key: cord-0931937-6ircaoz4
authors: Freeman, Daniel; Loe, Bao S.; Chadwick, Andrew; Vaccari, Cristian; Waite, Felicity; Rosebrock, Laina; Jenner, Lucy; Petit, Ariane; Lewandowsky, Stephan; Vanderslott, Samantha; Innocenti, Stefania; Larkin, Michael; Giubilini, Alberto; Yu, Ly-Mee; McShane, Helen; Pollard, Andrew J.; Lambe, Sinéad
title: COVID-19 vaccine hesitancy in the UK: the Oxford coronavirus explanations, attitudes, and narratives survey (Oceans) II
date: 2020-12-11
journal: Psychological medicine
DOI: 10.1017/s0033291720005188
sha: 85d63c143741a5a75fd8e45d8fe3b707cc6b37ee
doc_id: 931937
cord_uid: 6ircaoz4

BACKGROUND: Our aim was to estimate provisional willingness to receive a coronavirus 2019 (COVID-19) vaccine, identify predictive socio-demographic factors, and, principally, determine potential causes in order to guide information provision. METHODS: A non-probability online survey was conducted (24th September−17th October 2020) with 5,114 UK adults, quota sampled to match the population for age, gender, ethnicity, income, and region. The Oxford COVID-19 vaccine hesitancy scale assessed intent to take an approved vaccine. Structural equation modelling estimated explanatory factor relationships. RESULTS: 71.7% (n=3,667) were willing to be vaccinated, 16.6% (n=849) were very unsure, and 11.7% (n=598) were strongly hesitant. An excellent model fit (RMSEA=0.05/CFI=0.97/TLI=0.97), explaining 86% of variance in hesitancy, was provided by beliefs about the collective importance, efficacy, side-effects, and speed of development of a COVID-19 vaccine. A second model, with reasonable fit (RMSEA=0.03/CFI=0.93/TLI=0.92), explaining 32% of variance, highlighted two higher-order explanatory factors: ‘excessive mistrust’ (r=0.51), including conspiracy beliefs, negative views of doctors, and need for chaos, and ‘positive healthcare experiences’ (r=−0.48), including supportive doctor interactions and good NHS care. Hesitancy was associated with younger age, female gender, lower income, and ethnicity, but socio-demographic information explained little variance (9.8%). Hesitancy was associated with lower adherence to social distancing guidelines. CONCLUSIONS: COVID-19 vaccine hesitancy is relatively evenly spread across the population. Willingness to take a vaccine is closely bound to recognition of the collective importance. Vaccine public information that highlights prosocial benefits may be especially effective. Factors such as conspiracy beliefs that foster mistrust and erode social cohesion will lower vaccine up-take.

The success of a safe, efficacious coronavirus 2019 (COVID-19) vaccine will depend on uptake; if there are individuals who are reluctant or unwilling to be immunised, uptake will be limited. A low level of intent to be vaccinated can be understood within the concept of hesitancy, the behavioural 'delay in acceptance or refusal of vaccines despite availability of vaccine services' (SAGE Working Group on Vaccine Hesitancy, 2014). There is a continuum from acceptance to refusal of all vaccines, with vaccine hesitancy considered to reside between the two poles, and potential variation within individuals in stance on vaccination for different diseases. Vaccine hesitancy can have effects for both the individual (a greater risk of having the disease) and potentially the community (greater virus transmission). In this paper, by assessing expressed willingness to receive an approved vaccine, we wished to estimate in the UK the level of potential COVID-19 vaccine hesitancy (delay or refusal) and identify groups within the population in which it may be higher. Our primary focus was on explaining vaccine hesitancy at an individual psychological level in order to inform strategies to increase acceptance rates. Beliefs, which are potentially amenable to change, are wellestablished drivers of actions. Therefore, we focussed on identifying a broad cluster of cognitions that may inhibit or facilitate uptake of a COVID-19 vaccine. In essence, we wished to estimate the extent of the potential problem, identify pockets of pronounced hesitancy, and determine the content of key cognitions driving vaccine hesitancy.

