Varieties of Mistrust and American Epistemic Fragility

Introduction

In 2020, even as fatalities from COVID-19 reached 3,000 per day, millions of Americans defied public health recommendations for social distancing and mask-wearing. The following year, in the absence of any evidence that the presidential election had been stolen, violent insurrectionists at the United States Capitol attempted to block the constitutional transfer of power. These behaviors highlight the consequences of deep mistrust: noncompliance with public health guidelines contributed to the nation’s 1.1 million pandemic-related deaths, and the refusal to accept the presidential election results threatens American democracy; one third of Americans and two-thirds of Republicans still dispute the 2020 outcome (Monmouth University 2023). But how does distrust in government differ from defiance of pandemic public health advice? And are the drivers of such distrust similar?

We answer these questions by distinguishing between political trust and trust in “scientific authority” (Gauchat 2012) as different conceptual dimensions of “vertical trust,” and statistically identifying sources of mistrust across 3,000 American counties. Drawing upon research that blames mistrust for a “decline in reason” (Davies 2019), the rise of “post-truth” politics (Kakutani, 2018; McIntyre 2018), and the problem of “truth decay” (Kavanagh and Rich 2018), we argue these overlapping types of mistrust constitute a symptom of “epistemic fragility,” a state of receptivity to misinformation. Empirical tests find robust support for research linking partisanship to mistrust in government (Pew 2020; Zuckerman 2021). Such partisanship also strongly correlates with several proxies for mistrust in scientific authority, which research associates with health behaviors and attitudes about COVID-19 (Devine, Gaskill, Jennings et al. 2021; Gardarian et al., 2021). But political beliefs are only one source of mistrust. We identify a range of community characteristics that shape the social costs for acting upon mistrusting attitudes. In particular, we offer evidence that counties facing more mental health distress have higher levels of mistrust. This mistrust sometimes bridges partisan divides, suggesting that while mistrust indeed differs based on ideological beliefs (Goldstein and Wiedemann 2021; Kerr, Panagopoulos, and van der Linden 2021), relevant remedies may transcend politics. Our dataset and supplemental materials are available at https://prq.sagepub.com.

In what follows, a literature review summarizes prominent explanations for the decline in trust over the last five decades and common conceptual distinctions between types of trust. By trust, we have in mind relationships where those delegating authority see little need to monitor behavior (Citrin and Stoker 2018; Norris 2022). Political trust thus entails a set of expectations reasonably imposed on a government or another actor that does not require a personal relationship (Bauer and Freitag 2018). We treat political trust and trust in scientific authority as two dimensions of “vertical trust” (as opposed to “horizontal” trust), our dependent variable capturing the latitude that principals grant to agents. Next, we outline our empirical strategy. We proxy for political trust with original variables measuring tax compliance and U.S. Census self-response rates, operationalizations well-established in the literature (Chan, Supriyadi, and Torgler 2018). We proxy for trust in scientific authority with variables measuring rates of vaccination, vaccine hesitancy, and mask wearing. These operationalizations draw upon literature that uses vaccination as a proxy for “trust in science” (Goldstein and Wiedemann 2021) and that links trust to public health behavior (Goldberg et al. 2020; Latkin et al. 2021).

We then formulate and test two hypotheses. Our partisan mistrust hypothesis explores whether vertical mistrust is rooted in political preferences, visible in voting behavior. Using voting outcomes from the 2020 presidential election, statistical tests demonstrate that Republicans are more likely to mistrust both government and scientific authority. Next, our community distress hypothesis predicts that counties suffering from increased psychological distress will have lower levels of vertical trust. We ground this theoretical expectation in research linking distress and anxiety to a desire for predictability that impedes epistemic judgment (Farias and Pilati 2021; Kweon and Choi 2023). This quality distinguishes mistrust in scientific authority from political mistrust, which by itself can signal healthy democratic skepticism (Norris 2022; Warren 2018). Using a Centers for Disease Control (CDC) variable measuring the number of days in the last month that individuals reported poor mental health, tests link psychological distress to both types of vertical mistrust. Our empirical strategy captures the pre-pandemic depth of mental health challenges, while acknowledging how Coronavirus triggered significant increases in the problem (Thormeer et al. 2023). In sum, we find that partisanship and psychological distress systematically shape two types of vertical trust across 3,000 counties, using multiple operationalizations of our dependent variable. The results hold when controlling for socioeconomic conditions, ethnic and racial identities, religion, rural/urban demography, and COVID-19 infections and deaths. The evidence also withstands a broad range of robustness checks, including models interacting the mental health variable with partisanship, statistical specifications with state-level fixed effects, and an alternative measure of partisanship based on party registration.

