C. (2001). The influence of spirituality on substance use of college students. Journal of Drug Education, 31, 343–351.

Tichenor, P., Donohue, G., & Olien, C. (1970). Mass media flow and differential growth in knowledge. Public Opinion Quarterly, 34, 159–170.

Taddicken, M. (2014). The ‘Privacy Paradox’ in the social web: The impact of privacy concerns, individual characteristics, and the perceived social relevance on different forms of self-disclosure. Journal of Computer-Mediated Communication, 19(2), 248–273. doi: 10.1111/jcc4.12052.

Tichenor, P., Donohue, G., & Olien, C. (1980). Community conflict and the press. Beverly Hills, CA: SAGE.

p.54

Veenstra, A., Hossain, M., & Lyons, B. (2014). Partisan media and discussion as enhancers of the belief gap. Mass Communication and Society, 17, 874–897.

Washington Secretary of State. (2012, November 27). Nov. 6, 2012 election results: Initiative Measure No. 502 concerns marijuana. Available: http://vote.wa.gov/results/20121106/Initiative-Measure-No-502-Concerns-marijuana_ByCounty.html.

Willis, H. H., & DeKay, M. L. (2007). The roles of group membership, beliefs, and norms in ecological risk perception. Risk Analysis: An International Journal, 27(5), 1365–1380. doi:10.1111/j.1539-6924.2007.00958.

Zaller, J. (1992). The nature and origins of mass opinion. New York: Cambridge University Press.