BACKGROUND

The COVID-19 pandemic has presented an unprecedented challenge to public health on a global scale. In the United States, the Centers for Disease Control and Prevention (CDC) responded quite early by urging communities to practice social distancing and wear masks in public to slow the migration of the virus across populations. By November of 2020, 37 states and many local governments began instituting mandates requiring masks to be worn in public spaces (Harring, 2020). In a CDC analysis of county-level public health data, states with mask mandates experienced a drop in COVID-19 cases and related deaths while states without a mandate reported increased COVID-related cases and death rates (Guy, 2021). Despite evidence that the policies had a significant impact on virus transmission, some individuals refused to comply with mandates and leading up to the presidential election, the presence of “never-maskers” emerged to highlight how politicized public face coverings had become (Achenbach & Rozsa, 2020). Investigating public intentions to comply with COVID-19 reducing behaviors, the most effective behaviors, were actually ranked the lowest (Lennon et al., 2020). It seems that the politicization of this issue occupies an important role in understanding public responses to the pandemic.

From the beginning, states have played a primary role in policymaking around this issue and there is a body of evidence to indicate the influential power of political ideology in this process. Early state policy responses in the form of stay-at-home orders followed the political ideology of state leadership (Gusmano et al., 2020) and ideological differences influenced compliance with recommendations for social distancing policies impacting mortality rates (Gao & Benjamin, 2021). Partisan influence was even present in early communications regarding emerging treatments for COVID-19 infections (Brunell & Sarah, 2020). There is a large body of work connecting individually-held political ideologies with policy preferences (Converse, 1964; Jacoby & Sniderman, 2006; Zaller, 1992) but the strength with which one holds such ideologies or partisan preferences matter. Culturally oriented beliefs exhibit effects on policy preferences among the public that are distinct from political ideology (Jackson, 2015). The authors intend to contribute further understanding of what drives individual intentions to comply with recommended behaviors meant to slow the transmission of a disease during an emerging pandemic crisis. To do this, we investigate multidimensional beliefs among the US public to understand how various beliefs might inform behavioral intentions toward public mask wearing.

The Health Belief Model (HBM) has been used extensively to predict other health behaviors and is a relevant tool for crafting and evaluating health communications (Carpenter, 2010). During the emergence of the COVID-19 pandemic before communication toolkits were available, HBM was specifically used to guide healthcare worker-patient communications (Carico et al., 2021). The model has also proven useful for measuring the effectiveness of communication campaigns around vaccination behaviors (Jones et al., 2015). While the model has proven practically useful, it suffers from some important limitations, namely in its inability to account well for cultural differences (Glanz et al., 2008; Jones et al., 2015). An emerging area of research in policy studies have begun to employ Cultural Theory (CT) to conceptualize, measure, and predict the cultural influences on health risk management (Tansey & O'riordan, 1999). HBM and CT each identify beliefs as important. This study examines the relationship between HBM's health-centered beliefs and more intrinsic value-based beliefs specified by CT to understand what drives health-oriented behavioral intentions associated with COVID-19. In addition, the study accounts for other theoretically relevant factors such as demographics and political party identification. A brief review of HBM and CT will inform the analysis.

BELIEF CONSTRUCTS UNDER HBM AND CT

The HBM and the extended HBM (Bylund et al., 2011) share the basic assumptions attributed to value expectancy models (Miner, 2005), identifying beliefs as primary predictor of intentions/actions that will be taken to prevent, treat, or screen for serious illnesses (Glanz et al., 2008). Beliefs about the threats, benefits, and barriers associated with COVID transmission are acknowledged to be susceptible to an individual's background and experiences and play an important role in patient communications (Carico et al., 2021). Key constructs within the HBM include perceived susceptibility of an adverse health outcome and the perceived severity of that outcome (Jones et al., 2015). Together, these comprise the perceived threat of an illness. There is evidence that the model performs well under some conditions but, the model's ability to account for social or cultural influences (Glanz et al., 2008) has been identified as a primary limitation of the framework. Other approaches to assessing health-oriented risks within a social context have emerged in recent years.