Mistrust in government alone could merely reflect subjective evaluations of government performance, while views of science might simply measure scientific knowledge or literacy. By operationalizing vertical trust in different ways, one of our contributions is to bolster recent studies empirically justifying longstanding theoretical justifications for doing so (Dawson and Krakoff 2023). This approach also adds to literature on behavioral expressions of mistrust (Devine 2022). Second, our control variables constitute valuable contributions in themselves: high rates of vertical trust among Asian Americans contrast with low trust among Black Americans and other minorities, while tests with a Latinx variable confirm studies documenting unequal access to health care. Third, we find that evangelical Christianity rather than religion overall tends to weaken trust in scientific authority. Fourth, we show how America’s mental health crisis shapes mistrust. Unlike other studies that rely on surveys, lab experiments, or internet polls, our analysis enables identification of communities that share cognitive and contextual features of distress. Relatedly, our findings on partisanship offer evidence about how ideology conditions psychological distress: both conservative and liberal counties suffer from high rates of distress, but this manifests differently in Democrats (who fear Coronavirus) and Republicans (who fear loss of liberty). Our empirical approach does not specify the mechanism by which mental health impacts trust or vulnerability to misinformation—a topic explored by others (see Lewandowsky et al., 2013). Yet a core implication of this study nevertheless stands out: socioeconomic vulnerabilities and shared cognitive factors constitute important community contexts that weaken epistemic judgment.

Literature Review: Mistrust and its Epistemic Effects

The United States has witnessed a stark decline of trust in government since the 1950s, with brief recoveries during the 1980s and again after 9/11 (Kavanaugh and Rich 2018; Zuckerman 2021). Multiple datasets and surveys, including the American National Election Survey, the General Social Survey, Pew Research, and the World Values Survey all arrive at similar conclusions (Citrin and Stoker 2018). This decline has more recently been accompanied by doubts about various aspects of science (McIntyre 2018; Oreskes 2019). Such claims are not without nuance: for example, people trust local government more than the “government in Washington,” and low levels of confidence in the Supreme Court, career civil servants, and police are quite recent (Gallup 2023; Pew 2022). Similarly, trust in “scientists” and “medical scientists” has generally been much higher compared to other institutions. But public trust in these groups declined during the pandemic, the drop has been much steeper among Republicans compared to Democrats, and positive perceptions of scientific contributions have shrunk (Kennedy and Tyson 2023). In short, the evidence for mistrust in the United States is compelling.

In what follows, we draw upon democratic theory to organize research on these two broad trends in trust under the umbrella concept of “vertical trust.” Next, we describe some of the well-known drivers of mistrust such as populism and anti-intellectualism, which have inspired widespread skepticism of expertise and hostility towards government. The “post-truth” literature associates these phenomena with fake news, misinformation, and the proliferation of false conspiracies. Partisanship appears as a recurring variable in this research, framing trust relationships by shaping epistemologies and socially constructed truths. Partisanship has also long been a focus of political psychology research on conspiracy beliefs, and we discuss how the field acquired new urgency as COVID-19 became increasingly polarized.

Research on “horizontal” or social trust explores relationships and norms of reciprocity among people that foster cooperation. Dense interpersonal relationships, especially when they transcend social cleavages and are “general” rather than “particular,” are associated with a variety of prosocial outcomes (Uslaner 2018). We focus on “vertical” trust, which entails the delegation of authority by principals to institutions or people acting on their behalf as agents. A trustworthy relationship requires little oversight because agents demonstrate competence and impartiality, separating personal preferences from duties even under conditions of uncertainty (Goldstein and Wiedemann 2021; Norris 2022). Government institutions subject to partisan control, such as Congress, usually suffer from trust deficits compared to police, courts, or public health officials whose authority does not derive from electoral processes (Warren 2018). Corresponding declines in these different types of trust appear to support the “psychological propensity model,” which posits that people are either broadly trusting or not, regardless of whether the relationship involves government, family, or strangers (Newton et al. 2018). But few studies empirically explore different types of trust side by side (Dawson and Krakoff 2023).

Existing research attributes the overall decline in trust to a variety of factors. The collapse of traditional media leaves readers facing huge volumes of information and news outlets that no longer perform a “gatekeeper” role, filtering for reliability and accuracy (Kavanaugh and Rich 2018; McIntyre 2018). Other common explanations point to political polarization and online “echo chambers” that reinforce cognitive biases (Kavanaugh and Rich 2018; Kennedy and Tyson 2023; Pew 2020; Sunstein 2017). Populism also features prominently in the literature on the decline of trust (Citrin and Stoker 2018). Populists undermine trust by embracing a rhetorical style that challenges the legitimacy of government (Norris and Inglehart 2019), and they circumvent political institutions as a strategy to claim a direct line to “the people” as their sole authentic voice. Some research finds that populism generates a general distrust towards institutions of power, including science or scientists (see Rosenfeld 2020; Zapp 2022). In this regard, populism is closely related to both a general critique of expertise (Davies 2019; Norris and Inglehart 2019) and a resurgent anti-intellectualism, a negative affect towards elites or an “intellectual establishment” that supposedly suffers from liberal bias (Barker et al. 2022; Citrin and Stoker 2018).

Donald Trump’s populism marries conventional conservative mistrust of government with a denial of objectivity and technical neutrality (Chowkwanyun 2022). This stokes mistrust in “scientific authority,” meaning trust in expertise that serves as “cultural authority” for establishing political consensus (Gauchat 2012). Trump demeaned civil servants and sowed mistrust in the civil service and the “administrative state” (Gage 2022). Attacks on health officials during the pandemic exacerbated widening differences between Democrats' and Republicans' levels of trust in science (Kennedy and Tyson 2023). The gap had emerged in the 1980s as conservatives embraced skepticism of organized science due to its implications for government regulation. By the 2000s, this skepticism entailed rejection of climate science, marking an about-face from the 1970s when conservatives trusted science more than liberals (Gauchat 2012; Lewandowsky et al., 2013). Some experimental evidence supports this, linking ideological mistrust in science to mistrust in expertise (Parkhurst 2016).

