Introduction

As vaccinations rates for COVID-19 remain below the targets set by public health authorities in the U.S., it is important to understand both population-level willingness to vaccinate against COVID-19, as well as the reasons behind stated attitudes of vaccine acceptance or hesitancy in order to tailor interventions to increase uptake. Vaccine hesitancy, which has been defined as either a delay in acceptance or refusal of vaccination despite accessibility and affordability of vaccination services [1], remains a challenge in the United States: A recent review of the 106 articles from nationally representative samples found an overall COVID-19 vaccine acceptance rate ranging from 53.6% to 84.4% [2], and the population vaccination rate as of April 2022 was 66% [3]. As many COVID-19 related restrictions began to be lifted in early 2022, including vaccine mandates, many public health experts have warned about the further dampening effects on vaccine uptake, and the concurrent need for more informational campaigns to continue to encourage people to get vaccinated and boosted [4,5].

Even before COVID-19 vaccines became widely available to the general public, academic researchers and media reporters sought to identify subpopulations that expressed higher levels of vaccine hesitancy and lower intention to vaccinate against COVID-19. A review of 13 published studies with national samples in early 2021 found that African-Americans had significantly heightened hesitancy compared with other groups [6,7]. Further reinforcing these findings, in their review of 106 studies, Wang and Lie found that compared to non-Hispanic whites, identifying as non-Hispanic Black or Hispanic was associated with 2 to 6-fold higher chance of being COVID-19 vaccine hesitant, respectively [2]. However, as the vaccine roll-out began in early 2021, emerging polls and media content began to show that vaccine hesitancy has was declining in non-Hispanic Blacks, but remaining elevated in White Republicans. By March 2021polls showed that nearly 60% of white Republicans reported they would either not take the vaccine or were unsure if they would [8]. Research also increasingly shows supporters of former President Donald Trump to be particularly reluctant to vaccinate, despite the fact that Mr. Trump himself received the vaccine, and that his administration pushed for the rapid development of a COVID-19 vaccine [9,10]. This finding of elevated vaccine hesitancy in Republicans is consistent with broader polarization of views about COVID-19 along partisan lines, which emerged as early as the first weeks of lockdown in the US in March 2020 [11]. In a systematic review, Wang and Lie found at least 10 studies in which authors reported positive associations between identifying as a Republican or a conservative and intending to refuse vaccination [2].

To better understand these patterns, researchers have examined the sources of distrust in different groups, in order to unpack the mechanisms or attitudes which underly this distrust. A large literature examines broader vaccine hesitancy among African-Americans, rooting this mistrust of the medical establishment both in historical and contemporary practices and experiences [12–15]. While historic events have led to skepticism among African-Americans toward medical institutions in general and vaccines in particular, the roots of mistrust are not strictly historical. Overt racism and explicitly unethical practice has given way to less overt, implicit forms of racial bias and disregard in the health system [16,17]. Historic and on-going lived experience manifest as lower trust in medical research, scientists and doctors among non-Hispanic Blacks relative to other race-ethnic groups [16,18]. Less is known about vaccine hesitancy among Hispanic populations and its intersections with political partisanship, as party identification among Hispanics tends to vary by national origin [19]. However, a recent review of 13 studies found heightening COVID-19 vaccine hesitancy among Hispanics compared with non-Hispanics Whites, though to a lesser extent than non-Hispanic Blacks [6].

While medical mistrust has been identified as a source of hesitancy among non-Hispanic Blacks, less is know about how mistrust might affect intention to vaccinate among other groups. Mistrust of government has been growing generally in American society for some time [20]. Growing mistrust in mainstream institutions has also been identified as an important predictor of support for former President Trump in the U.S. [21–23], although it also long predated the Trump administration and has fueled populist movements globally [24]. Furthermore, there is evidence that far-right extremist groups that spread false information about the 2020 US presidential election have continued similar activities regarding vaccines [25,26]. Studies find that certain sub-groups of political conservatives are more likely to be vaccine-hesitant and vulnerable to misinformation than others [27]. Political populism has also been tied to both ethno-nationalism (nativist and nationalistic tendencies) and a preference for “common sense” knowledge over elite knowledge, which contributes to skepticism in medical and scientific experts [28]. This suggests that low levels of trust in institutions–whether driven ascriptive characteristics or political partisanship–may be an important underlying driver of vaccine hesitancy.