An emerging theoretical framework in risk and public policy studies conceives of risk as socially constructed and subject to culturally-biased beliefs. Adapted from an anthropological framework, Grid Group CT argues that behavioral patterns are driven by value-based beliefs and concepts of risk. Under this framework, value-based beliefs about what constitutes risk results in four distinct and mutually exclusive behavioral patterns that are reinforced through two primary modes of social control (grid and group). The grid dimension represents the degree to which social relations are prescribed by rules and institutions while the group dimension represents the degree of bounded social relations. As represented in Figure 1, CT's unique contribution applies these two dimensions to specify four typologies representing four distinct worldviews: egalitarianism (low grid, high group), individualism (low grid, low group), hierarchism (high grid, high group), and fatalism (high grid, low group) (Dake, 1992; Douglas & Wildavsky, 1982; Wildavsky & Dake, 1990). Each worldview reflects a particular set of values and perspectives of nature and risk which are shared through social interactions and function to bind groups and define group identities. Both egalitarianism and hierarchism have strong group dimensions and value social relations resulting in strong social networks. Variations in the grid dimension among these two worldviews result in distinct organizational patterns. Egalitarianism (low grid, high group) organizes social networks through values of equality and solidarity while hierarchism (high grid, high group) tends to rely on specialization and hierarchical ordering. Contrastingly, weak group affinity characterizes individualism and fatalism. Low grid tendencies characterized by individualism (low grid, low group) values competitiveness, opportunity, and libertarianism. Finally, fatalism (high grid, low group) perceived lack of social connection combined with subjugation to strong social prescription tends to emphasize situations of isolation and directs attention toward the role of fate. Recent application of this framework reveals that cultural biases like those specified by CT and psychological mechanisms may interact to strengthen the effect of culturally-biased beliefs in ways that work to protect self-identifying characteristics (Kahan, 2008, 2013; Kahan et al., 2007, 2010). CT has been applied across substantive policy domains including health (Kahan et al., 2010; Song, 2014; Song, Geoboo, et al., 2014; Tansey & O'riordan, 1999) and energy policy (Moyer & Geoboo, 2016, 2019; Tumlison & Song, 2019; Tumlison et al., 2017) and promises to offer some insight into understanding what drives health-relevant behavioral responses.

Grid group cultural theory worldview typologies1

To better understand what drives individual-level intentions to wear face coverings in public to decrease the transmission of COVID-19 among the US public, HBM and CT point to multidimensional beliefs as factors of behavioral intentions. It is expected in addition to beliefs about the threat of COVID, more intrinsic CT specified beliefs will guide behavioral intentions to avoid COVID-19 by wearing face coverings in public. More detailed information regarding the data and analytical methods used in this study are followed by analytical results and discussion.

DATA AND METHODS

OLS regression modeling is utilized to explore the relationships between HBM identified beliefs, CT specified beliefs, and the public's behavioral intentions toward mask-wearing using original survey data.

Survey participant selection and data

An original survey was conducted by the nonpartisan research organization PRRI (Public Religion Research Institute) with data collected by the nonpartisan and objective research organization NORC2 at the University of Chicago. The survey sample consists of 2538 randomly selected adults residing in the US across all 50 states plus the District of Columbia. Interviews were conducted both online using a self-administered design and by telephone using live interviewers. All interviews were conducted among participants in AmeriSpeak, a probability-based panel designed to be representative of the national US adult population run by NORC at the University of Chicago. Panel participants without Internet access, which included 42 respondents, were interviewed via telephone by professional interviewers under the direction of NORC. Interviewing was conducted in both Spanish and English between September 9 and September 22, 20203 and focused on a variety of policy issues.

Variables and measures

All measures used in the study are displayed in Table 1. The goal of this study is to estimate the effect of beliefs on the public's intention to avoid COVID-19 transmission by wearing a face covering in public. Therefore, the primary dependent variable was operationalized with the survey question, “When you are in public places, how often do you personally wear a mask to protect against COVID-19 transmission?” (0 = Never, 1 = Sometimes, 2 = Always).