In early May 2020, our first Oxford Coronavirus Explanations, Attitudes, and Narratives Survey (OCEANS) was focussed on coronavirus conspiracy beliefs and adherence to government social distancing guidelines (Freeman et al., 2020) . When survey respondents were asked whether they would accept a COVID-19 vaccine, 47.5% responded definitely, 22.1% probably, 18.4% possibly, 7.3% probably not, and 4.8% definitely not. Higher levels of vaccine hesitancy were associated with higher levels of coronavirus conspiracy beliefs. A connection between vaccine hesitancy and conspiracy beliefs has been replicated (Bertin et al., 2020) , and is consistent with work showing that a conspiracy mentality is associated with rejection of vaccine science (Lewandowsky et al., 2013) . In a UK survey conducted in July 2020, 2,237 people were asked how likely they would be to get a coronavirus vaccine (The Policy Institute, 2020); 30% of participants responded that they would be certain to get a vaccine, 23% very likely, 20% fairly likely, 9% not very likely, 3% not at all likely, 4% definitely not, and 11% did not know. These initial studies indicate that approximately 70% of the UK population are likely to accept a COVID-19 vaccine relatively readily, 20% may be ambivalent, and 10% are unlikely to accept a vaccine.

Vaccine hesitancy is typically viewed within the framework of complacency, confidence, and convenience, although it is notable that this is a complex research area in which terms, definitions, and measurement can vary in usage (Bedford et al., 2018; Dudley et al., 2020) . At the heart, an individual's decision as to whether to accept a vaccine can be understood as a weighing up of risk and benefit. Vaccine hesitancy is seen when there is a low perception of need for a vaccination (termed complacency) and concerns over the efficacy and safety (termed low confidence) (e.g. Larson et al., 2015; Crouse Quinn et al., 2019; Karlsson et al., 2019; Mesch & Schwirian, 2019) . A lack of confidence in a vaccine is exacerbated by misunderstanding of how immunisation works (e.g. Zingg & Siergrist, 2012) , distrust of government and healthcare authorities (e.g. Lee et al., 2016) , and newness of a vaccine (e.g. Karafillakisa et al., 2019) . Added to the balance is consideration of how easy it is to access a vaccine (termed convenience) (e.g. Fournet et al., 2018) . Social processes such as norms, altruistic intent, and collective responsibility, have also been highlighted (Brewer et al., 2017; Crouse Quinn, Hilyard et al., 2017; Betsch et al., 2018) .

We wished to build on this framework to understand COVID-19 vaccine hesitancy. In relation to a future UK approved COVID-19 vaccine, we aimed to assess a range of specific beliefs about the importance, benefits, and safety (i.e. complacency and confidence). There are specific aspects of the current epidemic that may affect complacency (e.g. inaccurate severity analogies to seasonal influenza) and confidence (e.g. speed of vaccine development). As the UK has universal free at point of access healthcare provision, we did not focus on convenience, although we did include questions concerning access via general practitioner surgeries, pharmacies, and schools, which are the most likely delivery routes. We also added an additional approach to assessing specific beliefs about the benefits and risks of a COVID-19 vaccine: general trust and mistrust. In essence, willingness to take a vaccine is about trust: that the vaccine is needed, that it will work, and that it is safe. Therefore, unwillingness to take a vaccine will be more likely when excessive mistrust is an individual's default position. If an individual is mistrustful of experts, authority, and institutions, the same tendency will apply to attitudes to a vaccination. Distrust will be more likely when individuals (i) generally feel disrespected and vulnerable to exploitation (marginalised), (ii) view doctors as looking down on them, (iii) hold conspiracy beliefs, (iv) hold particular worldviews (e.g. individualism), and (v) are discontented (e.g. express a 'need for chaos'). Therefore, we conducted a new survey of the UK population, in which we assessed willingness to be vaccinated for COVID-19 (as a marker of vaccine hesitancy) and a wide range of explanatory factors. We hypothesised that vaccine hesitancy would be closely connected to specific confidence and complacency beliefs about a COVID-19 vaccine and also associated with wider makers of mistrust.