The pandemic, along with Trump’s oversized role in spreading disinformation about it (Evangea et al. 2021), animated debates over the impact of partisanship on trust. Conservatives had long mistrusted government more than liberals (Citrin and Stoker 2018; Pew 2022) as part of a broader ideological critique of regulation (Zuckerman 2021). Mann and Ornstein (2012) write that by the early 2000s, the Republican Party was “unpersuaded by conventional understanding of facts, evidence and science” and was “all but declaring war on the government.” As COVID-19 spread, large partisan differences persisted, with Republicans resisting infection mitigation measures (Devine, Gaskill, Jennings et al. 2021; Gardarian et al., 2021) including social distancing (Goldstein and Wiedemann 2021) and mask wearing (Farias and Pilati 2021). Counties with Republican majorities also demonstrated higher rates of COVID-19 infections and less receptiveness to vaccines (Kerr et al. 2021; Rosenfield 2020). Mistrust in government led to fewer protective behaviors such as mask wearing because conservatives perceived the virus as less threatening (Kerr et al. 2021). Notably, this mistrust has lingered well past the end of the pandemic (Abutaleb, Roubein, and Arnsdorf 2023).

Scholars have long maintained that such mistrust in government is conducive to conspiracy theorizing. Hofstader (1964) described it as a “paranoid style” shaping attitudes about both government and science. Al Gore, in an under-appreciated book, associates pervasive anxiety with mistrust: the political weaponization of fear impedes epistemic judgment, amounting to an “attack on reason” (Gore 2007). A variety of research supports this connection, linking anxiety with “paralysis of cognition” and a desire for predictability (Birrell et al. 2011). Such psychological distress contributes to skewed threat perception (Jost, Sterling, and Stern 2017) and an increased willingness to believe false conspiracies (Farias and Pilati 2021).

Trump mainstreamed bizarre conspiracy theories (Muirhead and Rosenblum 2020), including the effort to overturn the 2020 presidential election (House Select Committee 2022). Stecula and Pickup (2021) maintain that Trump-style populism, with its distrust of scientists, government, elites and experts, has an even stronger effect than ideology on COVID-19 behaviors and conspiracy beliefs. Comparative research across 22 countries arrives at a similar conclusion (Kweon and Choi 2023). Where the causal model is reversed, belief in false conspiracies—even on topics unrelated to government—contributes to general feelings of mistrust (Albarracín, Albarracín, and Sally Chan 2022).

Yet psychology researchers widely debate the effects of political attitudes on mistrust and misinformation. Supporters of the “asymmetry hypothesis” argue that partisan identity filters fear, influencing individuals' receptiveness to misinformation. Jost et al. (2017) find that conservatives are more sensitive to threat than liberals, meaning their respective responses to threatening stimuli differ. This line of thinking suggests that conservatives resisted public health measures because the pandemic represented a threat to the status quo, confirming fears generated by uncertainty. By contrast, liberals responded to the same uncertainty and anxiety by trusting public health authorities and embracing risk mitigation behaviors including mask wearing (Kerr et al. 2021). Researchers doubtful of the asymmetry hypothesis claim climate change has been overused to study the relationship between partisanship and science, resulting in findings that do not represent overall attitudes about science; they also argue that context shapes psychological predilections for conspiracy (Enders, Farhart, and Miller 2023). Yet another school of thought claims that strong partisan identity, on both the left and the right, corresponds with “reduced cognitive flexibility,” understood as the ability to adapt to environments and a capacity to change modes of thinking. This inflexibility is associated with prejudice, ideological dogmatism, and willingness to believe false conspiracy theories (Zmigrod Jason Rentfrow, and Robbins, 2020). However, a major study rejects this notion of bipartisan extremes, finding that conservatives are far more likely to endorse conspiracy theories (van der Linden et al. 2021).

Much of the literature exploring why ideology conditions threat perception and risk mitigation behavior centers on uncertainty and fear. Magnetic Resonance Imaging (MRI) scans, for example, show that conservatives have increased activity in the parts of the brain closely associated with fear (Kanai et al. 2011). This cognitive reflex restrains the brain’s ability to reason, impairing the ability to make epistemic judgments pertaining to politics (Rule et al. 2010). Other (less deterministic) research examines how belief in conspiracy theories or misinformation constitutes a coping strategy in response to fear generated by ambiguous or stressful situations (Buhr and Dugas 2009) such as the pandemic (Maftei and Holman 2022). The associated feelings of uncertainty overlap with a broad range of mental health diagnoses (Mahoney and McEvoy 2012). Feelings of uncertainty also predict different types of psychological distress (Mertens et al. 2020; Rettie and Daniels 2021), which contribute to an erosion of trust (Simonetti et al. 2021).