Table 1. Variables and measures Variable Measure Intention to Wear Face Covering When you are in public places, how often do you personally wear a mask to protect against COVID-19 transmission? (0 = Never, 1 = Sometimes, 2 = Always) Egalitarianism Society works best if power is shared equally (1 = Completely disagree to 4 = Completely agree) It is our responsibility to reduce differences in income between the rich and the poor (1 = Completely disagree to 4 = Completely agree) What society needs is a fairness revolution to make the distribution of goods more equal (1 = Completely disagree to 4 = Completely agree) Egalitarianism Index Index using factor score of above three items (α = 0.74) Individualism We are all better off when we compete as individuals (1 = Completely disagree to 4 = Completely agree) Rewards in life should be based on initiative, skill, and hard work, even if that results in inequality (1 = Completely disagree to 4 = Completely agree) Even if some people are at a disadvantage, it is best for society to let people succeed or fail on their own (1 = Completely disagree to 4 = Completely agree) Individualism Index Index using factor score of above three items (α = 0.70) Hierarchism Society is in trouble because people do not obey those in authority (1 = Completely disagree to 4 = Completely agree) The best way to get ahead in life is to work hard and do what you are told to do (1 = Completely disagree to 4 = Completely agree) Society would be much better off if we imposed strict and swift punishment on those who break the rules (1 = Completely disagree to 4 = Completely agree) Hierarchism Index Index using factor score of above three items (α = 0.72) Fatalism The most important things that happen occur by chance (1 = Completely disagree to 4 = Completely agree) The course of our lives is largely determined by forces outside of our control (1 = Completely disagree to 4 = Completely agree) Succeeding in life is mostly a matter of luck (1 = Completely disagree to 4 = Completely agree) Fatalism Index Index using factor score of above three items (α = 0.69) Multidimensional Perceived Threat Financial Threat Dimension You or someone in your family lost a job in 2020 due to the coronavirus pandemic (0 = No; 1 = Yes) You or someone in your family had hours or pay cut in 2020 due to the coronavirus pandemic. (0 = No; 1 = Yes) Personal Threat Dimension You or someone in your household tested positive for COVID-19 in 2020 due to the coronavirus pandemic (0 = No; 1 = Yes) You or someone in your household was sick with COVID-19 symptoms in 2020 due to the coronavirus pandemic (0 = No; 1 = Yes) You or someone in your household has been hospitalized for COVID-19 in 2020 due to the coronavirus pandemic (0 = No; 1 = Yes) Other Threat Dimension You or someone in your household has known someone who has been hospitalized in 2020 due to the coronavirus pandemic (0 = No; 1 = Yes) You or someone in your household has known someone who has died in 2020 due to the coronavirus pandemic (0 = No; 1 = Yes) Perceived Multidimensional Threat Index Index with 0 = No threat and 7 = Extreme threat Party ID Do you consider yourself a Democrat, a Republican, an Independent or none of these? (0 = Independent; 1 = Republican; 2 = Democrat) Age Age in years Gender 0 = Female 1 = Male Race 0 = Nonwhite 1 = White/Non-Hispanic Education 1 = Less than High School to 5 = Postgraduate degree Income 1 = Less than $5000 to 18 = $200,000+

This study estimates the effects of multidimensional beliefs on behavioral intentions. Functioning as the primary variable of interest, this study operationalizes culturally-biased beliefs using measures defined by CT. A panel of 12 questions, three questions corresponding to each of the 4 grid-group typologies, were randomly ordered in the survey. Respondents were asked to indicate their level of agreement with these statements on a 4-point scale (1 = completely disagree, 2 = mostly disagree, 3 = mostly agree, 4 = completely agree). Factor analysis (with the varimax rotation method) was performed to extract four latent factors representing the four CT dimensions. These factors parallel the orthogonally distinct typologies in that consistently high factor loadings aligned with each of the three related CT measures (i.e., factor loading greater than 0.5) while loadings were low on remaining unrelated factors. Based upon this factor structure, factor scores were calculated for each of four latent dimensions (representing each of four cultural orientations with scores ranging from −3.08 to 3.36) and are used as an index for measuring each cultural orientation. Cronbach's α scores for the three survey items (constituting each CT index) range from 0.69 to 0.74 indicating that the related survey measures are reasonably reliable.

The analysis incorporates other multidimensional beliefs found to be relevant to health behaviors. HBM defines perceived threat as a combination of perceived severity and perceived susceptibility. Perceived threat is operationalized in this study using a multi-dimensional scale that captures the perceived personal and financial threat as well as perceived threat to others known by someone in their household. Respondents were asked to indicate any and all of the following experiences that they or someone in their household encountered in 2020 due to the coronavirus pandemic reasoning that the more personal experiences an individual has, the more likely they are to perceive the experience as a threat. The full measures of multidimensional threat variable are shown in Table 1. Respondents were asked to indicate whether they attributed a recent loss of job or wages (financial threat), a recent illness or hospitalization of someone in their household (personal threat), or someone known to their household (other threat) to COVID-19 (yes = 1 or no = 0). An additive index was then calculated to measure perceived threat where 0 indicates that the respondent answered no to all indicating no threat and 7 indicates that respondent answered yes to all questions indicating extreme perceived threat of COVID-19.

Measures for demographic factors and political party identification are also operationalized to statistically control for these factors in an attempt to isolate the effects of beliefs on behavioral intentions. Age was recorded as a continuous variable. Respondents indicated their gender as 0 for female and 1 for male. Race was recorded as 0 for non-White and 1 as White/non-Hispanic. Respondents' education was recorded with a scale ranging from 1 (less than high school) to 5 (postgraduate degree) and income was recorded with a scale ranging from 1 (indicating an annual household income of less than $5000) to 18 (indicating a household income of $200,000 or more).