An online survey with a new quota sampled UK participant group of 5,114 adults (18+ years old) was conducted from 24th September to the 17th October 2020 via a market research company. The quotas were based on UK Office for National Statistics population estimate data for gender, age, ethnicity, income, and region. Invited respondents did not know the topic of the survey before provisional agreement to complete it. OCEANSII was approved by the University of Oxford Central University Research Ethics Committee. (See online supplementary materials for further survey detail.)

All of the assessment items included in the survey, the details of the factor analyses, the items that comprise each factor, and Cronbach's alphas can be viewed in the online supplementary materials.

Oxford COVID-19 vaccine hesitancy scale There was no existing scale to assess expressed intent to accept a COVID-19 vaccine. We therefore developed a pool of 15 items (see online supplementary materials), with feedback from public involvement groups including representatives from ethnic minority health groups. Item-specific response options (Saris et al., 2010) , coded from 1 to 5, were used. A 'Don't know' option was also provided, which was excluded from scoring. Exploratory and confirmatory factor analyses were used to derive a final seven-item scale from the item pool ( Figure 3) . Higher scores indicate a higher level of vaccine hesitancy.

Oxford COVID-19 vaccine confidence and complacency scale We developed, with public involvement in item generation, 20 items concerning attitudes around vaccine complacency and confidence. Item-specific response options (Saris et al., 2010) , coded from 1 to 5, were used. A 'Don't know' option was also provided, which was excluded from scoring. Exploratory and confirmatory factor analyses identified four factors: collective importance of a COVID-19 vaccine; beliefs that the respondent may get COVID-19 and the vaccine will work; speed of vaccine development; and side effects. Higher scores indicate a greater degree of negative attitudes.

Daniel Freeman et al.

This scale (Shapiro et al., 2018) was included to test convergent validity with the Oxford COVID-19 vaccine hesitancy scale. The nine-item VHS asks parents about their views of childhood vaccines, and we re-worded the scale where appropriate to make it suitable for all survey respondents. A one-factor solution emerged. Each item was rated on a 1 (strongly disagree) to 5 (strongly agree) scale. Higher scores indicate greater hesitancy.

This questionnaire presents individuals with nine statements about vaccines that they are asked to rate as correct, incorrect, or do not know (Zingg & Siergrist, 2012) . Factor analysis identified two factors: general knowledge about vaccines and knowledge about childhood vaccines. Incorrect or do not know are scored as zero and correct is scored as one. Therefore, higher scores indicate better knowledge of vaccines.

We developed 11 new items about inter-personal disrespect from doctors and 5 items about mistrust of vaccine developers. Each item was rated on a 1 (disagree completely) to 4 (agree completely) scale, with a 'don't know' option that was not scored. Factor analysis indicated three factors: interpersonal disrespect by doctors; respect from doctors and negative views of vaccine developers. Higher scores indicated greater disrespect from doctors, less respect from doctors, and greater negative views of vaccine developers.

Nineteen items about doctors and medicine are rated on a scale from strongly disagree (1) to strongly agree (6) (Marteau, 1990) . Confirmatory factor analysis supported a four-factor model: positive attitude to doctors; negative attitude to doctors; positive attitude to medicine; and negative attitude to medicine. Higher scores indicate greater positive attitudes to doctors or medicine and greater negative attitudes to doctors or medicine.

The MacArthur scale of subjective social status Two separate items assess where people view themselves on a social ladder relative to other people in the UK or relative to other people in their community (Adler et al., 2000) . Each item is rated 0−10. Higher scores were coded to indicate lower subjective social status.

Brief core schema scales Twelve items assess beliefs about self, rated on a scale from do not believe (0) to believe it totally (4) (Fowler et al., 2006) . Confirmatory factor analysis indicated a negative self-factor and a positive selffactor. Higher scores reflect greater endorsement of items.

General practice assessment questionnaire-R2 Eight items were used from this questionnaire, assessing how the person has been treated by their GP (Roland et al., 2013) . Each item is rated on a scale from very good (1) to very poor (5). Factor analysis indicated a one-factor solution concerning positive GP experiences.

Higher scores indicate fewer positive GP experiences.

We developed eight items assessing positive and negative experiences of the NHS. Each item is rated on a three-point scale, from No (1) to Yes (3). Factor analysis identified two factors: positive NHS experiences; and negative NHS experiences. Higher scores indicate fewer positive NHS experiences, and greater negative NHS experiences.

From the first OCEANS study, we developed a seven-item general coronavirus conspiracy scale and a 14-item-specific coronavirus conspiracy scale (Freeman et al., 2020) . Each item was rated on a scale from do not agree (1) to agree completely (5). We also added a 'Don't know' response option, which was not included in the score. Higher scores indicate greater endorsement of coronavirus conspiracy beliefs. We only used the general conspiracy scale for the explanatory modelling. Confirmatory factor analysis confirmed a single factor.

This seven-item questionnaire asks participants how much they agree on a seven-point scale with vaccine conspiracy statements (Shapiro et al., 2016) . A one-factor CFA model indicated a good fit for the seven items. Higher scores indicate greater endorsement of conspiracy statements.

Individuals rate on a scale from almost everyday (1) to never (6) how often nine negative experiences happen to them (Williams et al., 1997) . Factor analysis indicated two factors: others disrespectful and others react negatively. Higher scores indicate fewer negative experiences.

Anger is assessed with five items, rated on a scale from 1 (none) to 5 (all of the time) (Forbes et al., 2014) . CFA resulted in a single factor with good model fit. Higher scores indicate higher levels of anger.

Eleven items assessing 'need for chaos', understood as a desire to bring down the established political order in order to increase one's own social status, are rated on a scale from strongly disagree (1) to strongly agree (7) (Petersen et al., 2020) . A one-factor CFA resulted in an excellent fit. Higher scores indicate a higher level of need for chaos.

Lifestyle and economic/government liberty Seven items assessing libertarian worldviews were rated on a scale from strongly disagree (1) to strongly agree (6) (Iyer et al., 2012) . Factor analysis identified a one-factor solution. Higher scores indicate greater libertarian beliefs.

Five items assessed populist views, each rated on a scale from strongly disagree (1) to strongly agree (6) (Akkerman et al., 2014) . Factor analysis identified a one-factor model. Higher scores indicate a greater level of populist beliefs.

Perceived religious influence on health behaviour and illness as punishment by God for sin This questionnaire contains two separate scales concerning the influence of religion on an individual's health views (Holt et al., 2009) . Each of the 15 items is rated on a scale from strongly disagree (1) to strongly agree (4). Factor analysis identified a religious influence on health behaviour factor and an illness as punishment for sin factor. Higher scores indicate higher influence of religion on health behaviour and greater belief in illness as punishment.

Following of UK government coronavirus guidance Participants were asked to rate how often they followed nine key aspects of government guidance at the time on a five-point scale.

Higher scores indicate greater adherence to guidance. Participants were also asked to rate on the same scale two general items assessing how much they follow the guidance. Participants were also asked to rate on a five-point scale how likely it was that they would accept a COVID-19 diagnostic test and a COVID-19 antibody test. Higher scores on these two items indicate less likelihood of accepting the tests.

Structural equation modelling (SEM) was used to assess two explanatory models. The first was between beliefs about a COVID-19 vaccine and hesitancy, and the second was between the broader psychological constructs of mistrust and vaccine hesitancy. SEM comprises a confirmatory measurement model, also known as the confirmation factor analysis (CFA), which estimates the relations among latent constructs and their observed indicators, and a structural model, which estimates the relations among constructs (Kline, 2015) . The lavaan R package was used to conduct the SEM analysis (Rosseel, 2012) . We utilised the full information maximum likelihood (FIML) estimation procedure with robust (Huber−White) standard errors and a scaled test statistic to estimate the SEM models. In assessing the relations among constructs in the structural model, we employed a backward elimination procedure to reduce multicollinearity issues (Hocking, 1976; Grewal et al., 2004) . Full details of the SEM methods can be seen in the online supplementary materials.

