key: cord-1027881-4jogln8t
authors: Kreps, S. E.; Kriner, D. L.
title: Factors Influencing Covid-19 Vaccine Acceptance across Subgroups in the United States: Evidence from a Conjoint Experiment
date: 2021-04-24
journal: Vaccine
DOI: 10.1016/j.vaccine.2021.04.044
sha: 632dc2df6a7d6d4ae236623291e56927e77ff70d
doc_id: 1027881
cord_uid: 4jogln8t

Public health officials warn that the greatest barrier to widespread vaccination against Covid-19 will not be scientific or technical, but the considerable public hesitancy to take a novel vaccine. Understanding the factors that influence vaccine acceptance is critical to informing public health campaigns to combat public fears and ensure broad uptake when a vaccine becomes available. Employing a conjoint experiment embedded on an online survey of almost 2,000 adult Americans, we show that the effects of seven vaccine attributes on subjects’ willingness to vaccinate vary significantly across subgroups. For example, vaccine efficacy was significantly more influential on vaccine acceptance among whites than among Blacks, while bringing a vaccine to market under a Food and Drug Administration Emergency Use Authorization had a stronger adverse effect on willingness to vaccinate among older Americans and women. Democrats were more sensitive to vaccine efficacy than Republicans, and both groups responded differently to various endorsements of the vaccine. We also explore whether past flu vaccination history, attitudes toward general vaccine safety, and personal contact with severe cases of Covid-19 can explain variation in group vaccination hesitancy. Many subgroups that exhibit the greatest Covid-19 vaccine hesitancy do not report significantly lower frequencies of flu vaccination. Several groups that exhibit greater Covid-19 vaccine hesitancy also report greater concerns about vaccine safety generally, but others do not. Finally, subgroup variation in reported personal contact with severe cases of Covid-19 does not strongly match subgroup variation in vaccine acceptance.

The human toll of the Covid-19 pandemic has continued unabated. Early questions about 48 how quickly societies would achieve herd immunity have been answered by evidence that even in 49 former hotspots such as Spain, seroprevalence was only about 5% after the first wave in spring 2020 50 [1] . The combination of ongoing transmission and likelihood that herd immunity will not be reached 51 expeditiously through community infection has led to a commensurate demand for a vaccine to 52 control the spread or intensity of the virus. By early 2021, multiple vaccines showing strong efficacy 53 data from clinical trials have been authorized for public use [2] . Yet even an effective vaccine may 54 have limited public health benefits if significant percentages of the public think the vaccine is unsafe 55 and are unwilling to vaccinate. In recent years, anti-vaccination groups have undermined public 56 uptake of vaccines, leading to recent increases in vaccine-preventable diseases such as measles in 57 Europe and the United States [3] . These groups have already begun mobilizing to challenge uptake 58 of the Covid-19 vaccine, potentially interfering with public health authorities' hopes of widespread 59 immunization. As former Director of the Centers for Disease Control, Thomas Frieden, observed, 60 "This is the first time we've had an anti-vaccine movement before we've had the vaccine [4] ." 61 Early public opinion surveys on likelihood of vaccination have offered widely varying 62 estimates of vaccine hesitancy among US adults [5] . Most of these surveys are based on generic 63 wording about vaccination that provide little context and no information about the specific 64 attributes of the vaccine. Evidence, first emerging from clinical trials and now from real-world data 65 about approved vaccines has begun to clarify the nature of the vaccines available to the public. The 66 appropriate public health question to pose is not about generic vaccination, but vaccination 67 conditional on the specifics of the vaccine: its efficacy, side effects, and other characteristics. 68 Further, because the human toll is uneven-minority groups and the elderly are disproportionately 69 affected-and rarely is vaccine hesitancy population-wide but rather localized demographically [6] , understanding the factors that most influence vaccine hesitancy among subgroups is crucial for 71 identifying particular vulnerabilities and informing efforts to address them. 72 Through a conjoint experiment conducted in the United States, we studied the vaccine 73 attributes that affect Americans' willingness to vaccinate against Covid-19 and, critically, how the 74 effects of those attributes vary across demographic subgroups. While previous research has 75 employed conjoint experiments to understand how COVID-19 vaccine attributes would affect 76 vaccine hesitancy [7] [8] [9] [10] , these studies focus primarily on the effects of vaccine attributes on 77 willingness to vaccinate among the population as a whole, while finding only limited evidence of 78 heterogenous treatment effects [8] . Previous research on other vaccines has shown population 79 subgroups may react differently to the same vaccine attributes [11, 12] . Our analysis builds directly on 80 one of the first conjoint experiments into the factors driving US adults' willingness to take a 81 COVID-19 vaccine [7] and investigates heterogeneity in how subgroups respond to different 82 attribute levels. Our findings of heterogeneous effects can aid the targeting and content of public 83 health outreach to specific communities to combat vaccine hesitancy. To assess the influence of a hypothetical vaccine's characteristics on willingness to vaccinate 88 and how the effects of these attributes vary across important subgroups of the population, we 89 employed a conjoint experiment. Conjoint analysis is a survey-based method commonly used in 90 market research to study how consumers value different attributes of a product or service [13] . The 91 methodological approach has been widely adopted in health research [14] , and has been shown to 92 approximate real-world decisions and reasonably predict health behaviors [15] . A fully randomized 93 conjoint is a full factorial design in which all possible combinations may not be observed. However, randomization assures that attributes are orthogonal, which allows the estimation of the marginal 95 effects of each attribute. An advantage of this approach is that it does not rest on an assumed 96 behavioral model of individual decision-making [16] . The estimated treatment effects are 97 nonparametrically identified under a modest set of assumptions, many of which (such as 98 randomization of attribute levels) are guaranteed by the experimental design [17] . 99 

