Datasets:

---

LLMDH

/

OpenScience

like 0

Follow

LLM - Digital Humanities 16

ERROR: type should be string, got "https://doi.org/10.1016/j.jvacx.2023.100299\n2590-1362/\u0001 2023 The Author(s). Published by Elsevier Ltd.\nThis is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). ⇑Corresponding author.\nE-mail address: w.j.jennings@soton.ac.uk (W. Jennings). a r t i c l e\ni n f o Article history:\nReceived 21 November 2022\nReceived in revised form 31 March 2023\nAccepted 1 April 2023\nAvailable online 6 April 2023\nKeywords:\nCOVID-19\nVaccination\nTrust\nMisinformation Article history:\nReceived 21 November 2022\nReceived in revised form 31 March 2023\nAccepted 1 April 2023\nAvailable online 6 April 2023 Previous studies of vaccine hesitancy in the context of COVID-19 have reported mixed results in terms of\nthe role played by political and institutional trust. This study addresses this ambiguity with a global anal-\nysis of the relationship between trust and vaccine hesitancy, disentangling the effects of generalized trust\norientations, trust in specific institutions and conspiracy mentality. It first draws upon a cross-national\nsurvey of 113 countries to demonstrate that trust in government is a predictor of vaccine hesitancy across\nglobal regions. It further draws on original surveys fielded in seven countries (France, Germany, Spain,\nArgentina, Croatia, Brazil, India), which deploy a diverse range of measures, to disentangle the\nindividual-level predictors of vaccine hesitancy. Our findings confirm the robust effects of trust in gov-\nernment across countries, but when including other trust measures in the same models, the most robust\neffects are those of trust in health institutions and conspiracy mentality. Weaker associations are\nobserved for right-wing ideology and online political engagement, while the consumption of traditional\nmedia tends to predict the willingness of individuals to be vaccinated. Keywords:\nCOVID-19\nVaccination\nTrust\nMisinformation Keywords:\nCOVID-19\nVaccination\nTrust\nMisinformation \u0001 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license\n(http://creativecommons.org/licenses/by/4.0/). The global battle against COVID-19 has drawn attention to the\nperceived critical importance of political trust for public health\ninterventions designed to mitigate the spread of infectious disease\n[3,4,8,9]. This perception seems well-founded: citizens’ trust in\nauthorities and social institutions speaks to whether they believe\nthat they are actually looking out for their interests and has thus\nlong been understood as fundamental to cooperation between gov-\nernment and the public [8,13]. Indeed, trust has been found to be\nimportantly related to countries’ general success in combating\nthe pandemic on the aggregate level [3,30] and with citizens’ com-\npliance with public health guidelines on the individual level\n[12,29,27,29]. they report non-significant effects for trust in government in all\nbut one of their samples. On the other hand, Lazarus et al. Trust and vaccine hesitancy during the COVID-19 pandemic: A cross-\nnational analysis Will Jennings a,⇑, Viktor Valgarðsson a, Lawrence McKay a, Gerry Stoker a, Eduardo Mello b,\nHasan Muhammad Baniamin c a University of Southampton, UK\nb Fundação Getulio Vargas, Brazil\nc North South University, Bangladesh Vaccine: X 14 (2023) 100299 Vaccine: X 14 (2023) 100299 https://doi.org/10.1016/j.jvacx.2023.100299\n2590-1362/\u0001 2023 The Author(s). Published by Elsevier Ltd.\nThis is an open access article under the CC BY license (http://c Original survey data For the second part of the analysis, we use data from online\nnationally representative surveys in seven major countries –\nFrance, Germany, Spain, Croatia, Argentina, Brazil and India –\nwhich we conducted between September and December 2021. Each of our surveys apply quotas on multiple demographic factors,\nwhich include age, gender, and region for all surveys (and addition-\nally social class for NetQuest surveys and education and past vote\nfor YouGov surveys). Full details of the survey fieldwork, sample\nsize, survey company and mode/design are reported in Table 1. The dependent variable is vaccine willingness, the inverse of vac-\ncine hesitancy. At the time of fieldwork, the distribution of free\nvaccines was becoming a reality rather than (as with the data in\nour previous analysis) a theoretical future possibility, even in\ncountries where the rate of doses had been stalled by poor avail-\nability (such as in the case of India, which may be reflected in its\nlow level of first doses and full vaccinations received), so the mea-\nsures reflect actual reported prior behaviour, rather than reported\nvaccine intentions. In the survey, we asked respondents the follow-\ning pair of questions, coding as vaccine ‘willing’ (1) those respon-\ndents who answered that they had received a dose or that they\nwould if offered. Those who had declined the dose offered, or\nwould not get a future dose if offered, are coded as unwilling\n(0).2 The questions were formulated as follows: p\nOur empirical findings confirm that political trust is a consistent\npredictor of vaccine hesitancy across the world, one that is robust\nto the inclusion of multiple controls and holds in over 90 percent of\nthe more than 100 countries covered here. In our more detailed\nanalysis of the role of trust, using original survey data, we find that\ngeneralised distrust in government is positively related to vaccine\nhesitancy and generalised trust is negatively related to it, whereas\nthe effects of political mistrust and social trust are non-significant,\nbut conspiracy mentality has the strongest effect by some margin. When separating political trust into trust in different institutions,\nwe find, in line with previous studies, that trust in health institu-\ntions stands out as by far the most robust, but this is fairly strongly\nrelated to confidence in political institutions, which has a signifi-\ncant effect when the former is not included in the same model. Methods lumped together in statistical models, without great clarity about\nthe reasons behind the choice of variables, the correlations\nbetween them or how the measures overlap. In this study, we\nmove\nbeyond\nthese\nprevious\napproaches\nby\ndistinguishing\nbetween generalized trust orientations and more specific confi-\ndence in particular institutions, by distinguishing between political\ntrust, mistrust, distrust and conspiracy mentality as distinct types of\nattitudes within the ‘‘trust” family, and by explicitly interrogating\nthe inter-relationships between different types of trust and vaccine\nacceptance. Finally, our original surveys contain a broader range of\nrelevant measures of personal, attitudinal and behavioural factors\nthan most others, allowing us to compare the effects of trust with\nfactors such as media use, personality traits and demographics to a\ngreater extent than prior studies. Wellcome global monitor data For the global analysis, we use data from The Wellcome Global\nMonitor (WGM) included in the Gallup World Poll. Between Octo-\nber 2020 and January 2021 inclusive, telephone surveys were com-\npleted by nearly 120,000 respondents in 113 countries, which\ncollectively represent 90 % of the global population aged 15 +. The surveys concentrated on themes related to COVID-19, includ-\ning its economic impacts, and crucially featured questions on both\nvaccine uptake intentions and on trust. The measures used here\nasked respondents: (a) how much they trusted the national gov-\nernment in their country, and (b) whether they would agree to\nbe vaccinated if a vaccine to prevent coronavirus ‘‘was available\nright now at no cost”.1 Our approach in this paper thus seeks to advance our under-\nstanding of the role(s) of trust in vaccine acceptance by addressing\nthese limitations of prior studies. Firstly, we undertake a global\nanalysis of the degree to which trust predicts vaccine hesitancy,\nbased on high quality survey data from over 100 countries. This\nenables us to determine how general the link between trust and\nvaccine hesitancy is, covering a broader scope of countries than\nany prior study on the individual level. Secondly, we report results\nfrom our original surveys conducted in seven countries, which\nincluded an extensive set of measures which allow us to disentan-\ngle the relationship between different types of trust and vaccine\nacceptance and compare with the role of related factors such as\npersonality traits, cognition and political ideology. 1 The full wording for these questions is included in Online Appendix A.\n2 Those who give ‘prefer not to say’ responses at Q13, or respond ‘don’t know’ at\nQ14, are dropped from the analysis. Methods In terms of\ntrust, we include measures of generalized political trust, mistrust\nand distrust, reflecting insights from recent studies that suggest\nthat lack of trust may either denote a type of cautious scepticism\n(mistrust) or settled dissatisfaction (distrust) – and that the two\nmay have different implications for vaccine hesitancy, in that we\nwould expect distrust but not necessarily mistrust to drive it\n([2]; [28]; [12]). In addition, we include a scale measuring ‘conspir-\nacy mentality’ (developed by 5 and more conventional measures of\ntrust in representative institutions, health institutions and the\nmedia as well as a measure of social trust, separating these in our\nanalysis and comparing their relative effects. This offers a more\ncomprehensive test of the precise ways that trust influences beha-\nviour and its relative importance. The WGM also included demographic information on respon-\ndents’ age, sex, education, income and employment status. In\nterms of sample size, 98,166 respondents in 105 countries are valid\non the vaccine and trust measures, with 94,620 in 102 countries\nvalid once control variables are added in a regression model. a r t i c l e\ni n f o In this study, we\nmove\nbeyond\nthese\nprevious\napproaches\nby\ndistinguishing\nbetween generalized trust orientations and more specific confi-\ndence in particular institutions, by distinguishing between political\ntrust, mistrust, distrust and conspiracy mentality as distinct types of\nattitudes within the ‘‘trust” family, and by explicitly interrogating\nthe inter-relationships between different types of trust and vaccine\nacceptance. Finally, our original surveys contain a broader range of\nrelevant measures of personal, attitudinal and behavioural factors\nthan most others, allowing us to compare the effects of trust with\nfactors such as media use, personality traits and demographics to a\ngreater extent than prior studies. a r t i c l e\ni n f o [16] find\na significant effect for trust in government in their 19-country\nstudy and Lindholt et al. [17] find in their 8-country study that\ntrust in government has an effect in addition to trust in scientists\nand health authorities, although that effect is only marginally sig-\nnificant in their full model. Finally, Allington et al. [1] find in their\nstudy of the UK that trust in government had a significant effect\nbut a much smaller one than trust in scientists and medical profes-\nsionals. Other studies on the topic have been limited by their use of\nconvenience samples, although these tend to find significant\neffects for political trust [13,25,25]. However, when it comes to the particular topic of the role of\npolitical trust in shaping vaccine acceptance, studies have reported\nmixed results. There appears to be no doubt that some kind of trust\nmatters, but while Rozek et al. [21] report in their 17-country\nstudy that trust in health institutions matter, they report no effects\nfor trust in political leaders. Similarly, Kerr et al. [14] find in their\n25-sample study that trust in experts plays an important role, but In summary, of the four large, cross-national studies published\nto date on the topic, two report significant effects for political trust\non vaccine acceptance, whereas two report no significant effects. In\nthis study, we directly address this ambiguity in the existing liter-\nature, using secondary survey data from over 100 countries and\noriginal survey data from seven countries, with the latter explicitly\ndesigned to disentangle the role of different types of trust in seven\ncountries. The broad concept of political trust is typically not mea-\nsured in a systematic or theoretically driven way in prior studies;\ninstead, different measures of trust in various entities tend to be https://doi.org/10.1016/j.jvacx.2023.100299\n2590-1362/\u0001 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Vaccine: X 14 (2023) 100299 Vaccine: X 14 (2023) 100299 W. Jennings, V. Valgarðsson, L. McKay et al. lumped together in statistical models, without great clarity about\nthe reasons behind the choice of variables, the correlations\nbetween them or how the measures overlap. Original survey data Indeed, trust in health institutions, conspiracy mentality and age\nare the only predictors whose significant effects persist even after\nincluding all 29 variables in our analysis in the same model. Con-\nsuming news in traditional media is also a strong (negative) pre-\ndictor of hesitancy and those who identify on the right in politics\nare more likely to be hesitant, whereas the effects of other factors\nare weaker and less consistent. By combining breadth and depth of\nanalysis, we highlight the consistency of trust as a predictor of vac-\ncine hesitancy and shed light on the dynamic role that different\ntypes of trust play, with conspiracy mentality and trust in health\ninstitutions standing out as the most potent predictors. Q13. Which of these applies to you? (1) I have received one or two doses of a COVID-19 vaccine\n(2) I have been offered a vaccine against COVID-19, but declined the\noffer to be vaccinated\n(3) I have not yet been offered a vaccine against COVID-19\n(4) Prefer not to say\nQ14. To what extent, if at all, do you agree, or disagree, with the\nfollowing statements? If I was offered a vaccine for COVID-19, I would get it [ASK IF\nQ13 = 3]\n(1) Strongly agree\n(2) Tend to agree\n(3) Neither agree nor disagree\n(4) Tend to disagree (2) I have been offered a vaccine against COVID-19, but declined the\noffer to be vaccinated (3) Neither agree nor disagree (4) Tend to disagree 2 2 Vaccine: X 14 (2023) 100299 W. Jennings, V. Valgarðsson, L. McKay et al. Table 1\nOriginal survey details. Source: Our World In Data [18]. Country\nFieldwork\nSample\nCompany\nDesign\nVaccination rate*\nFrance\n23.09.2021 – 20.11.2021\n1,548\nYouGov\nOnline, nationally representative\n65 % (75 %)\nGermany\n14.10.2021 – 23.11.2021\n1,558\nYouGov\nOnline, nationally representative\n65 % (70 %)\nSpain\n23.09.2021 – 26.09.2021\n1,022\nYouGov\nOnline, nationally representative\n77 % (79 %)\nCroatia\n29.09.2021 – 06.10.2021\n1,017\nYouGov\nOnline, nationally representative\n40 % (46 %)\nArgentina\n29.09.2021 – 08.10.2021\n1,113\nYouGov\nOnline, nationally representative\n50 % (66 %)\nBrazil\n06.12.2021 – 28.12.2021\n2,123\nNetQuest\nOnline, nationally representative\n64 % (77 %)\nIndia\n22.09.2021 – 30.09.2021\n1,040\nNetQuest\nOnline, nationally representative (of the literate population)\n15 % (44 %)\nTotal\n22.09.2021 – 28.11.2021\n9,421\nNote: France includes a top-up sample of 498 respondents 19.11.2021–20.11.2021; Germany includes top-up sample of 534 respondents 19.11.2021–20.11.2021. 3 In Brazil, this item specifically referred to the publicly funded healthcare system,\n‘‘Sistema Único de Saúde (SUS)”. Global analysis: Wellcome Global Monitor data We begin with an analysis of the relationship between political\ntrust and vaccine hesitancy across the globe. Is this robust, and\ndoes it consistently emerge globally? To answer this, we turn to\nthe Wellcome Global Monitor included in the Gallup World Poll,\nwhich includes measures of trust in government and willingness\nto be vaccinated, along with demographic measures. Further infor-\nmation about this data source and the variables used is presented\nin the methods section below and in Online Appendix B. We first\npresent the country-level aggregate relationships between trust\nin\ngovernment\nand\nvaccine\nintention\non\nthe\ncountry-level\n(Fig. 1).4 The y-axis represents the percentage of respondents in each\ncountry who said that they would agree to be vaccinated against the\ncoronavirus if a vaccine was available at no cost,5 while the x-axis\nrepresents the percentage of respondents who said they trusted\nthe national government in their country ‘a lot’ or had ‘some’ trust\nin it.6 The aggregate correlation between these two variables is pos- Following previous work, we took the first three statements (1–\n3) to indicate trust, the next two (4–5) to reflect mistrust, and the\nlast three (6–8) to measure distrust. For each set of statements, we\ncreated scales for each respondent taking the mean level of agree-\nment with the statements (on the Likert scale) before rescaling the\nvalues to 0–1. Cronbach’s alpha scores vary between 0.60 (mis-\ntrust) to 0.84 (distrust) and 0.90 (trust), all of which pass the 0.6\nthreshold established as acceptable by Nunally and Bernstein [20]. Following previous work, we took the first three statements (1–\n3) to indicate trust, the next two (4–5) to reflect mistrust, and the\nlast three (6–8) to measure distrust. For each set of statements, we\ncreated scales for each respondent taking the mean level of agree-\nment with the statements (on the Likert scale) before rescaling the\nvalues to 0–1. Cronbach’s alpha scores vary between 0.60 (mis-\ntrust) to 0.84 (distrust) and 0.90 (trust), all of which pass the 0.6\nthreshold established as acceptable by Nunally and Bernstein [20]. Turning to our other independent variables, we provide more\ndetail in Online Appendix B, and only summarise our measures\nbriefly here. In terms of demographics, we include age, gender,\nand a dummy for university level education. Our models include\nleft–right ideological self-placement, measured on a 0–10 scale. (5) Strongly disagree\n(6) Don’t know Turning to our independent variables, the survey includes mul-\ntiple measures of trust. First, it includes conventional measures of\nconfidence in institutions, drawn from the World Values Survey,\nfrom which we derive three separate measures of institutional\ntrust: 1) trust in health institutions, which is calculated for each\nrespondent as the mean of trust in the World Health Organisation\n(WHO) and ‘‘the health service in your country”,3 2) trust in the\nmedia, which is the mean of trust in ‘‘the press” and ‘‘TV”, and 3)\ntrust in political institutions, which is calculated as the mean of trust\nin ‘‘parliament”, ‘‘political parties” and ‘‘the government in your\ncountry”. We also include a module developed in previous studies\nto measure generalized orientations of trust, mistrust and distrust\ntowards government [6,14]. The following set of questions was\nasked in all seven countries: To what extent do you agree, or disagree, with the following\nstatements? statements? (1) The government is honest and truthful\n(2) In general, the government usually does the right thing\n(3) The government usually has good intentions\n(4) I am usually cautious about trusting the government\n(5) I am unsure whether to believe the government\n(6) The government acts unfairly towards people like me\n(7) The government usually ignores my community\n(8) The government doesn’t respect people like me\n(1)\nStrongly agree\n(2)\nTend to agree\n(3)\nNeither agree nor disagree\n(4)\nTend to disagree\n(5)\nStrongly disagree\n(6)\nDon’t know 4 Readers should bear in mind the timing of data collection (October 2020 through\nJanuary 2021) and that the level of trust in each country may have increased or\ndecreased since (for instance in Myanmar – ‘MM’ in Fig. 1 – where trust may have\nfallen significantly since the 2021 coup d’etat).\n5 The response options were binary: ‘‘Yes, would agree” and ‘‘No, would not agree”.\nThose who responded ‘‘don‘t know” or refused to respond are coded as missing for the\npurposes of this analysis.\n6 The other response options were ‘‘not much” and ‘‘not at all”. Don’t knows and\nrefusals are once again treated as missing. Results Global analysis: Wellcome Global Monitor data Original survey data *Share of\npeople vaccinated against COVID-19 at start of survey fieldwork: complete initial vaccine protocol (partially vaccinated). Table 1\nOriginal survey details. Source: Our World In Data [18]. 4 Readers should bear in mind the timing of data collection (October 2020 through\nJanuary 2021) and that the level of trust in each country may have increased or\ndecreased since (for instance in Myanmar – ‘MM’ in Fig. 1 – where trust may have\nfallen significantly since the 2021 coup d’etat). Table 1 We further incorporate measures of the ‘Big Five’ personality traits,\nbased on the ‘Short 15-item Big Five Inventory (BFI-S)’ devised and\nvalidated by Lang et al. [15], creating scales for the mean (dis)\nagreement with the statements measuring each of the Big Five. We also include a scale of ‘conspiracy mentality’, designed for\nuse across cultures and languages, based on the 5-item scale by\nBruder et al. [5], albeit with a slightly modified question and\nresponse scale. We take the simple mean of these items as our\nmeasure of conspiracy mentality. Finally, we include a measure\nof cognitive reflection based on Frederick (2005; see also [25]),\nwhich principally measures the ability or disposition of respon-\ndents to reflect on a question and resist reporting the first response\nthat comes to mind (using classic games/experiments from beha-\nvioural economics). We take the sum of correct answers (0–3) to\nbe the respondent’s cognitive reflection score. Finally, we include\nmeasures of media consumption and activity: how much respon-\ndents follow news in different types of media and whether they\nhave recently ‘fact-checked’ articles or posted political content\nonline. All these scales and all other variables used in our analysis\nare rescaled to 0–1 to enable direct comparison of effects. After\nremoving all cases with missingness in the final model, 6,210\nrespondents are retained for the multivariate analysis. (5) Strongly disagree\n(6) Don’t know 6 The other response options were ‘‘not much” and ‘‘not at all”. Don’t knows and\nrefusals are once again treated as missing. 5 The response options were binary: ‘‘Yes, would agree” and ‘‘No, would not agree”.\nThose who responded ‘‘don‘t know” or refused to respond are coded as missing for the\npurposes of this analysis. Global analysis: Wellcome Global Monitor data Turning to our other independent variables, we provide more\ndetail in Online Appendix B, and only summarise our measures\nbriefly here. In terms of demographics, we include age, gender,\nand a dummy for university level education. Our models include\nleft–right ideological self-placement, measured on a 0–10 scale. 3 W. Jennings, V. Valgarðsson, L. McKay et al. Vaccine: X 14 (2023) 100299 Fig. 1. Trust in government vs self-reported vaccine intentions (aggregate-level), Wellcome Global Monitor. Countries denoted by Alpha-2 ISO codes, see: https://www. iban.com/country-codes. Fig. 1. Trust in government vs self-reported vaccine intentions (aggregate-level), Wellcome Global Monitor. Countries denoted by Alpha-2 ISO codes, see: https://www. iban com/country-codes Fig. 1. Trust in government vs self-reported vaccine intentions (aggregate-level), Wellcome Global Monitor. Countries denoted by Alpha-2 ISO codes, see: https://www. iban.com/country-codes. orted vaccine intentions (aggregate-level), Wellcome Global Monitor. Countries denoted by Alpha-2 ISO codes, see: https://www levels, which may drive both trust and vaccine intention). The full\nresults from these regression models are presented in Online Appen-\ndix, Table A1, and they show that moving through each stage from\nthe lowest to the highest level of trust in government is associated\nwith a significant increase in vaccine willingness, when controlling\nfor individual and contextual factors. Fig. 3 presents the marginal\neffects of each category of trust from these models, illustrating that\nthe global prediction (holding all control variables constant) for vac-\ncine intention is about 54 % for those who say they trust their\nnational government ‘‘not at all”, whereas it is about 71 % for those\nwho say they trust it ‘‘a lot”. Thus, the global relationship between\npolitical trust and vaccine intention appears to be significant and\nsubstantial even when accounting for demographic and contextual\nfactors. itive and moderate (0.49, p < 0.001). The simple bivariate regression\nline shows that at the lowest observed levels of trust (20 %), fewer\nthan 50 % are expected to take vaccines, while at the highest levels\n(near to 100 %), nearly 80 % would accept a no-cost vaccine. To explore how far this relationship is universal, we run bivari-\nate logistic regression models of vaccine intention on trust within\neach country on the individual level and present the odds ratios for\neach country in Fig. 2 (where trust in government is an ordinal\nvariable that takes the values 1 ‘not at all’, 2 ‘not much’, 3 ‘some’\nand 4 ‘a lot’). 7 Here, we include the full categorical variable with dummies for each category and\n‘‘not at all” as the baseline. Global analysis: Wellcome Global Monitor data While there is significant cross-national heterogene-\nity, it is striking that there is statistically significant (at the 95 per\ncent confidence level) positive relationship between trust in gov-\nernment and vaccine willingness in 75 out of 105 countries. In just\na single country (Bosnia Herzegovina), the relationship between\ntrust and vaccine willingness is significantly negative (i.e. confi-\ndence intervals of the estimated odds ratio do not cross at 1.0). In another 29 countries, the relationship was not statistically sig-\nnificant, interestingly including both high trust countries, such as\nNorway and Sweden, and low trust countries, such as Slovakia,\nSenegal and Serbia. The average odds ratio across countries is equal\nto 1.4, meaning that each one unit increase in the ordinal scale of\ntrust in government typically leads to a 40 % increase in the likeli-\nhood of being willing to be vaccinated. Detailed analysis: original surveys in seven countries While the Wellcome Global Monitor provides impressive global\nscope in terms of the total number of countries and proportion of\nthe global population covered, it is understandably limited in\nhow far it explores alternative measures of trust, as well as other\nrelevant values, beliefs and behaviours which may account for or\novershadow the relationship. We build upon this through fielding\nour own bespoke survey that specifically explored multiple dimen-\nsions of trust and a range of related factors in multiple countries. For our study, we are therefore able to draw on an array of mea-\nsures that distinguish between categories of generalized trust, mis-\ntrust and distrust in government, as well as more conventional\nmeasures of trust in specific institutions, where we differentiate\nbetween political institutions (parliament, political parties and\nthe national government), health institutions (the national health\nservice and WHO) and the media (the press and TV). We also\ninclude various measures of conspiratorial mentality, political ide-\nology, cognitive reflection, personality traits and traditional and\nonline media use, as well as demographic information. This\napproach enables us to explore the degree to which a much wider Turning to a more sophisticated analysis of this relationship, we\nnext run multi-level regression models on the entire dataset (with\ncountry as the second level). These again predict vaccine intent as a\nfunction of trust in government,7 and add several individual-level\ncontrols: age (in four categories), sex, education (a binary variable\nwhere those with university education are coded as equal to 1),\nemployment status and income on the individual level (in quintiles). At the country-level, we include the Human Development Index\n(HDI) score from the UN Human Development Reports (to partly\naccount for the potential confounding effect of societal welfare 4 Fig. 2. Bivariate regression of trust in government on self-reported vaccine willingness, odds ratios (individual-level, by country). Countries denoted by Alpha-3 ISO\nsee: https://www.iban.com/country-codes. W. Jennings, V. Valgarðsson, L. McKay et al. Vaccine: X 14 (2023) 1 W. Jennings, V. Valgarðsson, L. McKay et al. Vaccine: X 14 (2023) 100299 Fig. 2. Bivariate regression of trust in government on self-reported vaccine willingness, odds ratios (individual-level, by country). Countries denoted by Alpha-3 ISO codes,\nsee: https://www.iban.com/country-codes. Fig. 3. Marginal predictions of vaccine intention by levels of trust in government in the 2020 Wellcome Global Monitor survey, from multi-level regression models controlling\nfor individuals’ demographics and countries’ HDI. W. Jennings, V. Valgarðsson, L. Detailed analysis: original surveys in seven countries However, these measures are moderately correlated (the\npairwise Pearson’s r coefficients are between 0.45 and 0.5) and\nwhen the other institutional trust measures are included without\nthe others but with controls, both are significant predictors, so this\nshould not be interpreted to mean that trust in political or media\ninstitutions does not matter for vaccine willingness. In other\nwords, people’s general trust orientations towards government\nand their trust in political institutions matter for vaccine accep-\ntance, but when accounting for their trust in different specific state\ninstitutions, it turns out that health institutions are the most clo-\nsely relevant institution. We begin by considering the degree to which each of our mea-\nsures in isolation is associated with willingness to be vaccinated\nagainst COVID-19. Bivariate logistic regression models are calcu-\nlated for all predictors for the full, pooled seven-country sample,\nwith the odds ratios plotted in Fig. 4. In the figure, the dark circles\nindicate the coefficient value, while the whiskers indicate the 95\nper cent confidence intervals. The red line indicates the point at\nwhich there is no significant relationship between the variables if\nthose confidence intervals cross. Several measures of trust have statistically significant bivariate\nrelationships with vaccine willingness. The strongest positive rela-\ntionship\nis\nobserved\nfor\ntrust\nin\nhealth\ninstitutions\n(odds\nratio = 16.32, 95 % CI 12.40 to 21.48): in other words, those most\ntrusting were between 12 and 21 times more likely to accept the\nvaccine than those lowest in trust. Trust in the media (OR 4.96,\n95 % CI 3.70 to 6.64) and political institutions (OR 3.72, 95 % CI\n2.78 to 5.01) also have significant, but weaker, bivariate relation-\nships with vaccination willingness. p\ng\nAs expected, generalized trust orientations towards govern-\nment also have a positive effect on willingness (OR: 3.61, 95 % CI\n2.83 to 4.62), whereas distrusting and mistrusting orientations\nhave negative effects with mistrusters (OR: 0.49, 95 % CI 0.38 to\n0.62) and distrusters (OR: 0.52, 95 % CI 0.40 to 0.68) each about half\nas likely. Perhaps surprisingly, the difference between the effects of\ndistrust and mistrust are small, and social trust has weak and non-\nsignificant effects. Besides trust, we observe numerous other sig-\nnificant relationships. Detailed analysis: original surveys in seven countries McKay et al. Vaccine: X 14 (2023) 100299\n5 Fig. 2. Bivariate regression of trust in government on self-reported vaccine willingness, odds ratios (individual-level, by country). Countries denoted by Alpha-3 ISO codes,\nsee: https://www.iban.com/country-codes. Fig. 3. Marginal predictions of vaccine intention by levels of trust in government in the 2020 Wellcome Global Monitor survey, from multi-level regression models controlling\nfor individuals’ demographics and countries’ HDI. Fig. 3. Marginal predictions of vaccine intention by levels of trust in government in the 2020 Wellcome Global Monitor survey, from multi-level regression models controlling\nfor individuals’ demographics and countries’ HDI. 5 W. Jennings, V. Valgarðsson, L. McKay et al. Vaccine: X 14 (2023) 100299 set of measures of political attitudes are predictive of vaccine will-\ningness/hesitancy, and to situate different types of trust in that\ncontext. We provide further details on these data and measures\nin the methods section and the full questionnaire in Online Appen-\ndix A. age, and consumption of traditional media. However, we run all\nour block models on the same sample of respondents with non-\nmissing values on all variables (N = 6,235), to enable comparison\nof effects between models. The block models are displayed in Fig. 5, with effects again\nexpressed in terms of odds ratios. The first model includes gener-\nalized trust orientations, in both government and other people,\nand shows that generalized trust retains its positive effect\n(OR = 3.08, 95 % CI 2.08 to 4.56) on vaccine willingness and distrust\nits negative effect (OR = 0.58, 95 % CI 0.38 to 0.87), whereas mis-\ntrust becomes a non-significant predictor. Here, social trust\nbecomes a weakly negative predictor of willingness, albeit not sig-\nnificant at the 95 % level (p = 0.08), when accounting for political\ntrust orientations. In model 2, we include measures of institutional\ntrust, with intriguing results. Here, trust in political institutions\nbecomes non-significant and trust in the media only significant\nat the 90 % level (p = 0.08), but trust in health institutions remains\na strong predictor (OR = 15.37, 95 % CI 10.06 to 23.47) – suggesting\nagain that the latter is the most robust driver of vaccine willing-\nness. Detailed analysis: original surveys in seven countries Most notably, those with most ‘conspiracy\nmentality’ are about 10 % as likely as the least conspiratorial to\nbe willing to be vaccinated (OR = 0.10, 95 % CI 0.07 to 0.14) and\nthe oldest respondents are four times as likely as the youngest\n(OR = 3.98, 95 % CI 2.94 to 5.40). Engagement with different infor-\nmation sources also appears related to vaccine hesitancy: online\ninformation consumption, sharing and fact-checking material\nonline are associated with slightly lower willingness to get vacci-\nnated, though the positive association with consumption of tradi-\ntional media is slightly larger in magnitude. People who identify\nas on the right of the political spectrum are also somewhat less\nlikely to get vaccinated, as are those who score higher on some\nof the Big Five personality trait scales; especially on ‘openness’,\nwhereas cognitive reflection shows no association. Finally, as in\nother studies, university education is related to greater willingness\nto get vaccinated for COVID-19. Finally, the third model shows that conspiracy mentality is an\neven stronger negative predictor (OR = 0.07, 95 % CI 0.04 to 0.10)\nwhen including demographic controls. Because political trust and\nconspiracy mentality are closely related attitudes [10],8 we explore\nwhether trust is a confounder for the effects of conspiracy mentality\nor vice versa, by presenting the results of models including all of\nthese variables together in Online Appendix Fig. A2: these show that\nthe effects of conspiracy mentality survive controlling for all of the\ntrust measures, whereas the only trust measure that still has a sig-\nnificant effect is that of trust in health institutions, suggesting that\nboth shape vaccine willingness above and beyond political trust.9 Returning to other factors, we find that the effects of most per-\nsonality traits lose their significance in the block model (neuroti-\ncism is significant at the 90 % level, p = 0.06) but here, cognitive\nreflection\nshows\na\npositive\neffect\non\nvaccine\nwillingness\n(p < 0.01) and information engagement still has significant effects. In particular, the frequency of consuming traditional media (news-\npapers, television, radio) remains a strong predictor of vaccine\nwillingness (OR = 4.27, 95 % CI 2.89 to 6.32), while getting informa-\ntion from ‘people’ and online has negative effects, although rela-\ntively weak ones (and non-significant for the former). 8 The correlation between the conspiracy mentality scale and the generalized trust\nscales is between 0.22 and 0.28 Pearson’s r in the pooled dataset, with stronger\ncorrelations in France, Germany, and Spain.\n9 Generalized trust in government is still a significant predictor when including\nthose and conspiracy mentality together in a model, but its effects become\ninsignificant when controlling for trust in institutions as well. 8 The correlation between the conspiracy mentality scale and the generalized trust\nscales is between 0.22 and 0.28 Pearson’s r in the pooled dataset, with stronger\ncorrelations in France, Germany, and Spain. 9 Generalized trust in government is still a significant predictor when including\nthose and conspiracy mentality together in a model, but its effects become\ninsignificant when controlling for trust in institutions as well. Detailed analysis: original surveys in seven countries We find\nthat more ‘active’ internet users – those who post political content\n– are more vaccine hesitant (OR = 0.51, 95 % CI 0.37 to 0.69),\nalthough the effect of fact-checking becomes non-significant. We\nalso find that moving from left to right on a left/right scale is asso-\nciated with lower levels of vaccine willingness in all models. The\neffect of university education becomes non-significant in most\nmodels, but age retains a strong positive effect on willingness\n(while the effects of gender are still non-significant). The latter is While informative in a descriptive sense, these bivariate rela-\ntionships may well be confounded by various related demographic\nand attitudinal factors. To get closer to a causal relationship, we\nnext run a series of multivariate regression models, to isolate and\ndisentangle the effects of specific factors. We estimate six separate\n‘block’ models of related variables here, each retaining a selection\nof closely related variables. Each of the block models controls for\ndemographic factors (gender, age, education), left–right ideology\nand country dummies (because the dependent and independent\nvariables are likely to vary by country and the other control vari-\nables, in ways which might result in spurious effects reported in\nthe absence of these controls). In Online Appendix A, we also report\nresults of a single model including all 29 variables. We focus on\nblock models in this discussion, because including so many corre-\nlated variables in one model creates problems of multicollinearity,\nwide confidence intervals and very unintuitive interpretations of\nindividual coefficients 29. We see this model primarily as a conser-\nvative test of which predictors are the most robust of all: these\nturn out to be trust in health institutions, conspiracy mentality, 6 W. Jennings, V. Valgarðsson, L. McKay et al. Vaccine: X 14 (2023) 100299 Fig. 4. Bivariate logistic regression of vaccine willingness, odds ratios. Fig. 4. Bivariate logistic regression of vaccine willingness, odds ratios. traits interestingly appear to play a bigger role in those countries\nthan in the others. perhaps unsurprising given the sharp age gradient of the severity\nof COVID-19. Finally, we explore whether the determinants of vaccine hesi-\ntancy (or willingness) differ importantly between the countries\nin our dataset by conducting the bivariate and block models sepa-\nrately for each country. Coefficient plots from these models can be\nfound in the Online Appendix, Figs. A3-A23. Detailed analysis: original surveys in seven countries Most of the central\nfindings remain the same across countries, but generalized trust\norientations have by some margin the strongest relationship with\nvaccine hesitancy in France and Germany, whereas in Brazil polit-\nical trust negatively predicts vaccine willingness. This finding for\nBrazil is consistent with other studies suggesting that partisan cues\ncan reverse expected relationships between political trust and\nCOVID-related attitudes and behaviours (e.g., [12]). President Bol-\nsonaro was notoriously vaccine- and COVID-sceptic, so those\nrespondents expressing trust in his government were more likely\nto also hold anti-vax attitudes. Generally, most variables have\nweaker and often non-significant effects in India and particularly\nArgentina than in the other countries, but conversely, personality Discussion Our analysis here has served to demonstrate the critical link\nbetween political trust and vaccine hesitancy across the globe,\ndemonstrating how trust in government is predictive of willing-\nness to get vaccinated against COVID-19 for much of the world’s\npopulation. Generalized trust orientations towards government\nand trust in political institutions are consistent predictors of vac-\ncine hesitancy, although when including multiple interrelated\npolitical trust measures in the same models, trust in health institu-\ntions and conspiracy mentality stand out as the most robust pre-\ndictors. The design of this study has offered unprecedented\nglobal scope in its consideration of the degree to which trust in\ngovernment predicts vaccine hesitancy for COVID-19 (113 coun-\ntries) and systematic comparison of alternative measures of differ-\nent types of trust as well as other beliefs, opinions and traits, again\nin a substantial cross-national design (7 countries). This allows us 7 7 W. Jennings, V. Valgarðsson, L. McKay et al. Vaccine: X 14 (2023) 100299 onclude with increased confidence that trust is an essential\nmedia) are less willing. There is little evidence to support the clai\nFig. 5. Multivariate block models of vaccine willingness, odds ratios. Fig. 5. Multivariate block models of vaccine willingness, odds ratios. Fig. 5. Multivariate block models of vaccine willingness, odds ratios. to conclude with increased confidence that trust is an essential\nresource for societies in their fight against COVID-19, and that this\nis true both relative to other attitudes and controlling for the\ndemographics of individuals. Our findings confirm that willingness\nto be vaccinated against COVID-19 is positively linked to trust in\ngovernment, and negatively linked to distrust of government. Prior\ncross-national studies have reported mixed results in terms of the\neffects of political trust on vaccine hesitancy, but our findings sug-\ngest that this is likely because those that report non-significant\neffects of political trust do so via regression models which simulta-\nneously include various closely related measure of trust [14,24]. In\nterms of other factors, we find that consumers of traditional media\n(i.e. television, newspapers) are more willing to get vaccinated,\nwhereas the politically online (i.e. people who post on social media) are less willing. There is little evidence to support the claim\nthat personality traits are drivers of vaccine hesitancy, even in\nmodels that exclude strong predictors such as conspiracy beliefs\nand political trust. Discussion We find that cognitive reflection is associated\nwith greater vaccine willingness, but this is a small effect relative\nto other attitudinal and demographic predictors. Of the 29 predic-\ntors included in our regression models, political trust orientations\nand conspiracy mentality emerge as the strongest predictors of\nvaccine willingness, along with respondents’ age. to conclude with increased confidence that trust is an essential\nresource for societies in their fight against COVID-19, and that this\nis true both relative to other attitudes and controlling for the\ndemographics of individuals. Our findings confirm that willingness\nto be vaccinated against COVID-19 is positively linked to trust in\ngovernment, and negatively linked to distrust of government. Prior\ncross-national studies have reported mixed results in terms of the\neffects of political trust on vaccine hesitancy, but our findings sug-\ngest that this is likely because those that report non-significant\neffects of political trust do so via regression models which simulta-\nneously include various closely related measure of trust [14,24]. In\nterms of other factors, we find that consumers of traditional media\n(i.e. television, newspapers) are more willing to get vaccinated,\nwhereas the politically online (i.e. people who post on social We have here built on a still growing literature examining the\nlink between trust and vaccine hesitancy in the context of\nCOVID-19. Our study draws upon high quality survey data from\nthe Wellcome Global Monitor, enabling us to cover a substantially\ngreater number of countries than most empirical studies con- 8 W. Jennings, V. Valgarðsson, L. McKay et al. Vaccine: X 14 (2023) 100299 [7] Devine D, Gaskell J, Jennings W, Stoker G. Exploring Trust, Mistrust and\nDistrust (1st Digital Workshop of the ESRC ‘‘TrustGov” Project); 2020. https://\ntrustgov.net/s/TrustGov-Trust-mistrust-distrust-20042020.pdf. ducted during the pandemic, and through our own cross-national\nsurvey programme, allowing us to disentangle the roles of different\ntypes of trust and to directly compare their effects with other com-\nmonly cited factors. While the COVID-19 pandemic has eased in\nseverity by now, these results point to the importance of rebuilding\npolitical trust for the potential success of future mass vaccination\nprogrammes. [8] Devine D, Gaskell J, Jennings W, Stoker G. Trust and the Coronavirus Pandemic:\nWhat are the Consequences of and for Trust? An Early Review of the Literature. Polit\nStud\nRev\n2021;19(2):274–85. https://doi.org/10.1177/\n1478929920948684. [9] Devine D, Valgarðsson V, Smith J, Jennings W, di Vettimo MS, Bunting H,\nMcKay L. References [23] Nunnally JC, Bernstein IH. Psychometric Theory. McGraw-Hill; 1994. [24] Rozek LS, Jones P, Menon A, Hicken A, Apsley S, King EJ. Understanding Vaccine\nHesitancy in the Context of COVID-19: The Role of Trust and Confidence in a\nSeventeen-Country Survey. Int J Public Health 2021;66(May):. https://doi.org/\n10.3389/ijph.2021.636255636255. [1] Allington D, McAndrew S, Moxham-Hall V, Duffy B. Coronavirus conspiracy\nsuspicions, general vaccine attitudes, trust, and coronavirus information\nsource as predictors of vaccine hesitancy among UK residents during the\nCOVID-19\npandemic. Psychol\nMed\n2021. https://doi.org/10.1017/\nS0033291721001434. [25] Schernhammer E, Weitzer J, Laubichler MD, Birmann BM, Bertau M, Zenk L,\net al. Correlates of COVID-19 vaccine hesitancy in Austria: Trust and the\ngovernment. J Public Health 2021;44(1):E106–16. https://doi.org/10.1093/\npubmed/fdab122. [2] Bertsou E. Rethinking political distrust. Eur Polit Sci Rev 2019;11(2):213–30. https://doi.org/10.1017/S1755773919000080. [26] Schmeisser Y, Renström E, Bäck H. Who follows the rules during a crisis? Personality traits and trust as predictors of compliance with containment\nrecommendations during the COVID-19 pandemic in Sweden. Front. Polit. Sci. 2021;3(November):1–13. https://doi.org/10.3389/fpos.2021.739616. [3] Bollyky TJ, Hulland EN, Barber RM, Collins JK, Kiernan S, Moses M, et al. Pandemic preparedness and COVID-19: an exploratory analysis of infection\nand fatality rates, and contextual factors associated with preparedness in 177\ncountries, from Jan 1, 2020, to Sept 30, 2021. Lancet (London, England)\n2022;399(10334):1489–512. https://doi.org/10.1016/S0140-6736(22)00172-\n6. (\n)\np\n//\ng/\n/ p\n[27] Seyd B, Bu F. Perceived risk crowds out trust? Trust and public compliance\nwith coronavirus restrictions over the course of the pandemic. Eur Polit Sci Rev\n2022;14(2):155–70. https://doi.org/10.1017/S1755773922000078. [4] Brezzi M, Gonzalez S, Nguyen D, Prats M. An Updated OECD Framework on\nDrivers of Trust in Public Institutions to Meet Current (OECD Working Papers\non\nPublic\nGovernance,\nIssue\n48);\n2021. https://www.oecd-ilibrary. org/governance/an-updated-oecd-framework-on-drivers-of-trust-in-public-\ninstitutions-to-meet-current-and-future-challenges_b6c5478c-en. ( )\np\n//\ng/\n/\n[28] Stecula DA, Pickup M. Social Media, Cognitive Reflection, and Conspiracy\nBeliefs. Front. Polit. Sci. 2021;3(June):1–8. https://doi.org/10.3389/\nfpos.2021.647957. [29] Weinberg J. Can Political Trust Help to Explain Elite Policy Support and Public\nBehaviour in Times of Crisis? Evidence from the United Kingdom at the Height\nof the 2020 Coronavirus Pandemic. Polit. Stud. 2020. https://doi.org/10.1177/\n0032321720980900. [5] Bruder M, Haffke P, Neave N, Nouripanah N, Imhoff R. Measuring Individual\nDifferences in\nGeneric Beliefs\nin Conspiracy Theories Across Cultures:\nConspiracy Mentality Questionnaire. Front Psychol 2013;4(April). https://doi. org/10.3389/fpsyg.2013.00225. [30] Yuan H, Long Q, Huang G, Huang L, Luo S. Different roles of interpersonal trust\nand institutional trust in COVID-19 pandemic control. Ethical review [10] Easton D. A Re-Assessment of the Concept of Political Support. Br J Polit Sci\n1975;5(4):435–57. https://doi.org/10.1017/S0007123400008309. [11] Frederick S. Cognitive reflection and decision making. J Econ Perspect 2005;19\n(4):25–42. https://doi.org/10.1257/089533005775196732. This study was approved by the Faculty of Social Sciences\nResearch Ethics Committee at the University of Southampton\n(Ethics/ERGO Number: 61735). [12] Garry J, Ford R, Johns R. Coronavirus conspiracy beliefs, mistrust, and\ncompliance:\nTaking\nmeasurement\nseriously. Psychol\nMed\n2021;1–11. https://doi.org/10.1017/S0033291720005164. p\n//\ng/\n/\n[13] Hetherington MJ. Why Trust Matters: Declining Political Trust and the Demise\nof American Liberalism (Kindle Edi). Princeton University Press; 2005. Discussion Political trust in the first year of the COVID-19 pandemic: a meta-\nanalysis of 67 studies. J Eur Publ Policy 2023. https://doi.org/10.1080/\n13501763.2023.2169741. Declaration of Competing Interest 10.1007/s11135 017 0584 6. [16] Karic´ T, Mededovic´ J. Covid-19 conspiracy beliefs and containment-related\nbehaviour: The role of political trust. Pers Individ Differ 2021;175. https://doi. /10 1016/j\nid 2021 11069 /\n[16] Karic´ T, Mededovic´ J. Covid-19 conspiracy beliefs and containment-related\nbehaviour: The role of political trust. Pers Individ Differ 2021;175. https://doi. org/10.1016/j.paid.2021.110697. The authors declare the following financial interests/personal\nrelationships which may be considered as potential competing\ninterests: Will Jennings reports financial support was provided by\nthe UK Economic and Social Research Council (ESRC). Will Jennings\nreports financial support was provided by Research England QR\nGlobal Challenges Research Fund (GCRF). [17] Kerr JR, Schneider CR, Recchia G, Dryhurst S, Sahlin U, Dufouil C, et al. Correlates\nof\nintended\nCOVID-19\nvaccine\nacceptance\nacross\ntime\nand\ncountries: Results from a series of cross-sectional surveys. BMJ Open\n2021;11(8). https://doi.org/10.1136/bmjopen-2020-048025. [18] Lang FR, John D, Lüdtke O, Schupp J, Wagner GG. Short assessment of the Big\nFive: Robust across survey methods except telephone interviewing. Behav Res\nMethods 2011;43(2):548–67. https://doi.org/10.3758/s13428-011-0066-z. Acknowledgments [19] Lazarus JV, Ratzan SC, Palayew A, Gostin LO, Larson HJ, Rabin K, et al. A global\nsurvey of potential acceptance of a COVID-19 vaccine. Nat Med 2021;27\n(2):225–8. https://doi.org/10.1038/s41591-020-1124-9. Financial support was provided by the ESRC (ES/S009809/1) and\nResearch\nEngland\nQR\nGlobal\nChallenges\nResearch\nFund\n(519016146). [20] Lindholt MF, Jørgensen F, Bor A, Petersen MB. Public acceptance of COVID-19\nvaccines: Cross-national evidence on levels and individual-level predictors\nusing observational data. BMJ Open 2021;11(6):1–12. https://doi.org/10.1136/\nbmjopen-2020-048172. [21] Mathieu E, Ritchie H, Rodés-Guirao L, Appel C, Giattino C, Hasell J, et al. ‘‘Coronavirus Pandemic (COVID-19)”. Published online at OurWorldInData. org; 2022. Retrieved from: ’https://ourworldindata.org/coronavirus’ [Online\nResource]. Data will be made available on request. Data will be made available on request. [15] Johnston R, Jones K, Manley D. Confounding and collinearity in regression\nanalysis: a cautionary tale and an alternative procedure, illustrated by studies\nof British voting behaviour. Qual Quant 2018;52(4):1957–76. https://doi.org/\n10.1007/s11135-017-0584-6. Appendix A. Supplementary material Supplementary data to this article can be found online at\nhttps://doi.org/10.1016/j.jvacx.2023.100299. ]\n[22] McCarthy M, Murphy K, Sargeant E, Williamson H. Examining the relationship\nbetween conspiracy theories and COVID-19 vaccine hesitancy: A mediating\nrole for perceived health threats, trust, and anomie? Anal Soc Issues Public\nPolicy 2022;22(1):106–29. https://doi.org/10.1111/asap.12291. Data availability [14] Jennings W, Stoker G, Valgarðsson V, Devine D, Gaskell J. How trust, mistrust\nand distrust shape the governance of the COVID-19 crisis. J Eur Publ Policy\n2021;28(8):1174–96. https://doi.org/10.1080/13501763.2021.1942151. References Soc Sci Med 2022; 293\n(August 2021): 114677. doi:10.1016/j.socscimed.2021.114677. [6] Bunting H., Gaskell J., Stoker G. Trust, Mistrust and Distrust: A Gendered\nPerspective\non\nMeanings\nand\nMeasurements. Front. Polit. Sci. 2021:3\n(November):1-15. doi:10.3389/fpos.2021.642129. 9 9"