This study seeks to advance literature on the epistemic effects of mistrust by testing a version of the asymmetry hypothesis. Our empirical strategy captures both social context and cognitive traits that play a role in trust relationships. By studying communities cross-sectionally, we aim to explore how “truth decay” relates to demography, ideology, and the particularities of place. Moreover, by including observational data, we consider behavioral manifestations of trust, rather than conceptualizing it solely in attitudinal terms. This seems important, since defiance of public health recommendations during the pandemic and the violent denial of presidential election results in 2020 both highlight the grave consequences of attitudes shaped by mistrust. Third, by exploring political trust alongside trust in scientific authority, we test the “psychological propensity model” (Newton et al. 2018). Political mistrust by itself reflects Western democracy’s Enlightenment roots and can theoretically inspire citizen participation. But mistrust in scientific authority conveys information about how people assess empirical claims. In this spirit, we claim that the convergence of these two types of mistrust fosters “epistemic fragility,” understood as community vulnerability to misinformation and falsehoods.

Empirical Strategy

We disaggregate our dependent variable, vertical trust, into two dimensions: political trust and trust in scientific authority. Both use variables representative at the county level. Political trust, which we understand here as a general “trust in government,” is operationalized with three proxy variables based on observational data. First, we examine Census self-response rates, which the Census Bureau itself recognizes as a powerful proxy for trust (Singer, Bates, and Hoewyk 2011). A recent Harvard study of education performance during the pandemic used response rates to measure “trust in government institutions,” for example (Fahle et al. 2023). Other research blames poor Census response rates on mistrust in government (Hogan 2020; O’Hare and Lee 2021). By definition, this variable captures the voluntary sharing of potentially sensitive or private personal information and belief that information will be handled appropriately. In our statistical models, Census2020 represents the overall response rate, while Census1020 measures response rate change from 2010 to 2020. Drawing on both variables is useful since the 2020 Census was administered during the pandemic, and we aim to capture trust beyond a single moment in time.

We also use tax compliance—another common indicator of vertical trust—to proxy for political trust in government (Chan, Supriyadi, and Torgler 2018). Tax compliance is conceptually related to trust because a willingness to pay taxes implies some confidence that the political system works as intended. Studies of the US (Jimenez and Iyer 2016) and broad cross-national samples (Batrancea et al. 2019) point to a strong connection between trust and tax compliance. Conditions such as civic mindedness, religiosity, or perceived fairness of taxation may influence compliance rooted in vertical trust (Scholz and Pinney 1995; Torgler and Schneider 2007).

Alternatively, some research argues that fear works alongside trust as a determinant of tax compliance (Batrancea et al. 2019). This distinction informs the operationalization of the TaxCR variable, drawing on Hwang and Nagac (2021). We calculate the tax compliance ratio by comparing county-level average household income reported to the IRS in 2018 with average household income reported to the U.S. Census. Discrepancies between the two reflect not only intentionally hiding income from the IRS but also a lack of fear of being caught: if deterrence were the core driver of compliance, cheaters would not risk reporting different numbers to two government agencies. ¹ Instead, they would simply underreport to both. While deterrence may still play some role, we believe that the tax compliance ratio used here is a strong proxy for trust in government. With TaxCR, a high rate of compliance refers to taxpayers accurately “declaring their entire income” and “paying all taxes by fulfilling their legal obligations” (Aktaş Güzel, Özer, and Özcan 2019, 81).

Our other dependent variables measure trust in scientific authority, similar to a “cultural authority” of science rather than knowledge of science or trust in scientists (Gauchat 2012). Americans generally support scientific research and demonstrate levels of scientific literacy comparable to other advanced democracies. However, understanding of the process of scientific inquiry is weak, and only four in ten Americans express confidence in scientific leadership (Hallman 2017). Measuring trust in scientific authority is especially relevant during the specific period of the pandemic we study, which corresponded with significant uncertainty and therefore a need to trust public health guidance (Latkin et al. 2021).

We operationalize trust in scientific authority in four ways. First, COVIDhesitancy calculates the observed difference between COVID-19 vaccine availability and the actual rate of vaccination. Research empirically links vaccine hesitancy with belief in the QAnon conspiracy (Public Religion Research Institute 2021b), suggesting that as a proxy for trust it is indeed associated with epistemic judgments. Next, COVIDvaccination measures the actual vaccination rate. The distinction from hesitancy is important since minorities and some communities might be willing to get vaccinated but face barriers to doing so. We use July 2021 data because it captures a critical moment in the pandemic when vaccines were widely available, but many vaccine mandates—public or private—did not come into effect until the fall of 2021. Our variable therefore captures those who trusted scientific authorities when vaccines were readily accessible but not necessarily mandated. We also include FluVaccine, taken from the CDC’s Behavioral Risk Factor Surveillance System (BRFSS) data for 2020. Comparing this with COVID-19 variables provides an important check on our theory since the common flu is less politicized and thus helps us test whether mistrust in scientific authority is indeed recent.

Finally, we consider face masks, whose connection with trust in science is well-established (Stosic, Helwig, and Ruben 2021). Mask opponents explicitly challenge scientific authority with claims that masks can cause carbon dioxide poisoning (Reuters Fact Check 2021), spread other illnesses (Groth 2021), have little public health value (Mollenauer 2021), or are unnecessary because religious faith offers protection (St. George 2021). Conversely, pro-mask advocates wear face masks because they trust scientific advice (Mollenauer 2021; Dolan 2021). Protesters calling for mask requirements in Texas, for example, carried signs that said, “Trust Science, Mask Up” (Ramirez 2021). The variable Masks measures the rate of mask usage, drawing upon a nationwide, county-level survey from July 2020. Again, our choice of timing is deliberate: because no vaccines or COVID-specific antivirals were available yet, masks and social distancing were critical defenses against infection. Although the CDC did not recommend face masks until early April 2020—several weeks into the first wave of the pandemic—masks were widely recommended by public health officials by that summer (Crump and Placheril 2020).