Methods

First, a comprehensive estimation of the effects of perceived threat associated with COVID-19 and more intrinsic beliefs on behavioral intentions are estimated using OLS regression (n = 2395). Further investigation into the conceptual moderation effect of multidimensional threat variable on the relationship between culturally biased beliefs and behavioral intentions is accomplished by dividing the data set into two subsets based on the threat level, fitting the said OLS regression model to these two subsets of data, and comparing main effects (i.e., the effects of culturally-biased beliefs on behavioral intention) in the estimated models. One subset represents members of the general public who indicated some perceived threat associated with COVID in at least one but perhaps more dimensions (n = 1723) while the other subset represents members of the public who indicated no perceived threat (n = 672) at all.

ANALYTICAL RESULTS Sequential OLS regression

The data used in this step of the analysis consists of 2395 observations. The typical survey respondent may be characterized as a 49-year-old White/non-Hispanic female who reports an annual household income of between $40,000-60,000 and some education beyond high school (see Tables 2 and 3).

Table 2. Descriptive statistics for full sample Statistic n Mean St. Dev. Min Pctl(25) Pctl(75) Max Intention 2389 1.750 0.496 0.000 2.000 2.000 2.000 Egalitarianism 2395 0.000 1.000 −3.047 −0.632 0.636 2.862 Individualism 2395 0.000 1.000 −3.426 −0.648 0.601 2.862 Hierarchism 2395 0.000 1.000 −3.174 −0.670 0.652 3.100 Fatalism 2395 0.000 1.000 −2.440 −0.540 0.671 3.377 Threat 2395 1.321 1.232 0 0 2 7 Age 2395 48.603 16.972 18 33 63 97 Education 2395 3.240 1.042 1 3 4 5 Income 2395 9.986 4.189 1 7 13 18 Table 3. Frequency table Sample Variable n Category (%) Full Race 2395 Non-White (36.2%) White/Non-Hispanic (63.8%) No threat subsample 672 (21.9%) (78.1%) Threat subsample 1723 (44.2%) (55.8%) Full Gender 2395 Female (50.9%) Male (49.1%) No threat subsample 672 (53.7%) (46.3%) Threat subsample 1723 (47.4%) (52.6%) Full Political party identification 2395 Independent (30.6%) Republican (30%) Democrat (39.4%) No threat subsample 672 (34.4%) (29.8%) (35.8%) Threat subsample 1723 (27.0%) (31.8%) (41.2%)

OLS regression estimates the effects of demographics on behavioral intentions (Model 1) before adding political party identification (Model 2), perceived threats (Model 3), and culturally-biased beliefs (Model 4) to the regression model (see Table 4). The base model (Model 1) results indicate that respondents who are older (+0.004, or an increase of 0.2% on the dependent variable scale for each additional year of age, p < 0.05) and more educated (+0.041, or an increase of 2.05% on the dependent variable scale for each additional year of age, p < 0.05) report stronger intentions to wear a mask in public. White males are the least likely to report public mask-wearing (−0.097, or a 4.85% reduction on the dependent variable scale for non-Hispanic whites, p < 0.05 and −0.061, or a 3.05% reduction on the dependent variable scale for males, p < 0.05 respectively). Controlling for demographics, those who identify as Republican are less likely (Model 2, −0.174, or a decrease of 8.7% on the dependent variable scale for Republicans, p < 0.05) to wear masks in public while Democrats are more likely to do so (+0.133, or an increase of 6.65% on the dependent variable scale for Democrats, p < 0.05). Model 3 incorporates perceived threat of COVID transmission into the regression showing that the higher the perceived threat, the more likely one is to wear a mask in public (+0.028, or an increase of 1.4% on the dependent variable scale for each unit increase in threat perception, p < 0.05). Results of the final model (Model 4) indicate a positive relationship between egalitarian beliefs and public mask-wearing (+0.094, or an increase of 4.7% on the dependent variable scale for each unit increase in egalitarianism, p < 0.05).

Table 4. OLS regression analysis results Regression results Dependent variable: Intention to wear face covering in public (1) (2) (3) (4) Egalitarianism 0.094*** (0.011) Individualism −0.060*** (0.010) Hierarchism −0.027*** (0.010) Fatalism 0.008 (0.010) Threat 0.028*** 0.025*** (0.008) (0.008) Republican −0.174*** −0.172*** −0.104*** (0.026) (0.026) (0.026) Democrat 0.133*** 0.132*** 0.063*** (0.024) (0.024) (0.024) Race (white) −0.097*** −0.047** −0.034 −0.032 (0.022) (0.022) (0.022) (0.022) Gender (male) −0.061*** −0.045** −0.042** −0.036* (0.020) (0.020) (0.020) (0.020) Age 0.004*** 0.004*** 0.004*** 0.004*** (0.001) (0.001) (0.001) (0.001) Educat