The experiment presented each subject with five pairs of hypothetical vaccine profiles. Authorization; the national origins of the vaccine, and the person or entity endorsing it. Table 1 109 summarizes the attributes and levels. A sample choice set is presented in the Supporting Information 110 (SI Figure 1 ). After viewing each pair of hypothetical vaccine profiles, subjects were first asked to 111 indicate whether they would choose Vaccine A, Vaccine B, or neither. Subjects were then asked to 112 evaluate how likely they would be to accept each vaccine individually on a seven-point ordinal scale 113 from "extremely unlikely" to "extremely likely." Our analysis here focuses on responses to this 114 second measure of vaccination intentions. 

Estimating treatment effects in a fully randomized conjoint experiment is straightforward.

As described in Hainmueller, Hopkins, and Yamamoto [17] , the regression coefficients from a Our analysis here re-analyzes data from Kreps et al's study [7] subgroups and ensuring that any observed differences (or lack thereof) are not an artifact of the 152 chosen baseline level of each attribute [25] . However, the two quantities are directly related; in a 153 fully randomized conjoint, the AMCEs are the differences between marginal means of a given 154 attribute-level and that attribute's baseline level, all else equal.

In this analysis, we look for evidence of differential treatment effects along four dimensions: 156 race/ethnicity; age; gender; and political partisanship. We focus on the first three categories because 157 prior research shows that vaccine hesitancy is likely stronger among people of color [26] and women 158 [27] , and because people of color and the elderly are at disproportionate risk from Covid-19.

Moreover, a recent study also employing a conjoint analysis reported evidence of heterogeneous 160 treatment effects by age [8] . 

Each vaccine attribute summarized in Table 1 significantly affected willingness to vaccinate 169 in the aggregate [7] . However, this masks significant variation in the effects of different attributes 170 across key demographic subgroups. Covid-19 vaccine than were whites. Moreover, several vaccine attributes had significantly different 174 effects on uptake across racial and ethnic groups (SI Table 2 We find little evidence that vaccine attributes had significantly different effects on the 205 vaccination intentions of Latinos vs. whites. As seen in the right panel of Figure 1 , the general 206 pattern of effects is quite similar, with the only exception being protection duration. A longer 207 protection duration significantly increased willingness to vaccinate among whites, but had no effect 208 among Latinos. acceptance. For example, a longer protection duration had a greater positive effect on willingness to 252 vaccinate among women than it did among men (SI Table 4 ). An FDA Emergency Use

Authorization significantly reduced vaccine acceptance among women by 4%. Among men, the 254 difference was roughly .5% and not statistically significant (the difference in effect sizes is 255 statistically significant: see SI differences in effect sizes are statistically significant (SI Table 5 ).