The evolution of these partisan dynamics presents a dilemma for message framing aimed at increasing vaccine uptake for COVID-19 and beyond. On the one hand, to increase vaccine confidence among non-Hispanic Blacks, health communication scholars have recommended counteracting mistrust by acknowledging that medical mistrust is justifiable based on past and present harms [29]. This may also involve “role-modeling” from public officials, health authorities and influencers to help in building public trust as well as race concordance with the messenger [30].

On the other hand, it is also possible that framing COVID-19 as a disease that primarily affects minorities and/or prioritizing minority communities for vaccination may stimulate racial animus in non-Hispanic Whites, and thereby reduce their own intentions to vaccinate, especially among Whites with higher levels of racial bias. Thus, framing vaccine messages around race potentially could be a double-edged sword, as shown by several recent studies. Harell and Lieberman demonstrate that informing White Americans about racial disparities in COVID-19 mortality rates decreased support for public health measures among those with racially biased views, while increasing support among those with less bias [31]. Similarly, Skinner-Dorkenoo et al. found that manipulating exposure to information about COVID-19 racial disparities in infections and deaths reduced fear of COVID-19, empathy for those vulnerable to COVID-19, and support for safety precautions among those randomized to read about persistent inequalities [32]. Acknowledging racism and suggesting racial prioritization could build vaccine confidence in racial and ethnic minorities, but it could also raise resentment and reduce confidence among Whites.

In this context, we examine racial, ethnic and partisan differences in vaccine hesitancy, particularly the contribution of mistrust in institutions on intention to vaccinate, through a multi-method approach. Given variation in trust in the medical establishment and the novel context of a pandemic and new vaccine, we explore three key questions. First, does experimental exposure to a news report favoring minority prioritization for the vaccine, due to the disproportionate burden of COVID-19 on these groups, affect intent-to-vaccinate? If so, does the impact of reading this news report vary along race-ethnic and partisan lines? Second, how do respondents explain their intent-to-vaccinate (or not) in their own words? Do these explanations vary across race-ethnic and partisan lines? Third and finally, to what extent are race-ethnicity and party identification associated with intent-to-vaccinate for COVID-19 and what factors, including mistrust, are correlated with elevated hesitancy in these groups?

We studied these questions with data gathered from an online survey and experiment with a diverse sample of 1,353 respondents in New York State. New York State is an important context in which to study race-ethnic disparities in vaccine hesitancy as the New York City metro area was the epicenter of the first US wave of COVID-19, and non-Hispanic Black and Hispanic populations, which are concentrated in New York City, were disproportionately affected by COVID-19 morbidity and mortality. However, NYS is also demographically and politically diverse; much of upstate is more rural, white and conservative than highly diverse communities in New York City and other locations downstate. We utilized linked closed-ended and open-ended questions to generate deeper insights on respondents’ intentions to get a COVID-19 vaccine before the vaccine was available.

Methods

Analysis

Results

Quantitative results

Descriptive results.