A summary of the socio-demographic characteristics of the participants is provided in Table 1 .

The mean score on the Oxford COVID-19 vaccine hesitancy scale was 13.6 (SD=7.3). The factor score for the Oxford COVID-19 vaccine hesitancy scale was significantly associated with the factor score for the vaccine hesitancy scale, r=0.47, p < 0.001. The frequencies of endorsement of each of the Oxford COVID-19 vaccine hesitancy scale items are summarised in Table 2 . In all, 3892 (76.1%) participants did not endorse any clear vaccine hesitancy response (a response rating of 4 or 5) on any of the seven items, while 1222 (23.9%) endorsed at least one of the items with a hesitant response. Further, 596 (11.7%) participants endorsed four or more of the seven items (i.e. over half) with a clear vaccine hesitancy response (a rating of 4 or 5). This group can be considered strongly vaccine hesitant. 196 (3.8%) participants endorsed all seven items with a clear vaccine hesitancy response, and this can be considered a very extreme group. 4193 (82.0%) participants endorsed at least one of the seven items with a clear positive response (a rating of 1 or 2), while 921 (18.0%) did not rate any of the items with a clear positive response. 3669 (71.7%) participants endorsed four or more of the seven items with a clear positive response (a rating of 1 or 2). This group can be considered clearly positive about a COVID-19 vaccine. 2823 (55.2%) participants endorsed all seven items with a clear positive response. 1842 (36.0%) participants endorsed at least one of the items with the middle response (a rating of 3), while 3272 (64.0%) did not endorse any middle responses. 466 (9.1%) participants endorsed four or more of the seven items with the middle response. This group can be considered as consistently very unsure about a COVID-19 vaccine. 737 (14.4%) participants endorsed at least one of the items with a do not know response, while 4377 (85.6%) did not use this response option. 106 (2.1%) participants endorsed four or more of the seven items with a do not know response. This group can also be considered as consistently very unsure.

There were also individuals who showed two different sets of three consistent responses (but no set of four consistent responses). 34 (0.7%) participants endorsed extreme hesitancy responses (a response rating of 4 or 5) for three of the seven items and three undecided middle options, 47 (0.9%) endorsed three clear positive responses (a rating of 1 or 2) and three undecided middle options, 19 (0.4%) endorsed three undecided middle options and three don't know options, and 11 (0.2%) endorsed three clear positive and three clear negative views.

In summary, 71.7% of the population responded in a consistently positive way towards taking a COVID-19 vaccine, 16.6% were very unsure about taking a COVID-19 vaccine, and 11.7% were strongly hesitant.

The associations of vaccine hesitancy and socio-demographic factors, tested in individual linear regressions, are summarised in 

Higher levels of vaccine hesitancy are associated with less following of all guidelines and less likelihood of taking a diagnostic or antibody test (all p-values<0.001) (see Table 4 ).

Tables 5 and 6 provide a summary of prevalence rates for misinformation and conspiracy beliefs about the pandemic and vaccinations. It can be seen that there is appreciable endorsement of coronavirus conspiracy beliefs and vaccination conspiracy beliefs. 

The simple associations of the factor scores for vaccine hesitancy and the explanatory factors are shown in Table 6 . Every factor score was significantly associated with vaccine hesitancy (p < 0.001) ( Table 7) .

There were high correlations between the four factors derived for beliefs about a COVID-19 vaccine (assessed by the Oxford COVID-19 vaccine confidence and complacency scale) (see online supplementary materials, Table S1 ). Therefore, the four factors were merged into a higher-order construct labelled as 'Beliefs about a COVID-19 vaccine'. The CFA model with a Figure 1 (also see online supplementary materials, Table S2 ). Table S4 and Fig. S1 . The backward elimination procedure reduced the number of predictors from 14 to 10. Direct effects of four factors (religious influence on health behaviour, others reacting negatively, libertarian beliefs, and populist beliefs) were fixed to zero in the structural model. The SEM model after this backward elimination procedure indicated reasonable fit (FIML, χ2(5697, N=5114)=28926.01, p < 0.001, RMSEA=0.03; SRMR=0.07; CFI=0.93; TLI=0.93). Since the reduced SEM model is considered nested from the baseline model where the relations between the constructs were freely estimated, an analysis of variance (ANOVA) was conducted to test for the best-fitting model from all the models, which indicated that the final reduced model (SEM model 5) was preferable.

Further review of the beta estimates suggested that several predictors may be acting as suppressors (see online supplementary materials). Thus, the direct effects of five predictors (negative self, anger, illness as punishment, others disrespectful, positive self) were fixed to zero in the structural model. The variance inflation factor and tolerance levels of all remaining variables were less than 5 and tolerance values greater than 0.1, indicating an absence of multicollinearity (see online supplementary materials, Table S7 ). The final SEM model indicated reasonable fit to the data (FIML, χ2(5702, N=5114)=29308.76, p < 0.001, RMSEA=0.03; SRMR=0.070; CFI=0.93; TLI=0.92) and the variance explained by the structural model was 32%. Online supplementary materials, Table S8 summarises the parameter estimates of the structural model. The reduced SEM model (i.e. SEM model 5) is shown in Figure 2 . The higher-order factors had the strongest associations with vaccine hesitancy. The higher-order factor of positive experiences of healthcare was associated with lower levels of COVID-19 vaccine hesitancy; the higher-order factor of mistrust was associated with higher levels of COVID-19 hesitancy. A lower subjective sense of self in relation to others (the UK or community) was also associated with greater vaccine hesitancy.

The substantial majority view in the UK public that vaccines are safe, effective, and important remains intact, but appears vulnerable given the presence of significant proportions of the population either very doubtful or rejecting of a COVID-19 vaccine. The danger is of a tipping point being approached, whereby mistrust of vaccines becomes mainstream, with negative consequences for individuals, healthcare services, and potentially herd immunity. With the caution that this was an online nonprobability survey, the majority of the populationaround 72% − are likely to accept reasonably readily a COVID-19 vaccine. It is possible that this majority will remain robust in the months ahead, but this is likely to be tested. The remainder of the population, a sizeable minority, are clearly very ambivalent (17%) about a COVID-19 vaccine or are unlikely to take it (12%). Around one in 20 people would even describe themselves as explicitly anti-vaccination for COVID-19. When the time comes, the behaviour of those who are ambivalent about a vaccination will be a defining factor in the success of implementation of an approved vaccine. Socio-demographics do not explain vaccine hesitancy to any helpful degree. Doubt is spread relatively evenly in the population − it is not circumscribed to specific groups − which may be another indicator of the issue being at a tipping point. The understanding of the causes of vaccine hesitancy has never been more important.

There are a set of beliefs, tightly bound to willingness to take a COVID-19 vaccine, that are plausible drivers of vaccine up-take. Fused with acceptance of a vaccine are beliefs about the collective importance: that a vaccine will save lives, help the community, and that it will be dangerous if many people do not get vaccinated. This chimes with evidence that emphasising collective − rather than personalresponsibility may lead to greater change in individuals' behaviour (e.g. Obradovich & Guenther, 2016). Pro-social behaviour may be especially self-rewarding and therefore affect behaviour (e.g. Aknin et al., 2012) . Betsch and colleagues (2015) in an online experiment with fictitious disease scenarios found that explaining the social benefits of herd immunity can increase intent to be vaccinated. In the OCEAN survey, beliefs about whether a vaccine will provide freedom or restriction were part of the collective importance factor too. There were three other key types of beliefs about a COVID-19 vaccine: if a respondent thought it likely that they would be infected and the vaccine would work; whether speed of development of the vaccine would affect safety and efficacy; and the degree to which receiving the vaccine may be physically unpleasant and that the recipient would feel experimented on. All these findings are highly consistent with the framing in the vaccine hesitancy literature of the importance of complacency and confidence in vaccine decision-making.

We also set out to determine the more distal individual factors that may affect vaccine up-take. The hypothesis was that mistrust may skew vaccine decision making. The modelling indicated a key higher-order factor of excessive mistrust that was associated with vaccine hesitancy. This factor included coronavirus conspiracy 

beliefs, which were at a troubling prevalence comparable to our first OCEAN survey in May (Freeman et al., 2020) . We also note that a significant minority hold general vaccine conspiracy beliefs, while an even higher proportion is neutral as to whether vaccine conspiracy beliefs are true or false. These fringe beliefs do appear to have become mainstream. Negative perceptions of doctors ('They do not really care about me'), vaccine developers ('They just want to make money'), and healthcare experiences ('The NHS treats my community badly') feed into the distrust. There also appears to be a contribution of anger: a need for chaos ('I think society should be burned to the ground'), anger directed at societal structures, contributed to the higher-order factor. Individuals who were vaccine hesitant also viewed themselves as lower on the social ladder, indicating a degree of perception of vulnerability with regards to other people. Against this mistrust and vulnerability, perceived positive healthcare experiences were connected to lower vaccine hesitancy. Positive GP experiences ('My GP is polite and considerate'), views of doctors ('They have my best interests at heart'), medicine ('The improved health of the nation is due to effective medicine'), and NHS experiences ('Staff have gone out of their way to help') were connected with greater enthusiasm about a COVID-19 vaccine. In sum, factors that erode trust in healthcare and society are likely to be pernicious for vaccine up-take but positive experiences of institutions, those that rebuild trust, are likely to maintain the consensus of the importance of being vaccinated. The overall immediate implications of our explanatory work are the potential importance of emphasising in public health messaging the prosocial benefits of a COVID-19 vaccine, as well as the necessity to be transparent about safety and efficacy. Public health communication may need to be carefully attuned to the different kinds of collective identities and benefits in order to resonate with pro-social motives. In this way, such messaging may not only help consolidate the majority willing to be vaccinated but perhaps increase the willingness of those who are hesitant. Careful testing and refining of messaging across the spectrum of hesitancy will be needed. The survey findings also indicate that materials may benefit from highlighting the many positive contributions that NHS staff make. There is an urgent need to counter misinformation, ideally by 'prebunking' or inoculation (Wong, 2016; van der Linden et al., 2017) , and provide strong presentation of accurate information. The findings also reiterate the longer-term work needed to rebuild trust in experts and institutions. However there are limitations to the survey. Foremost, we used a non-probability online quota sampling method, which will have introduced bias to who was approached to take part. We do know that, taken as a whole, the respondents in this survey were broadly representative of the adult general population on a number of basic demographic features (although, for example, levels of higher education were slightly high) but not that individual respondents were representative of the general population. This means that prevalence estimates especially must be treated with caution, as well as the identification of demographic predictors. It is plausible that the survey method will have introduced less bias into the explanatory parts of the study (Pasek, 2015) . A further caution concerns our primary measure: we do not know the extent to which expressed intent to take a vaccine is associated with actual behaviour. The key limitation for the conclusions concerning the explanatory variables is that the survey is cross-sectional. We do not know whether the beliefs, attitudes, and perceived experiences actually cause willingness to take a COVID-19 vaccine. Our plan is to use detailed qualitative interviewing, guided by the results of the survey, to deepen our understanding of vaccine hesitancy, and then conduct experimental tests to assess change and causation.

Supplementary material. The supplementary material for this article can be found at https://doi.org/10.1017/S0033291720005188

Relationship of subjective and objective social status with psychological and physiological functioning: Preliminary data in healthy

How populist are the people? Measuring populist attitudes in voters

Giving leads to happiness in young children

Vaccine hesitancy, refusal and access barriers: The need for clarity in terminology

Conspiracy beliefs, rejection of vaccination, and support for hydroxychloroquine: A conceptual replication-extension in the COVID-19 pandemic context

On the benefits of explaining herd immunity in vaccine advocacy

Beyond confidence: Development of a measure assessing the 5C psychological antecedents of vaccination

Increasing vaccination: Putting psychological science into action

The influence of social norms on flu vaccination among African American and White adults

Measuring vaccine hesitancy, confidence, trust and flu vaccine uptake: Results of a national survey of White and African American adults

Words matter: Vaccine hesitancy, vaccine demand, vaccine confidence, herd immunity and mandatory vaccination

Utility of the dimensions of anger reactions-5 (DAR-5) scale as a brief anger measure

Under-vaccinated groups in Europe and their beliefs, attitudes and reasons for non-vaccination; two systematic reviews

The Brief Core Schema Scales (BCSS): Psychometric properties and associations with paranoia and grandiosity in non-clinical and psychosis samples

Coronavirus conspiracy beliefs, mistrust, and compliance with government guidelines in England

Multicollinearity and measurement error in structural equation models: Implications for theory testing

A biometrics invited paper. The analysis and selection of variables in linear regression

Development and validation of instruments to assess potential religion -health mechanisms in an African American population

Understanding libertarian morality: The psychological dispositions of self-identified libertarians

HPV vaccination in a context of public mistrust and uncertainty: a systematic literature review of determinants of HPV vaccine hesitancy in Europe

The association between vaccination confidence, vaccination behavior, and willingness to recommend vaccines among Finnish healthcare workers

Principles and practice of structural equation modeling

Measuring vaccine confidence: Introducing a global vaccine confidence index

Hurdles to herd immunity: Distrust of government and vaccine refusal in the US

The role of conspiracist ideation and worldviews in predicting rejection of science

Attitudes to doctors and medicine: The preliminary development of a new scale

Vaccination hesitancy: fear, trust, and exposure expectancy of an Ebola outbreak

When will nonprobability surveys mirror probability surveys? Considering types of inference and weighting strategies as criteria for correspondence

The "Need for Chaos" and Motivations to Share Hostile Political Rumors

Coronavirus uncertainties: vaccines, symptoms and contested claims

GPAQ-R: development and psychometric properties of a version of the General Practice Assessment Questionnaire for use for revalidation by general practitioners in the UK

Lavaan: An R package for structural equation modeling and more. Version 0.5-12 (BETA)

Comparing questions with agree/disagree response options to questions with itemspecific response options

Validation of the vaccine conspiracy beliefs scale

The vaccine hesitancy scale: Psychometric properties and validation

Inoculating against misinformation

Racial differences in physical and mental health: Socio-economic status, stress and discrimination

Vaccinations are safe and effective": inoculating positive HPV vaccine attitudes against antivaccination messages

Measuring people's knowledge about vaccination: Developing a one-dimensional scale

Acknowledgements. The project was funded by a grant from the University of Oxford COVID-19 Research Response Fund (Project Reference: 0009519). It was also supported by an award from the National Institute of Health Research (NIHR) invention for innovation (i4i) programme (Project II-C7-0117-20001) and the NIHR Oxford Health Biomedical Research Centre (BRC-1215-20005). DF is supported by an NIHR Research Professorship (NIHR-RP-2014-05-003). The views expressed are those of the authors and not necessarily those of the National Health Service, NIHR, or the Department of Health. Alberto Giubilini is supported by an UKRI/ AHRC grant (AH/V006819/1).Conflicts of Interest. AJP is Chair of UK Dept. Health and Social Care's (DHSC) Joint Committee on Vaccination & Immunisation (JCVI), but does not participate in discussions on COVID19 vaccines, and is a member of the WHO's SAGE. The views expressed in this paper do not necessarily represent the views of DHSC, JCVI, or WHO. Oxford University has entered into a partnership with Astra Zeneca for the development of a coronavirus vaccine.