Partisanship also significantly moderated the effects of endorsements on vaccine acceptance.

Unsurprisingly, Republicans were significantly more likely to vaccinate when the vaccine was 273 endorsed by President Trump than by then-former Vice President Biden, and vice versa for Democratic respondents. Democrats were also significantly more likely to accept vaccination than 275 Republicans when the vaccine was endorsed by the CDC. The first model of Table 2 shows that older Americans were more likely to report frequent against studies showing that those who question vaccine safety are less likely to vaccinate generally 320 [33] . Moreover, Latinos were significantly more skeptical of vaccine safety in our sample than 321 whites, but they were only marginally less likely to vaccinate against Covid-19 in our experiment.

Older Americans were significantly more confident in general vaccine safety than were younger 323 Americans, and yet they were more hesitant to take a Covid-19 vaccine in our study.

While we did find that women were both more skeptical of vaccine safety in general and 325 more hesitant to take a Covid-19 vaccine, recent research suggests that a primary driver of this 326 hesitancy among women may be online misinformation baselessly alleging that Covid-19 vaccines 327 can jeopardize fertility [34] . If this is the dominant force behind the gender gap, we would expect 328 the gender gap to be greater among women and men of child-bearing age than between older men 329 and women. Figure 5 offers little support for this hypothesis. The gender gap between men and 330 women under 40 is, if anything, slightly smaller than the corresponding gender gap for those 40 and 331 over. 3 332 3 We chose 40 as the cut-off because the 2018 birth rate for women 40-44 was 11.8 per 1,000 and for women 45-49 just .9 per 1,000, vs. 52.6 per 1,000 for women 35-39 [48] . Alternate cut-offs yield similar results. Our research corroborates studies on other viruses that find lower vaccination rates for 400 women versus men [27] and emerging evidence about Covid-19 vaccine preferences [5] . Vaccine 401 safety has been cited as a cause of concern [43] , and indeed our analysis shows that women tend to surprising, though additional analyses (SI Table 6 ) show that the relationship between age and 415 vaccine acceptance in our study was curvilinear Most opinion surveys find older adults are more 416 likely to vaccinate than younger adults [5] . However, most of these survey questions ask about 417 willingness to take a generic vaccine. The relationship here is consistent with a pair of additional 418 studies that also recruited subjects from the Lucid platform and employed a conjoint experiment to 419 examine the effects of vaccine attributes on public willingness to vaccinate [8, 30] . The framing of 420 vaccination intention questions affects responses [5] , and it could be that the conjoint design source of concern among many older adults. Similarly, we found that CDC endorsements were 433 particularly effective in boosting vaccine acceptance among older individuals. The messenger 434 matters, and our results show that older Americans were even more responsive than younger 435 Americans to endorsements from public health officials vs. endorsements from politicians.

Since subgroups vary considerably in their hesitancy and response to different vaccine 437 attributes, subgroup-specific mobilization cues should be considered. For example, public health 438 authorities should develop outreach strategies that reassure and incentivize individuals in these 439 subgroups [46] , whether through dialogue-based strategies that deploy religious or traditional leaders 440 in these groups, incentive-based strategies that provide food or other goods as an incentive, or 441 reminders by phone or mail [47] . Recognizing these subgroup differences can help identify and then 442 craft those subgroup-specific policies. 