In this sample, non-Hispanic Blacks were 6 percentage-points more likely than non-Hispanic Whites or Hispanics to say that they definitely would not get a COVID-19 as soon as it is available to them: 18% of non-Hispanic Blacks said they definitely would not vaccinate as soon as possible compared with 12% for both non-Hispanic Whites and Hispanics (Fig 1 & Table 1). Non-Hispanic Whites were more likely to say that they definitely would vaccinate (52%) compared with non-Hispanic Blacks (33%) and Hispanics (42%) (Fig 1 & Table 1). Respondents who reported voting for Mr. Trump in 2020 were 8 percentage-points more likely to say they definitely would not vaccinate as soon as possible than 2020 Biden voters (19% versus 11%); Trump voters were similar to respondents who did not vote or voted for another candidate (18%) (Fig 2 & Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. COVID-19 vaccination intentions by race/ethnicity.

https://doi.org/10.1371/journal.pone.0277043.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. COVID-19 vaccination intentions by vote choice 2020.

https://doi.org/10.1371/journal.pone.0277043.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Bivariate associations: Vaccine intentions and predictors.

https://doi.org/10.1371/journal.pone.0277043.t001

Survey experiment.

We found no effect of the survey experiment in adjusted models on intent-to-vaccinate across the sample as a whole (see Table 2), and no significant effects at the 95% confidence level within racial, ethnic and voter subgroups (Fig 3). We consider this to be a true null.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Predictive margins, experimental conditions by race/ethnicity and vote choice.

Notes: DV = Definitely will not get the vaccine (3), Probably will not get the vaccine (2), Probably will get the vaccine (1), Definitely will get the vaccine (0). Controls included: gender, age, education, income, co-morbidities, family member died from COVID.

https://doi.org/10.1371/journal.pone.0277043.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Stepwise ordered logistic regression, disparities in vaccine hesitancy.

https://doi.org/10.1371/journal.pone.0277043.t002

Stepwise regression.

Table 2 summarizes the results of the stepwise regression. Non-Hispanic Blacks were nearly twice as likely to be unlikely to vaccinate as soon as possible compared with non-Hispanic Whites (Model 3: OR = 1.82, p<0.01) adjusting for controls and confounders, and remained more hesitant after introducing explanatory variables (institutional trust, alternative media, individualism) (Model 4: 2.01, p<0.01). Hispanics were also significantly more likely to be hesitant compared with non-Hispanic Whites though to a lesser degree than non-Hispanic Blacks (Model 3: OR = 1.37, p<0.05), and remained more hesitant after introducing controls and explanatory variables (Model 4: OR = 1.43, p<0.01).

Voting for Trump in 2020 was also associated with lower intention to vaccinate adjusting for controls and confounders (Table 2, Model 3: OR = 1.74, p<0.01). Not voting in 2020 or voting for other candidates besides Biden or Trump was also associated with a lower intent to vaccinate (Table 2, Model 3: 1.50, OR = p<0.05) However, this effect was attenuated and no longer significant after the introduction of mediating variables (Table 2, Model 4, OR = 1.15, 1.22 respectively, p>0.1).

In all specifications, gender was a strong predictor of hesitancy. Women were 95% more likely to be hesitant than men even after adjusting for conrols, confounders and potential mediators (Table 2, Model 4, OR = 0.95, p<0.01). Respondents older than age 40 were also more likely to be hesitant, including in models with explanatory mechanism variables included (Table 2, Model 4). Factors associated with reduced hesitancy included higher income, having 3 or more comorbidities, and having a family member die of COVID-19. These remained significant with the introduction of mediating variables. Watching alternative media was strongly associated with hesitancy (Model 4: OR = 1.93, p<0.01), as was distrust in medico-pharmaceutical institutions. Respondents who reported having very little trust in the medico-pharmaceutical institutions were nearly four times more likely to not intend to vaccinate (Table 2, Model 4, OR = 3.76, p<0.01). Scoring high on the individualism scale was also associated with a lower intent to vaccinate (Table 2, Model 4: OR = 1.42, p<0.01).

Qualitative results

Free-response qualitative data—even thin, short-response answers—sheds light on what is top-of-mind (and willing-to-share) for respondents as they score their intent-to-vaccinate as soon as possible (see Supplmenatary Materials 4, S4 1–9 Fig in S1 File).

Among those who expressed willingness to get vaccinated as soon as possible, four themes emerged most strongly. The first is protection: almost half of respondents noted belief in the protective value against Covid-19 for themselves and/or their families as motivating them to get the shot as soon as possible. For example, one person noted, “I work outside the home and need to protect myself to protect my family,” while another noted wanted to take proactive “steps to protect myself.” Many wrote in the exact phrase “to protect myself and my family.”

Second, and offering another dimension to why protection matters to them, some high intent-to-vaccinate respondents specifically noted high perceived suspectibility to Covid-19 or complicated sequelae from infection, judging themselves at high risk for negative outcomes given their current health status (e.g., “I am black and old”) and/or profession (e.g., “I am an essential worker”). A few in this situation noted being “scared of Covid” and that vaccination would bring “piece of mind.”

An additional set of responses (roughly a tenth in each case) evoked a sense of collective responsibility, framing vaccination as “doing [one’s] part” or as a “public duty,” extending the idea of protection beyond the self or household. Finally, people expressed a desire to return to normal life, and understood their vaccination decision to be a key route to “be free.” In general, these themes were articulated with similar frequency across our demographic and political groups of interest, though non-Hispanic Whites chose to write about both about perceived susceptibility to Covid-19 and collective responsibility more than other groups (see S4 8 Fig in S1 File).

Among those who did not want to get vaccinated as soon as possible, two sometimes overlapping themes came out most strongly: concerns about the safety of the new Covid-19 vaccination and an explicit desire to “wait and see” about potential side effects of the vaccination. This included wanting others to “get it first” and to be “cautious” about their own decision. A subset wanted to hear about the safety of the vaccine specifically for people like them (age, race, disease status) and to hear it from their own or another healthcare professional. Others noted that it was “too soon” or “rushed;” a few among these cited a specific mental model that a vaccine “should take ten years to develop” and therefore needed “further testing.” A small (n = 8) subset of these—all non-Hispanic Black respondents—explicitly stated not wanting to “be a guinea pig” or a “lab rat.”

Among othose who said they would never get the vaccine, a small number of respondents wrote both about generalized vaccine refusal or specific rumors circulating along with the Covid-19 pandemic. Generalized vaccine concerns and refusal—expressed by Biden-voters, Trump-voters, and non-voters—were articulated as not “liking,” “trusting,” “believing in,” or being “keen on” vaccines across the board. Meanwhile, respondents (n = 12) who elaborated on never wanting to get the vaccine with conspiratorial references included skepticism about the vaccine (e.g., “I have concerns about microchips in the vaccine”) and about Covid-19 (e.g., “I think there is something not kosher about this whole alleged ‘crisis.’”). These respondents identified as Trump-voters. Unexpectedly, no one wrote in with specific concerns about the vaccine and pregnancy and/or fertility.

Though there were differences across race, ethnic and partisan groups in the themes mentioned, the differences were smaller than differences in the close-ended questions, suggesting that all race-ethnic groups share similar underlying motivations and concerns but may weigh these differently and/or express those differently on a Likert-type scale—a measurement point that warrants further investigation in future research.

Discussion

A large majority (86%) of respondents in this diverse sample of New York-based adults expressed some degree of willingness to be vaccinated for COVID-19, although an important segment (14%) of the sample stated that they definitely would not get the vaccine as soon as possible. Consistent with previous literature [41–43], we found racial and ethnic differences in vaccine hesitancy in the pre-rollout period, with non-Hispanic Blacks being 6 percentage points more likely to say that they definitely would not vaccinate as soon as possible for COVID-19 compared with non-Hispanic Whites and Hispanics. Hispanics were also more likely to be hesitant than non-Hispanic Whites, with fewer saying they would definitely vaccinate as soon as possible (42% compared with 52%). Trump voters were 8 percentage points more likely to say they definitely would not vaccinate compared with Biden voters and those who as well as people who did not vote in 2020 or voted outside of the two main candidates were 7 percentage points more likely to say they definitely would not vaccinate. This is consistent with literature showing heightened hesitancy among people who identify as conservative and/or Republican [6,44–48].

These descriptive findings should be interpreted in light of the situation in November/Decmeber of 2020 when the survey was fielded. This was a period before COVID-19 vaccines became widely available and began being rolled out to essential workers, which started in January 2021. People’s intentions were more hypothetical at this point since vaccines had not yet been made available yet. It was also a period just following the 2020 Presidential election when Donald Trump lost the election. Media reporting at this time suggested that the vaccine may have been rushed for political reasons to re-open the economy [49]. Political right opposition to vaccination may have therefore been lower at this time given its association with the Trump administration, but also changing given his political loss.

To examine whether a racial justice message about vaccination might reduce hesitancy in non-Hispanic Blacks, but increase hesitancy in non-Hipanic Whites, we experimentally tested the effect of reading a news article about racial prioritization for the COVID-19 vaccine on intent to vaccinate. These messages did not affect expressed willingness to vaccinate overall or for any race, ethnic or voting group; there was no positive effect among non-Hispanic Blacks/Biden voters nor was there a backlash effect among non-Hispanic White/Trump voters. These results run counter to our hypotheses that exposure to a message acknowledging justified mistrust in African-Americans could reduce hesitancy in non-Hispanic blacks and Biden voters/Democrats, and would likely have no effect or could even possibly backfire among non-Hispanic Whites and Trump voters/Republicans. The overarching null findings should be interpreted in light of the fact that there was relatively limited power to detect small heterogeneous treatment effects in this sample. Furthermore, other studies have examined the moderating effects of racial attitudes and prior knowledge of COVID-19 disparities on similar exposures [31]. Heterogeneity in treatment effects according to these factors could account for the null effect in this setting.

In observational analyses of the survey data, we found that the lower intent to vaccinate among Trump voters and non-voters was no longer significant after controlling for a hypothesized set of mediators, including distrust in medico-pharmaceutical institutions, consumption of alternative media, and individualistic values. This suggests that hesitancy based on political partisanship can largely be explained by these factors. Distrust has been highlighted both as an independent predictor of support for former President Trump [50] and a predictor of hesitancy [51]. This pattern may highlight an underlying latent set of trust-related characteristics driving both attitudes. This group has been shown to have more anti-establishment attitudes [35] and may also be a target for anti-vaccine messages on social media [26].

Unlike vote choice, the significance of race and ethnicity persisted even after inclusion of measures of mistrust, alternative media use, and individualism. It is possible that the particular measures of mistrust were insufficient to capture the nuanced mechanisms through which distrust in these institutions shapes COVID-19 vaccine attitudes. Prior studies have found that mistrust among Black populations in the U.S. may be further exacerbated by the general sociopolitical climate in the US [52], or potential confounding with socioeconomic status that is diificult to measure accurately and adjust for in observational studies with online samples (e.g., lack of financial resources or housing insecurity) [53].

While we found race, ethnicity, 2020 vote choice, and partisanship to be significant predictors of hesitancy, gender and age were even more consistently and strongly associated with hesitancy even after adjustment for other factors, including institutional mistrust. This is consistent with previous literature that has shown women to be more vaccine hesitant [45]. The age effect is likely representative of a cohort effect whereby there are generational differences in acceptance of COVID-19 vaccines, but this cannot be separated in observational studies. Other studies have found an inverted U-shaped relationship in the relationship between age and vaccine hesitancy as well [54]. This suggests that in addition to considering race and ethnicity, gender and certain age cohorts may be additional groups that can be targeted for additional vaccine promotion.

Using thin qualitative data of how respondents explain briefly their intent-to-vaccinate choice, we found hesitancy views were present in all race, ethnic and voter groups and that the major categories of objections did not manifest differently across groups. The complexity of vaccine hesitancy and its presence across racial, ethnic, gender, and age groups highlights the need for a deeper understanding of the relationship between the expression of hesitant attitudes and their translation into intention to vaccinate.

The qualitative responses also revealed relatively few explicit references to politics, although trust in the political system and the skepticism that a “safe” vaccine could be produced in such a short time frame was prevalent. This may be surprising given that the politics of vaccination has been salient throughout COVID-19 in the United States over the period preceding the survey, which took place several weeks after the US presidential election in November 2020. Despite widespread concern about the rise of misinformation and fake news spread through social media, and potential susceptibility of people already mistrustful of mainstream institutions to mis/dis-information [55–58], write-in references to outright conspiracies were rare and were no more likely to be held by racial or ethnic minority groups. More frequent (although still rare) references to conspiracy theories among Trump supporters may reflect a sense of mistrust among this group, or greater exposure to misinformation.

The fact that ascriptive characteristics of individuals, i.e., characteristics corresponding to one’s external identity assigned by society, remained significant and even increased in magnitude after adjusting for potential confounders and potential mediating factors, including mistrust, deserves further scrutiny. This suggests that these identity-based characteristics are meaningful in themselves, and cannot be reduced to a set of opinions or ideas often theorized to explain or mediate these effects (i.e., lack of trust). Furthermore, the fact that the experimental condition did not significantly alter vaccination intentions also suggests that message framing acknowledging justified mistrust in minorities may not be sufficient on its own to overcome hesitant attitudes. Future research may want to assess how much impact message framing has on its own compared with other factors known to be important in vaccine uptake including peer influence, communication from physicians and mandates.

Conclusions

Generalized vaccine hesitancy has been observed in many groups and geographies. In the context of COVID-19, while initial concerns about vaccine hesitancy focused on non-Hispanic Blacks, increasingly concern has fixated on the low vaccination rates among Republicans and supporters of former President Trump. This polarization of attitudes poses challenges for public health strategies to build vaccine confidence. To increase vaccine confidence, health communication scholars recommend acknowledging disparities and the reality that medical mistrust has a basis in histories of discrimination. However, it is possible that this framing could come at the expense of reducing hesitancy in other groups.

We find that an experimental message conveying racial vaccine prioritization and acknowledging structural racism in COVID-19 disparities has no effect on intent to vaccinate overall or in subgroups. Moreover, we find lower intentions to vaccinate among non-Hispanic Blacks, Hispanics, females and certain age cohorts, that are not accounted for by higher degrees of institutional mistrust, whereas for Trump voters and non-voters, the correlation between this identity and hesitancy is no longer evident after controlling for low trust and ideological factors.

These results reinforce the idea that for some groups, strong hesitancy views may be relatively fixed and difficult to change at least with relatively simple messaging campaigns. However, qualitatively we also find strong similarities in the kinds of reasons offered by individuals across racial and ethnic groups, suggesting that for some “softer” elements of hesitancy (notably the “wait and see” group), including that general observation of safe and effective vaccination among respondents’ social circles, may reduce hesitancy. When introducing new technologies in the future, health communicators might consider how to make the (lack of) side effects and limited disruption to daily life more visible. More positively, the racial-prioritization prime did not backfire and produce lower intentions in non-Hispanic Whites nor Trump voters, suggesting that framing minorities as disproportionately at risk and deserving of prioritization is not necessarily a polarizing or zero-sum message.

In the short run, people may eventually choose to vaccinate in spite of hesitancy in order to get back to “normal” life, but the deep and enduring mistrust of public institutions that underlie hesitancy, are unlikely to reduce any time soon and may to continue to hinder public health efforts into the future. More attention should be given to how to build trust in medical institutions and to depoliticize and depolarize opinions on vaccines, so that this and future vaccination campaigns will be starting from a period of high trust among all social, ethnic and racial groups.

Supporting information