Tests and Analysis

The regression models use ordinary least squares (OLS) with different specifications and operationalizations of vertical trust; to correct for potential heteroskedasticity, all models obtain robust estimators using the STATA VCE [robust] option. The results offer strong support for both hypotheses. Democratic voters have higher levels of trust in government and in scientific authority, supporting our partisan mistrust hypothesis and research that finds asymmetrical effects of partisanship on trust (Gauchat 2012; Jost et al. 2017). We also find support for our community distress hypothesis: communities with more reported days with poor mental health have less political trust and less trust in scientific authority. The evidence of psychological distress across party lines suggests that despite other partisan asymmetries, epistemic fragility has roots in public health problems. Table 1 offers a baseline regression, where the coefficients on independent variables are statistically significant and in the expected direction for all the models with the exception of Mask regressed on MentalHealth (model 7). Next, we introduce our control variables, run new estimates with them, and describe several robustness checks.

OPEN IN VIEWER

Table 1.

Baseline Regression Without Control Variables.

	Political Trust			Trust in Scientific Authority
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Census20	Census 1020	Tax CR	COVID Hesitancy	COVID Vaccine	Flu Vaccine	Mask
VoteDem, %	34.42*** (20.86)	31.04*** (28.84)	0.48*** (9.61)	−0.09*** (−22.71)	0.22*** (13.80)	0.16*** (13.80)	0.53*** (35.81)
MentalHealth	−8.73*** (−21.97)	−4.05*** (−13.86)	−0.13*** (−14.72)	0.02*** (19.53)	−0.05*** (−16.47)	−0.02*** (−5.21)	0.02*** (5.66)
Constant	62.59*** (34.59)	−16.51*** (−12.39)	1.22*** (33.77)	0.07*** (12.72)	0.46*** (34.88)	0.42*** (30.32)	0.25*** (14.48)
Observations	3112	3091	2830	3112	2836	3109	3112
Adjusted R2	0.24	0.25	0.13	0.25	0.15	0.08	0.32

Controls for Confounding Factors

Numerous social, economic, and political contexts could interfere with the hypothesized relationships. First, we control for religious affiliation. Dozens of lawsuits challenged mask and vaccine mandates based on religious freedom. Some of the most vocal opposition came from evangelical Baptist preachers (Dias and Graham 2021; Hoffman 2021), many of whom wrote vaccine exemption letters (Graham 2021). While Gauchaut (2012) found that church attendance overall reduces public confidence in science, we believe religious faith by itself predicts neither mistrust in government nor science. Indeed, our analysis of public statements by every major religious group recognized by the federal government found no doctrinal basis for vaccine opposition. The variable evangelical controls for religion with data from the 2010 U.S. Religion Census, which measures Christian evangelical adherence rates. Using this particular faith as a reference point is useful since white evangelicals had the highest COVID vaccine “refusal rate” at the time (Public Religion Research Institute 2021a), and they contributed to the proliferation of false conspiracies (Survey Center on American Life 2021).

We control for race and ethnicity with several different variables. This offers important advantages over approaches that treat race as binary (Gadarian et al. 2021) or that reduce diversity to the proportion of white residents (Norris and Inglehart 2019). For example, Asian Americans feared being blamed for the pandemic. They endured a nearly 150 percent increase in hate crimes in 2020 (Center for the Study of Hate and Extremism 2021). Racism is also a critical reference point for Black Americans, and many mistrust vaccines because of the Tuskegee experiments during the 1950s (Green et al. 1997). Vaccine hesitancy appears to be declining among Black Americans compared to other groups, including Latinx (Doherty et al. 2021). Other research suggests that health care inequalities, rather than overall mistrust in science, limits the personalized access to scientific expertise conducive to trust. Within Latinx communities, for example, this sometimes means countering myths about identification requirements or costs (Kornfield 2021). Our variables, expressed as a percent of the county’s population in 2020 and using U.S. Census categories, capture these different experiences: Hispanic for Hispanic or Latinx; AIAN for American Indian or Alaska Native; Asian for East, Southeast, and South Asian; Black for Black and African American; and NHOPI for Native Hawaiians and Other Pacific Islander. Such variables are of particular interest with our community distress hypothesis: the pandemic amplified mental health challenges affecting minority groups, who suffer unequal access to health care as well as anxiety triggered by fear of being blamed for spreading COVID-19 (McKnight-Eily et al. 2021; Thomeer, Moody, and Yahirun 2023). Finally, COVID-19cases measures the share of the county population testing positive, while COVID-19deaths measures fatalities per 100,000 people as of July 2021. An increase in either one could theoretically inspire people to get vaccinated. There are several reasons to control for urban demography (as of 2019) with our variable, Rural. The virus’s initial wave disproportionately impacted densely populated (and heavily Democratic) cities. But soon thereafter, individuals who voted for Trump in 2016 were up to 30 percent more likely to die from COVID, an outcome driven by age and rural residency (Pueyo 2022). More generally, research identifies widespread resentment in rural areas due to perceived federal neglect which undermines trust (Lister and Jourdey 2023). This mistrust correlates closely with the anti-intellectual streak mentioned earlier (Trujillo 2022), reinforcing our view that Rural is relevant for analyzing the vertical relationships of interest here. Moreover, few studies of rural vaccine hesitancy include partisanship (Callaghan 2023). Supplemental materials included as appendices provide additional details about data sources and collection dates (Table A1), descriptive statistics (Table A2), and bivariate correlation statistics (Table A3).

Table 2 offers strong support for our partisan mistrust hypothesis by demonstrating a statistically significant correlation between VoteDem and all seven measures of vertical trust. Model 8 shows that each one percent increase in Democratic voters leads to a 19.25 percent increase in political trust as measured by Census self-response rates (R² = 0.63). That figure rises to 23.81 percent in model 9, measuring change in response rates over time from 2010 to 2020 (R² = 0.42). Model 10 indicates that Democratic voters also have higher rates of tax compliance. Models 11–14 show much higher rates of trust in scientific authority: a one percent increase a county’s share of Democratic voters is associated with a 0.11 percent decrease in vaccine hesitancy (R² = 0.39), a 0.47 percent increase in COVID-19 vaccine uptake (R² = 0.34), and a 0.41 percent increase in mask adherence (R² = 0.47). The results with the flu vaccine variable suggest that the mistrust in science is not merely driven by politicization of the Coronavirus pandemic, COVID-19 vaccines, or mask wearing: in model 13, a one percent increase in Democratic vote share led to a 0.23 percent increase in flu vaccination rates. This implies that even before polarized reactions to the COVID-19 pandemic, predominantly Republican counties had statistically significant lower levels of trust in scientific authority. Compared to our multivariate regressions and correlation matrices (included as Table A3 in supplemental materials), these relationships remain highly stable.

OPEN IN VIEWER

Table 2.

Partisanship (VoteDem) and Vertical Trust.

	Political Trust			Trust in Scientific Authority
	(8)	(9)	(10)	(11)	(12)	(13)	(14)
	Census20	Census 1020	Tax CR	COVID Hesitancy	COVID Vaccine	Flu Vaccine	Mask
VoteDem, %	19.25*** (9.43)	23.81*** (14.21)	0.36*** (4.99)	−0.11*** (−15.78)	0.47*** (17.27)	0.23*** (13.77)	0.41*** (18.91)
MentalHealth	−5.80*** (−18.14)	−1.80*** (−6.15)	−0.11*** (−11.60)	0.02*** (13.84)	−0.02*** (−4.97)	−0.00 (−1.31)	0.04*** (11.40)
Evangelical	0.24 (0.18)	−5.38*** (−4.39)	0.21*** (5.60)	0.00 (0.25)	−0.04* (−2.31)	0.07*** (5.38)	−0.03* (−2.00)
AIAN, %	−31.71*** (−12.03)	−18.63*** (−6.26)	−0.35*** (−6.89)	0.13*** (10.32)	0.04 (0.82)	−0.32*** (−9.71)	−0.30*** (−6.49)
Asian, %	41.24*** (4.11)	21.46** (2.94)	2.99*** (5.72)	−0.09*** (−3.46)	−0.28* (−2.37)	0.04 (0.61)	0.48*** (4.73)
Hispanic, %	−34.23*** (−27.40)	−9.72*** (−8.79)	−0.31*** (−7.73)	−0.05*** (−14.21)	−0.19*** (−8.99)	−0.26*** (−21.67)	0.32*** (20.10)
Black, %	−30.49*** (−19.46)	−18.22*** (−12.76)	−0.30*** (−5.76)	0.06*** (11.46)	−0.43*** (−19.44)	−0.21*** (−14.49)	−0.02 (−1.07)
NHOPI, %	−271.62*** (−5.68)	−75.31* (−2.10)	−10.29*** (−5.49)	0.35** (3.02)	−0.74 (−0.78)	−1.62*** (−5.26)	−0.82* (−2.11)
Rural, %	−30.27*** (−42.19)	−12.66*** (−19.00)	−0.18*** (−8.40)	−0.00 (−0.34)	0.00 (0.45)	−0.09*** (−14.30)	−0.04*** (−4.35)
COVID19 cases				0.25*** (9.88)	0.50*** (3.70)	0.56*** (8.64)	−1.18*** (−11.19)
COVID19 deaths				−1.28 (−1.78)	5.53* (2.19)	−2.84 (−1.54)	4.57* (2.01)
Constant	79.91*** (52.63)	−11.73*** (−8.51)	1.26*** (28.53)	0.08*** (11.06)	0.26*** (9.96)	0.38*** (22.02)	0.33*** (14.19)
Observations	3093	3073	2811	3088	2813	3085	3088
Adjusted R2	0.63	0.42	0.24	0.39	0.34	0.33	0.47

t statistics in parentheses.

*p < 0.05, **p < 0.01, ***p < 0.001.

The results in Table 2 also support our “community distress” hypothesis. Counties reporting higher rates of psychological distress have a statistically significant relationship across all of our measures of vertical trust with the exception of FluVaccine. In terms of political trust, each additional MentalHealth day corresponds with a 5.80 percent decrease in the 2020 Census self-response (model 8), a 1.8 percent decrease in Census responses over time (model 9), and a 0.11 percent drop in the tax compliance rate (model 10). Models 11 and 12 indicate these communities also mistrust scientific authority: MentalHealth corresponds with a slight increase in COVID vaccine hesitancy and a small decrease in COVID vaccination rates. However, in model 7, psychological distress manifests differently than anticipated: MentalHealth correlates with an increase in mask wearing. This could indicate that the logic of wearing a mask differs from the anxiety about vaccines. This would align with Kerr et al.’s (2021) suggestion that psychological distress could stimulate trust in public health authority leading to mask wearing as a low-cost health mitigation behavior in response to health risk uncertainty; vaccination (especially early on) entailed more perceived risk.

Our results remain robust across a broad range of controls, many of which offer meaningful findings in their own right. For starters, every model except for those testing Mask indicates that counties with higher shares of Black Americans mistrust both government and scientific authority at a statistically significant level. The consistency of results between COVID-19 and flu vaccines suggests a lingering link between the historical experience with medical racism and contemporary mistrust in scientific authority. It is also worth repeating how this period corresponds with historic declines in trust in the police and other formally non-partisan, bureaucratic institutions. Latinx populations (Hispanic) also show low levels of political trust, but the coefficient in model 8 changed directions compared to the original correlation matrix, indicating potential multicollinearity, likely due to the Census Bureau’s definition of “Hispanic.” ² Importantly, an inverse relationship with hesitancy (model 11) and low vaccination rates for both COVID (model 12) and flu (model 13) point to an unmet demand for vaccination. American Indian and Alaska Natives (AIAN) communities are strongly associated with low political trust, and little trust in scientific authority as measured by high rates of COVID-19 vaccination hesitancy (Model 11), and low rates of both flu vaccination (model 13) and mask wearing (model 14). Counties with more Asian Americans tell a very different story: these communities show unusually high levels of trust, including a large increase (21.46 percent) in Census self-reporting over the last decade (model 10), along with low rates of COVID vaccine hesitancy (model 11) and high rates of mask wearing (model 14). Like Latinx, they faced some barriers to access for vaccines (Model 12).

With regard to religion, the Evangelical variable overall demonstrates an inverse relationship with trust. Counties with more evangelicals have lower rates of political trust, as indicated by declines in Census self-response rate (model 9) and COVID vaccinations (model 12) and mask wearing (model 14). Other measures of trust are somewhat mixed. Model 10 indicates an increase in the likelihood of tax compliance. More interestingly, FluVaccine shows a slight positive relationship with trust. We attribute this to extensive church involvement with flu vaccine drives (NBC 2020; Wilkes-Barre 2021). Since the flu vaccine has been available for decades, it also highlights the politicization of mistrust in scientific authority as a recent phenomenon.

Finally, our results demonstrate a strong negative effect of rurality on political trust, as models 8–10 indicate that rural counties have lower rates of Census response and tax compliance. But interestingly, while tests for trusts in scientific authority show that people in rural areas are less likely to wear a mask or get vaccinated for the flu, neither COVIDhesitancy nor COVIDvaccine is significant. Overall, this is consistent with previous studies, using Facebook surveys and other data, which find low vaccine uptake without high hesitancy. As mentioned earlier, public health scholars have noted that this finding needs to be explained in the context of rural, conservative political polarization (Callaghan 2023), as well as federal neglect.

Our correlation matrix (Table A3 in supplemental materials) shows a statistically significant, positive relationship between psychological distress and flu vaccines. It demonstrates a positive relationship between an increase in mental health days and mask adherence. This is contrary to our expectations but consistent with literature suggesting that anxiety can impact protective health behaviors (Kerr et al. 2021), including divergent partisan responses: Democrats may wear masks out of fear of the disease while Republicans refuse to do so out of fear they undermine personal liberty. Finally, these findings imply, unsurprisingly, that the uncertainty associated with a novel pandemic generated more distress than the familiar, seasonal flu. COVID-19 vaccines and masks on the one hand, and the flu vaccine on the other, therefore capture different aspects of trust in scientific authority.

Table 3 displays the results with a variable interacting VoteDem and MentalHealth. We find that counties with higher rates of mental health distress have less political trust – and the insignificant coefficient measuring change since 2010 (model 16) compared to the highly significant variable measuring 2020 (model 15) suggests this is a recent phenomenon. These communities also generally have less trust in scientific authority. In fact, the coefficient for VoteDem more than doubles in value in the models with Census1020, TaxCR, COVIDvaccine, and FluVaccine, suggesting that mental health distress is even more important than voting behavior in predicting vertical trust. There are no changes to the Adjusted R² other than very slight changes in models 16 and 19.

OPEN IN VIEWER

Table 3.

Interaction of VoteDem * Mental Health.

	Political Trust			Trust in Scientific Authority
	(15)	(16)	(17)	(18)	(19)	(20)	(21)
	Census20	Census 1020	Tax CR	COVID Hesitancy	COVID Vaccine	Flu Vaccine	Mask
Vote Dem, %	34.02*** (3.69)	47.64*** (6.71)	1.62*** (4.36)	−0.10** (−2.83)	0.02 (0.22)	0.59*** (7.03)	0.75*** (8.01)
MentalHealth	−4.73*** (−6.76)	−0.08 (−0.14)	−0.02 (−0.75)	0.02*** (6.27)	−0.05*** (−6.71)	0.02** (3.22)	0.06*** (8.93)
Evangelical Adher rate	−0.18 (−0.13)	−6.07*** (−4.83)	0.17*** (4.64)	0.00 (0.18)	−0.03 (−1.44)	0.06*** (4.45)	−0.04* (−2.57)
AIAN, %	−30.29*** (−11.25)	−16.72*** (−5.69)	−0.25*** (−4.48)	0.13*** (11.04)	−0.00 (−0.07)	−0.29*** (−8.20)	−0.26*** (−5.75)
Asian, %	36.31*** (3.35)	13.49 (1.77)	2.58*** (4.89)	−0.09** (−3.23)	−0.12 (−1.03)	−0.08 (−1.14)	0.36*** (3.43)
Hispanic, %	−33.91*** (−26.82)	−9.21*** (−8.26)	−0.29*** (−7.15)	−0.05*** (−13.85)	−0.20*** (−9.29)	−0.25*** (−20.84)	0.33*** (20.51)
Black, %	−29.15*** (−17.67)	−16.02*** (−10.36)	−0.18** (−3.26)	0.06*** (10.38)	−0.47*** (−18.83)	−0.18*** (−11.63)	0.01 (0.44)
NHOPI, %	−265.28*** (−5.52)	−64.85 (−1.82)	−9.80*** (−5.22)	0.35** (3.02)	−0.89 (−0.94)	−1.46*** (−4.68)	−0.68 (−1.71)
Rural, %	−30.24*** (−42.08)	−12.63*** (−18.95)	−0.18*** (−8.14)	−0.00 (−0.34)	0.00 (0.44)	−0.09*** (−14.35)	−0.04*** (−4.39)
Int Dem voters *MH	−3.86 (−1.73)	−6.24*** (−3.49)	−0.33*** (−3.76)	−0.00 (−0.34)	0.12*** (4.45)	−0.10*** (−4.45)	−0.09*** (−3.70)
COVID19 cases				0.25*** (9.77)	0.53*** (3.88)	0.53*** (8.20)	−1.20*** (−11.37)
COVID19 deaths				−1.26 (−1.73)	4.51 (1.78)	−2.03 (−1.08)	5.31* (2.33)
Constant	75.81*** (26.01)	−18.30*** (−7.35)	0.91*** (8.98)	0.08*** (6.47)	0.38*** (10.79)	0.28*** (9.65)	0.24*** (7.31)
Observations	3093	3073	2811	3088	2813	3085	3088
Adjusted R2	0.63	0.42	0.25	0.39	0.35	0.33	0.47

t statistics in parentheses.

*p < 0.05, **p < 0.01, ***p < 0.001.

Conclusion

Overall, our empirical analysis offers strong support for both of our predictions. Trust in government and in scientific authority is strongly linked to electoral support for Democrats, supporting our “partisan mistrust” hypothesis. In addition, counties suffering from psychological distress have a significant inverse relationship with vertical trust, regardless of partisan preferences (and with the partial exception of flu vaccination and mask wearing). These statistical findings question research that “bias is bipartisan” (Ditto et al. 2019), and generally affirm the asymmetry hypothesis (Jost et al. 2017).

A further contribution of this study relates to our empirical strategy. Most research on political trust uses individual-level data to capture attitudinal dimensions (Albarracín et al. 2022; Devine et al. 2021). However, our observational data captures what people actually do, not just what they say. It thus enables us to demonstrate the behavioral manifestations of low trust, which were visible in tax compliance and Census response rates—and one might add in the January 6 insurrection. While political mistrust has been low for decades, and some argue it is not necessarily harmful (Norris 2022), results here using variables such as flu vaccinations highlight that mistrust in scientific authority is indeed relatively recent. Our strategy also emphasizes the construction of mistrust and the experience of psychological distress at the community level. Distress is strongly impacted by social, economic and cultural conditions that are difficult to manipulate in laboratory settings. Therefore, while our findings about different types of vertical mistrust aligning largely supports the psychological propensity model, which emphasizes individual characteristics and learned experiences, the results here on demography, ethnicity and other factors suggest a fusion of the psychological and the social experiences. This is a useful insight, since such factors are typically treated separately in constructs of trust (Newton et al. 2018). Moreover, since our measures of scientific authority convey information about how people assess fact claims, our findings add to emerging research highlighting how the decline of trust in nonpartisan institutions presents risks to democracy (Dawson and Krakoff 2023).

The high rates of vertical trust by Asian Americans, in contrast to every other racial and ethnic minority, indicate that restoring trust and perhaps countering misinformation will require culturally nuanced communication. Other controls point to high levels of mistrust among evangelicals, underscoring concerns about potential sympathy for extremist ideologies in this conservative religious community (Cox 2021). With regard to rural Americans, our results show a strong and consistent inverse relationship with political trust, but mixed effects on trust in scientific authority. This is consistent with the anti-intellectualism embedded in rural mistrust (Trujillo 2022), but it also may suggest that interactions with medical professionals in rural areas involve more inter-personal trust—separating scientists or doctors from populist notions of elites.

Our findings also emphasize the value of proactive public policy: the steep increase in vaccinations, accompanied by a decline in hesitancy among Native Americans suggest that culturally sensitive health messages can help build vertical trust (Read 2021) and reduce epistemic fragility. We are learning that legitimacy derives from more than just good governance or adequate government capacity. It requires the community resilience and trust that foster sound epistemic judgment.

Supplemental Material