Prevalence of SARS-CoV-2 in Spain (ENE-COVID): a nationwide, population-based seroepidemiological study

COVID-19 vaccines: where we stand and challenges ahead

Europe's new normal: the return of vaccine-preventable diseases

Former CDC Chief Says It's Important to be Honest with People about the Risks of a Vaccine

Confidence and receptivity for covid-19 vaccines: A rapid systematic review

Exploring the reasons behind parental refusal of vaccines

Factors Associated With US Adults' Likelihood of Accepting COVID-19 Vaccination

Can a COVID-19 vaccine live up to Americans' expectations? A conjoint analysis of how vaccine characteristics influence vaccination intentions

COVID-19 vaccine hesitancy in a representative working-age population in France: a survey experiment based on vaccine characteristics

Influence of a COVID-19 vaccine's effectiveness and safety profile on vaccination acceptance

The impact of vaccination and patient characteristics on influenza vaccination uptake of elderly people: A discrete choice experiment

Quantifying preferences around vaccination against frequent, mild disease with risk for vulnerable persons: A discrete choice experiment among French hospital health care workers

Commercial Use of Conjoint Analysis: A Survey

What attributes should be included in a discrete choice experiment related to health technologies? A systematic literature review

How well do discrete choice experiments predict health choices? A systematic review and meta-analysis of external validity

Discrete choice experiments are not conjoint analysis

Causal inference in conjoint analysis: Understanding multidimensional choices via stated preference experiments

Public preferences for vaccination programmes during pandemics caused by pathogens transmitted through respiratory droplets-A discrete choice experiment in four European countries

Acceptance of vaccinations in pandemic outbreaks: A discrete choice experiment

The effects of convenience and quality on the demand for vaccination: Results from a discrete choice experiment

Girls' preferences for HPV vaccination: A discrete choice experiment

Parental preferences for rotavirus vaccination in young children: A discrete choice experiment

Drivers of vaccine decision-making in South Africa: A discrete choice experiment

Validating the demographic, political, psychological, and experimental results obtained from a new source of online survey respondents

Measuring Subgroup Preferences in Conjoint Experiments

Determinants of influenza vaccination among high-risk Black and White adults

Major motives in non-acceptance of A/H1N1 flu vaccination: The weight of rational assessment

Elusive consensus: Polarization in elite communication on the COVID-19 pandemic

Health Behavior, and Policy Attitudes in the Early Stages of the COVID-19 Pandemic. SSRN Electron J 2020

Public Attitudes toward COVID-19 Vaccination: The Role of Vaccine Attributes, Incentives, and Misinformation

Knowledge, attitudes, and beliefs about seasonal influenza and H1N1 vaccinations in a low-income, public health clinic population

Long-standing influenza vaccination policy is in accord with individual self-interest but not with the utilitarian optimum

Association between health care providers' influence on parents who have concerns about vaccine safety and vaccination coverage

Professionally Responsible COVID-19 Vaccination Counseling of Obstetric/Gynecologic Patients

Excess Deaths from COVID-19, Community Bereavement, and Restorative Justice for Communities of Color

The Three Key Hurdles for a Coronavirus Vaccine to Clear

Effective messages in vaccine promotion: A randomized trial

Does correcting myths about the flu vaccine work? An experimental evaluation of the effects of corrective information

Not Just Asking Questions: Effects of Implicit and Explicit Conspiracy Information About Vaccines and Genetic Modification

Republicans, Democrats Move Even Further Apart in Coronavirus Concerns

Socioeconomic status, demographics, beliefs and A(H1N1) vaccine uptake in the United States

Parents Concerned About Vaccine Safety. Differences in Race/Ethnicity and Attitudes

Low acceptability of A/H1N1 pandemic vaccination in french adult population: Did public health policy fuel public dissonance?

The effects of anti-vaccine conspiracy theories on vaccination intentions

Just 50% of Americans plan to get a COVID-19 vaccine. Here's how to win over the rest. Science (80-) 2020

Strategies for addressing vaccine hesitancy-A systematic review

Births: Final Data for 2018 Figure 1. Number and rate of triplet and higher-order multiple births: United States

☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: