Policy Research Working Paper 10191 COVID-19 Vaccine Hesitancy in 53 Developing Countries Levels, Trends, and Reasons for Hesitancy Julia Dayton Eberwein Ifeanyi Edochie David Newhouse Alexandru Cojocaru Gildas Deudibe Jakub Kakietek Yeon Soo Kim Jose Montes Poverty and Equity Global Practice & Health, Nutrition and Population Global Practice September 2022 Policy Research Working Paper 10191 Abstract This paper presents new evidence on the levels and trends levels of hesitancy except in Iraq, Malawi, and Uzbekistan, of vaccine hesitancy in developing countries based on har- where hesitancy increased. COVID-19 vaccine hesitancy monized high-frequency phone surveys from more than is higher among female, young, less educated, and rural 120,000 respondents in 53 low- and middle-income coun- respondents, after controlling for selected observable char- tries. These countries represent a combined 30 percent of acteristics. Country estimates of vaccine hesitancy from the the population of low- and middle-income countries. On high-frequency phone surveys are correlated with but lower average across countries, one in five adults is hesitant about than those from earlier studies, which often relied on less the COVID-19 vaccine, with the most cited reasons for representative survey samples. The results suggest that vac- hesitancy being concerns about the safety of the vaccine, cine hesitancy in developing countries, while less prevalent followed by concerns about its efficacy. Between late 2020 than previously thought, will be an important and enduring and the first half of 2021, there tended to be little change in obstacle to recovery from the pandemic. This paper is a product of the Poverty and Equity Global Practice and the Health, Nutrition and Population Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http:// www.worldbank.org/prwp. The authors may be contacted at jdayton@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team COVID-19 Vaccine Hesitancy in 53 Developing Countries: Levels, Trends, and Reasons for Hesitancy Julia Dayton Eberwein, Ifeanyi Edochie, David Newhouse, Alexandru Cojocaru, Gildas Deudibe, Jakub Kakietek, Yeon Soo Kim, and Jose Montes1 JEL Codes: I12, I15, Keywords: vaccine hesitancy, COVID-19, high frequency phone surveys Topics: Coronavirus (COVID-19), vaccines, household surveys, health promotion and disease prevention 1 Affiliations: Dayton Eberwein, Kakietek: Health, Nutrition and Population Global Practice, The World Bank; Edochie, Cojocaru, Deudibe, Kim, Montes: Poverty and Equity Global Practice; Newhouse: Data, Analytics and Tools, Development Economics. Dayton Eberwein is the corresponding author: jdayton@worldbank.org. The authors are grateful to Renos Vakis and Sven Neelsen for useful comments on an earlier draft of the paper. All views and any remaining errors are those of the authors. 1. Introduction The world is entering the third year of the global COVID-19 pandemic, which has caused enormous devastation to both people’s health (450 million cases and 6 million deaths as of March 2022) and national economies, in the form of a global recession that has pushed millions into poverty (Dong, Du and Gardner 2020; World Bank 2022). With the continued emergence of new variants and limited treatments available, it is commonly accepted that widespread vaccination is the world’s best bet to contain the virus and it is also expected to play an important role in economic recovery, as widespread vaccinations would enable the reopening of the economy (IMF 2021; Hoogeveen and Lopez-Acevedo 2022). 2 As of early March 2022, over 60 percent of the world’s population had received at least one dose of a COVID-19 vaccine. However, there are stark disparities in vaccination rates across countries: only 14 percent of people in low-income countries had received at least one dose as of March 22, 2022, compared to 79 percent in high-income countries and 81 percent in upper-middle income countries (Ritchie et al. 2022) (Figure 1). The lag in vaccine distribution in lower income countries put the focus mainly on supply-side constraints until recently. However, the latest industry estimates predict that total vaccine production will exceed global demand by mid-2022, more than two years after the start of the pandemic and about a year and half since a COVID-19 vaccine was first authorized for emergency use. In fact, there are multiple reports of a possible “supply glut” and doses going to waste (Sanjay and Bloomberg 2022). Moreover, low- and middle-income countries have also been catching up with high-income countries in terms of the rates of vaccine rollout (Glassman, Kenny and Young, 2022; see also Figure 2). Despite this progress, a significant share of the world’s population remains unvaccinated, and if COVID-19 becomes endemic, the necessity of periodic boosters will reinforce the importance of addressing barriers to vaccination both on the supply and demand sides. To this end, it is increasingly important to understand the extent to which individuals in low- and middle-income countries elect to be vaccinated, and if not, to understand the nature of the concerns of those who choose not to. This will help address underlying bottlenecks to achieving immunity among the wider population, especially given the fear that persistent pools of unvaccinated people around the world present a greater risk for the emergence of new SARS-CoV-2 variants (Mallapaty 2022). 2 Note that lagged increases in vaccination rates for other diseases were also found to be positively associated with increases in GDP growth rates across countries, and increasingly so over time (Masia et al. 2018), but for non- COVID-19 vaccines the primary channel would be that of improved population health, rather than an enabling environment for economic activity. 2 Figure 1: Vaccination rates by income group Figure 2: Vaccine rollout rates by income group Share of people vaccinated against COVID-19, Mar 22, 2022 How many vaccine doses were administered in the previous 12 months? Per 100 people in the population. The value shown for each date is the total number of vaccine doses administered in the 12 months preceding that date. All doses, including boosters, are counted individually. Share of people with a complete initial protocol Share of people only partly vaccinated Upper middle income High income 100 Upper middle income 76% 4.8% 81% World Lower middle income Low income High income 74% 5.2% 79% 10 World 57% 6.9% 64% 1 Lower middle income 49% 10% 59% 0.1 Low income 11% 14% 0.01 0% 10% 20% 30% 40% 50% 60% 70% 80% Dec 15, 2020 Jun 4, 2021 Sep 12, 2021 Mar 22, 2022 Source: Official data collated by Our World in Data CC BY Source: Official data collated by Our World in Data CC BY Note: Alternative definitions of a full vaccination, e.g. having been infected with SARS-CoV-2 and having 1 dose of a 2-dose protocol, are ignored to maximize comparability between countries. Data source: Ritchie et al. 2022 as of March 22, 2022. Vaccine hesitancy refers to delay in acceptance or refusal of vaccination despite availability of vaccination services (MacDonald et al. 2015). It is not a new phenomenon, and some level of hesitancy exists for most vaccines, usually revolving around concerns about efficacy or safety. Even before the COVID-19 pandemic began, the World Health Organization (WHO) identified vaccine hesitancy as one of the top threats to global health (WHO 2019). The 2019 Wellcome Global Monitor from Gallup assessed general attitudes about the importance, safety, and effectiveness of vaccines across the globe prior to the onset of COVID- 19. Overall, on average, globally 7 percent of respondents either somewhat or strongly disagreed that vaccines are safe, and 5 percent either strongly or somewhat disagreed that vaccines are effective (Gallup 2019). Levels of vaccine hesitancy were lower in low- and middle-income countries compared with high- income countries. Across regions, countries in Eastern Europe and former Soviet states were among the least likely in the world to believe vaccines are safe and effective. Although these beliefs were expressed prior to the COVID-19 pandemic, they suggest that higher levels of COVID-19 hesitancy might be expected in Central and Eastern Europe as compared to other regions of the developing world. Furthermore, they suggest that concerns about safety are greater than concerns about efficacy. Despite the importance of vaccinations against COVID-19, little is known about the prevalence of COVID- 19 vaccine hesitancy, and even less about the extent to which concerns about the COVID-19 vaccine are similar or different to longstanding concerns about other vaccines. Anecdotally, several factors likely contribute to COVID-19 vaccine hesitancy: for example, the accelerated speed of development of the COVID-19 vaccine, combined with its newness, has intensified concerns about unknown longer-term side- effects. There are also reports of specific effects, such as of myocarditis in men under the age of 40 (Husby et al. 2022). Uncertainty about the duration of the vaccine’s effectiveness raises the possibility of needing annual vaccination. The availability of multiple types of vaccination, each with its own levels of efficacy and dosing schedule may also lead to confusion, and hence hesitancy. Relatedly, both the type of vaccine (e.g., mRNA) and country of origin of the vaccine have been identified as a factor in hesitancy. For example, a study from Brazil showed greater rejection among Brazilians for vaccines developed in China and the Russian Federation, as compared to vaccines from the United States or England, which could be related to local politics and foreign policy (Gramacho et al. 2021), as well concerns about data reporting related to vaccine development/approval (Bucci et al. 2021). Wong and colleagues (2021) reported greater hesitancy toward vaccines produced using the mRNA technology in Southeast Asia as compared with Europe and the Americas. 3 The multi-country studies that exist show relatively high levels of hesitancy to take the COVID-19 vaccine, but many of them focus on high-income countries. A systematic review and meta-analysis of 28 large (>1,000) nationally representative samples in 13 high-income countries collected between March and October 2020 found, across all 13 countries, that 72.9% of the sample intended to vaccinate (95% CI: 66.6% to 78.4%) (Robinson et al. 2021). The study also found that as the pandemic progressed (between March and October 2020), the percentage of people intending to vaccinate decreased and the percentage of people intending to refuse vaccination increased, although intentions varied between samples and countries. Another study of nationally representative surveys from mid-2020 in 19 countries, including 12 high-income countries (HICs), found an average across countries of vaccine acceptance rate across the 12 HICs of 66.6%, with wide variation ranging from 80% in the Republic of Korea to 55% in Russia (Lazarus et al. 2021). Both studies observed that being female, younger, of lower income or education level, and belonging to an ethnic minority group was consistently associated with being less likely to intend to vaccinate. Less is understood about the levels and reasons of COVID-19 vaccine hesitancy in low- and middle- income countries. A few studies reported multi-country survey results from developing countries (de Figueiredo and Larson, 2021, Facebook 2021, Gallup 2021, Kanyanda et al. 2021, Lazarus et al. 2021, Solis Arce et al. 2021, and Wouters et al. 2021) or systematic reviews of individual country surveys (Sallam 2021), but few of the studies were based on nationally representative samples of respondents or included comparable results for many countries. Taken together, the available literature from low- and middle- income countries suggests that vaccine acceptance is higher in East and South Asia and Latin America and lower in the Middle East and Eastern and Central Asia. Results from Africa were mixed, as were results for different subpopulations. Some studies indicated women were more hesitant than men, younger respondents were more hesitant than older respondents, and those with less education were more hesitant than persons with more education. The predominant reason for hesitancy was concern over side effects from the vaccine. Given the limited availability of randomly selected samples that are representative at the national level, an important contribution of this analysis will be to validate these findings. This paper extends the existing literature in the following four ways. First and most importantly it provides estimates of vaccine hesitancy for 53 developing countries, which are comparable across countries and largely based on household surveys that are more representative of the national population. In total, these 53 countries represent approximately 30 percent of the population of all low- and middle-income countries and 53.3 percent excluding India and China. Second, it investigates heterogeneity in hesitancy across demographic characteristics of the respondent such as urban or rural residency, gender, age and level of education. Better understanding this heterogeneity can inform the design of vaccination rollouts by helping to target specific population groups. Third, the study expands the evidence base on the reasons behind individuals’ reluctance to take the COVID-19 vaccine, even when it is available and free to them. Finally, the analysis enables a comparison between representative and previously published results from less representative sources. Taken together, the paper provides a rich set of results which can aid policy makers in the design of vaccination campaigns in developing countries. The rest of the paper is structured as follows. Section 2 describes the data and methods, and section 3 provides the results, first for levels and correlates of vaccine hesitancy, followed by changes over time and concluding with reasons reported for being hesitant. Section 4 discusses the results, compares them with the existing literature and suggests some lessons including policy implications. 4 2. Data and Methods 2a. Data This analysis describes the levels of vaccine hesitancy and its reasons in 53 developing countries between October 2020 August 2021 using data from the World Bank’s COVID-19 high-frequency phone surveys (HFPS) which were implemented to monitor the impact of COVID-19 on households around the world (World Bank 2020). Data are available for one survey round from 39 countries and two or more survey rounds for 14 countries. The countries in the pooled database represent five out of six regions defined by the World Bank: Latin America and the Caribbean (24 countries), Sub-Saharan Africa (14), East Asia and Pacific (7), Europe and Central Asia (6) and the Middle East and North Africa (2). 3 By income group, the sample includes 12 low-income countries, 13 lower-middle income countries, 23 upper middle-income countries and five high-income countries (Appendix Table 1 provides a list of countries with survey month, sample size, region and country income group). A global core questionnaire served as the basis for each country’s survey but was then customized to fit the local context. As a result, the exact questions may vary across countries, but topics typically included knowledge and concerns about COVID-19, access to food, health care and education, employment and income loss, and safety nets and coping strategies. The survey data was then harmonized ex-post to enable cross-country comparability. The questionnaire was flexible and adapted over time to the pandemic context. The vaccine module, for example, was included more recently. The sampling frame was drawn from pre-existing nationally representative household surveys in 19 countries, random digit dialing (RDD) in 29 countries, and a list of phone numbers typically obtained from mobile phone operators in 5 countries. The RDD design samples from all active mobile and landline phone numbers, such that RDD surveys are representative of the population aged 18 and over with an active phone number, conditional on response and survey completion. Sampling weights were then constructed to correct for selection bias due to the inability to contact households that did not participate in the survey either due to non-response or lack of access to a phone, with the goal of obtaining estimates that are as close to being nationally representative as possible. Information was collected from one respondent per household. In the case of countries where the sampling frame was derived from a previous survey, this was typically the household head. In other cases, such as when the sample utilized RDD, the respondent was more representative of individuals within the household. While the non-random selection of individuals in the former cases is not ideal, in an earlier analysis of labor market indicators derived from phone surveys, Kugler et al. (2021) find little evidence of differences in employment outcomes or trends over time being affected by the oversampling of household heads with respect to variables such as level of education, gender or location, the main source of bias being for age comparisons. Within the pooled data set, the household sampling weights are scaled such that countries are weighted equally, such that estimates of vaccine hesitancy reported in this study are averages of country averages. 4 2b. Outcome measures The questions asked about vaccine hesitancy in the surveys varied depending on whether a vaccine was available in the country at the time of the survey. In surveys that took place before the vaccine rollout had 3 South Asia is not represented due to lack of available data. The Middle East and North Africa includes two countries—Iraq and Lebanon. 4 Using population weights would have resulted in the results being driven by a small number of large countries. 5 begun in that country, the question asked was “If an approved vaccine to prevent COVID-19 was to become available at no cost, would you agree to be vaccinated?” In surveys that took place after the vaccine rollout had begun, the wording of the question was either “Are you planning to be vaccinated?” or “When a vaccine to prevent COVID-19 is available to you, are you planning to be vaccinated?” In the sample, 47 countries offered three answer options (yes, not sure, and no), while the other six offered only two categories (yes and no). For this analysis, we combine the “no” and “not sure” answers to obtain the measure of vaccine hesitancy. To obtain the respondent’s reason for vaccine hesitancy, survey respondents who answered “no” or “not sure” were asked “What is your (main) reason/concern for not wanting to be vaccinated / not being sure if you want to be vaccinated?” The answer categories varied widely across surveys. To make these more comparable across countries, answers were remapped into the following nine most common categories (see Appendix Tables 2a and 2b for the original answer options in each survey and a mapping of how they were harmonized): 1. Safety, which includes concerns about side effects 2. Efficacy 3. Distrust of government, pharmaceutical industry, international community 4. Dislike of vaccines in general 5. Preference for natural immunity, which included perceiving self as low risk and already having had COVID-19 infection 6. Lack of knowledge or access, which included “I do not have enough information about the vaccine,” “too hard to get,” “health facility to far” and “I don’t have the time” 7. Religious reasons 8. Not eligible, which included “counter-indication,” “recent medical discharge, “have an underlying health condition and believe taking the vaccine will make it worse” 9. Other The surveys also differed across countries in terms of whether a single concern or multiple concerns were collected. To account for the difference in the number of response options, the results are presented separately. 2c. Contextual data Country-level contextual data were drawn from other sources. New COVID-19 cases per million, measured as a 7-day rolling average prior to the midpoint of the month household survey data was collected in each country, were obtained from Dong and colleagues (2022), and the Oxford Stringency Index was drawn from Hale et al. 2022. The latter is an index, ranging from 0–100, that indicates the degree to which various restrictions were put in place by governments to control the pandemic (for example, school closures and shelter-in-place requirements). Confidence in the press, the government, and the WHO was drawn from the World Values Survey Wave 7, fielded in 2017-2020, and aggregated up to the country level (Inglehart 2022). Excess deaths due to COVID-19 draw on WHO data (WHO 2022). Country geographic region and income group are based on the World Bank classifications. 2d. Analytical strategy Data from 53 countries was pooled into a single data set. First, we report point estimates and 95% confidence intervals for each country based on standard errors clustered at the state or province level within each country. We also report the simple average across all countries and stratified by: (i) World Bank region, (ii) World Bank country income group, (iii) urban vs. rural residence, (iv) gender of the 6 respondent, (v) whether the respondent is head of household, (vi) age of the respondent (under 35 years, 35-64 years, and 65 years and older, and (vii) educational attainment of the respondent (some primary school, some secondary school, some tertiary). Tukey’s test of multiple comparisons is used to test the significance of differences in rates of vaccine hesitancy across the above-mentioned groups. Second, we use multivariate regression analysis to assess the relative association of vaccine hesitancy and the correlates and contextual variables. Trends over time in levels of vaccine hesitancy are examined in 14 countries with more than one wave of results available. Finally, the analysis describes the reasons for vaccine hesitancy. 3. Results 3a. Levels of vaccine hesitancy across 53 countries Overall, when taking a simple average across countries, the average level of vaccine hesitancy across the most recent survey round in each country was 20.0% (Confidence Intervals (CI) 17.2-22.7%). Across regions, average levels of hesitancy were highest in Europe and Central Asia (58.8%, CI 55.0-62.6%), followed by the Middle East and North Africa (47.4%, CI 38.8-56.0%), East Asia and Pacific (26.2%, CI 21.4- 31.0%), Sub-Saharan Africa (15.5%, CI 11.8-19.2%), and Latin America and the Caribbean (8.0%, CI 6.5- 9.5%) (Figure 3 and Appendix Table 3). Differences between all regions were statistically significant at the 1% level, except for the difference between Europe and Central Asia and Middle East and North Africa (p- value=.017). (See Appendix Table 4 for all results of tests of differences.) When considered by country income group, the highest average level of hesitancy was in lower middle-income countries (27.7%, CI 23.8-31.7%), followed by low-income countries (14.6%, CI 7.8-21.4%), upper middle-income countries (12.7%, CI 9.8-15.6%), and high-income countries (5.9%, CI 3.4-8.4). The level of vaccine hesitancy across country income groups was significant at the 5% level across all pairs with the exception of the comparison between low-income and upper-middle income groups. Figure 3: Share of households that were hesitant to be vaccinated against COVID-19 in 53 countries Note: Weighted estimates, with weights scales such that countries are given equal weights. 7 These averages across countries mask the wide range of country-level averages of vaccine hesitancy. The highest levels of vaccine hesitancy were reported in Kazakhstan (75.3%), Bulgaria (66.2%) and Georgia (65.2%), and the lowest levels in Chile (3.4%), Ethiopia (3.4%), and Brazil (3.1%) (Figure 4 and Appendix Table 3). People who are “not sure” about getting vaccinated are likely to be more easily persuaded than those who answer that they would not get vaccinated, and thus there would be more opportunities to increase vaccine uptake in countries with higher shares of respondents who responded “not sure.” Figure 5 shows these results for the 46 countries that included all three answer options (yes, not sure, and no). In four countries, over 20% of respondents reported being unsure of whether they would get vaccinated: these countries are the Philippines (26.3%), Georgia (25.9%), Jamaica (24.5%), and Kazakhstan (22.6%). In another 11 countries, between 10% and 20% of the sample was unsure whether they would be willing to be vaccinated against COVID-19: these countries include Iraq, St. Lucia, Belize, Haiti, Zimbabwe, Mongolia, Indonesia, Dominica, The Gambia, Mali, and Antigua and Barbuda. Figure 4: Share of households that were hesitant to be vaccinated against COVID-19, by country Notes: Weighted estimates, with weights scales such that 53 countries are given equal weights. Black bars indicate confidence intervals. Point estimates are given in Appendix Table 3. 8 Figure 5: Strength of sentiment: share of households in 46 countries reporting being “not sure” if they will take the COVID-19 vaccination when available Note: Figure only includes the 46 countries with three response options (yes, not sure, no). 3b. Bivariate correlates of vaccine hesitancy As expected, the extent of hesitancy varies for different types of respondents. Rural households reported significantly higher levels of vaccine hesitancy than urban households (23.2% vs. 17.7%, p-value=0.018), and female respondents (22.5%) were significantly more likely than male respondents (17.3%) to be vaccine hesitant (p-value=0.021) (Figure 6). Among households in the 33 surveys with information on the educational attainment of the respondent, respondents with lower levels of education had on average higher levels of hesitancy: 22.8% of respondents with no education were hesitant, compared with 19.8% of those with any primary education, 19.0% of respondents with any secondary education, and 13.7% of those with some tertiary education. Only the rates of hesitancy of those with at least some tertiary education were significantly lower than those with no education (p-value=0.027), some primary (p- value=0.019), and some secondary (p-value=0.052). Younger respondents reported higher levels of hesitancy than older respondents: 20.3% of respondents under 35 years of age were hesitant to get vaccinated, 20.1% of respondents aged 35-59 years, and 17.7% of those over the age of 60 years. None of the differences in age were statistically different from each other in bi-variate tests. 9 Figure 6: Share of households that were hesitant to be vaccinated against COVID-19 in 53 countries, by individual characteristics Notes: Weighted estimates, with weights scales such that 53 countries are given equal weights, except for results by education which include only results from the 33 country surveys with information on the educational attainment of the respondent. 3c. Multivariate analysis of correlates of vaccine hesitancy Results from a multivariate regression generally confirm the bivariate results. Results from OLS regression models are shown in Table 1: a baseline model with the correlates described above plus month of survey (expressed in terciles), 5 and extended models with additional contextual variables (expressed in terciles). Results from these models are consistent, as are the marginal effects from logit regressions (see Appendix Table 5a). Men are less COVID-19 vaccine hesitant than women, although the magnitude is small. Having less formal education is associated with being more hesitant and the differences are more substantial. Respondents with no formal education are 12.3 percentage points more likely to be hesitant than respondents with some tertiary education. Those with some primary or some secondary education are 6.1 and 6.4 percentage points more likely, respectively. Age is inversely correlated with hesitancy – adults over 65 years of age are 10.3 percentage points less likely to be hesitant compared with respondents under the age of 35 years, and respondents aged 35-64 years are 4.7 percentage points less likely. There are no statistically significant differences between hesitancy among rural and urban respondents. 5 Time of survey was split into the following 3 groups: (i) November 2020 – January 2021; (ii) March – May 2021; and (iii) June – August 2021. There are no observations in the data for February and April of 2021. 10 Differences across regions were statistically significant and, for some regions, quite large: those living in Europe and Central Asia are 36.3 percentage points more likely to be hesitant than those living in the reference region of Latin America and the Caribbean, those in the Middle East and North Africa are 23.1 percentage points more likely and those in East Asia and the Pacific, 21.1 percentage points more likely. Differences between Sub-Saharan Africa and Latin America and the Caribbean were not statistically significant, nor were differences by country income group. Over time, the probability of being hesitant to take the COVID-19 vaccine declined but not significantly. Table 1: Multivariate analysis of vaccine hesitancy in 53 developing countries Dep. Var: Hesitancy (1) (2) (3) (4) (5) (6) Male -0.028* -0.027* -0.028 -0.027* -0.031* -0.038** (0.013) (0.013) (0.014) (0.013) (0.013) (0.012) Head of HH 0.012 0.009 0.019 0.000 0.011 0.007 (0.015) (0.014) (0.013) (0.017) (0.015) (0.013) Education of respondent (ref – Tertiary) No education 0.123*** 0.125*** 0.116*** 0.134*** 0.125** 0.143*** (0.033) (0.035) (0.032) (0.037) (0.040) (0.036) Any primary 0.061** 0.063** 0.060** 0.067** 0.052* 0.070** (0.018) (0.019) (0.018) (0.019) (0.020) (0.021) Any secondary 0.064*** 0.065*** 0.062*** 0.065*** 0.058*** 0.062*** (0.011) (0.011) (0.011) (0.012) (0.011) (0.011) Age group (ref. -- 34 and younger) Working age (35-64) -0.047** -0.048*** -0.042** -0.050*** -0.045** -0.049*** (0.014) (0.013) (0.015) (0.013) (0.015) (0.012) Retirement age (65+) -0.103*** -0.102*** -0.096*** -0.103*** -0.099*** -0.099*** (0.023) (0.024) (0.023) (0.022) (0.024) (0.023) Rural area 0.016 0.014 0.013 0.010 0.012 0.011 (0.019) (0.018) (0.018) (0.015) (0.015) (0.011) Region (ref – LAC) EAP 0.211** 0.218* 0.241** 0.127 0.299 0.249 (0.076) (0.086) (0.084) (0.076) (0.155) (0.137) ECA 0.363*** 0.364*** 0.347*** 0.376*** 0.369*** 0.345*** (0.075) (0.076) (0.073) (0.076) (0.056) (0.068) MNA 0.231*** 0.230*** 0.262*** 0.240*** 0.288*** 0.393*** (0.043) (0.043) (0.062) (0.036) (0.059) (0.050) SSA 0.082 0.089 0.078 -0.005 0.072 0.057 (0.075) (0.079) (0.074) (0.101) (0.086) (0.103) Income group (ref. – LIC) LMIC -0.073 -0.071 -0.052 -0.090 -0.049 -0.053 (0.127) (0.124) (0.122) (0.129) (0.113) (0.081) UMIC -0.052 -0.050 -0.071 -0.096 -0.026 -0.130 (0.154) (0.154) (0.146) (0.154) (0.135) (0.107) HIC -0.161 -0.160 -0.183 -0.208 -0.175 -0.280* (0.159) (0.162) (0.153) (0.162) (0.136) (0.108) Survey month (ref -- Nov 2020 - Jan 2021) March - May 2021 -0.073 -0.062 -0.102 -0.056 -0.113 -0.110 (0.095) (0.108) (0.104) (0.101) (0.112) (0.095) June - August 2021 -0.018 -0.012 -0.018 -0.028 -0.049 -0.026 (0.090) (0.099) (0.099) (0.089) (0.097) (0.079) New COVID-19 cases per million, terciles -0.022 0.023 Cases (middle tercile) (0.065) (0.057) -0.009 -0.011 Cases (top tercile) (0.054) (0.049) 11 Dep. Var: Hesitancy (1) (2) (3) (4) (5) (6) Oxford stringency index terciles (ref. – bottom tercile) Stringency (middle tercile) 0.061 0.148 (0.073) (0.077) Stringency (top tercile) -0.005 0.028 (0.060) (0.059) WHO excess deaths due to COVID-19 (ref.- bottom tercile) Excess deaths (middle tercile) -0.110* -0.123 (0.050) (0.080) Excess deaths (top tercile) -0.059 0.045 (0.067) (0.106) Confidence in government index tercile (ref.- top tercile) Confidence in government 0.073 0.105 (0.054) (0.064) Confidence in government (top 0.016 0.055 (0.131) (0.135) Constant 0.282* 0.286* 0.265 0.395** 0.206 0.190 (0.136) (0.140) (0.141) (0.143) (0.149) (0.145) R-squared 0.090 0.090 0.095 0.097 0.102 0.123 N 65088 65088 65088 65088 65088 65088 Notes: Weighted logit regressions. Marginal effects reported. Standard errors clustered at country level. *, **, *** indicates significance at the 95%, 99%, and 99.9% level. In the extended regression model, we test two additional hypotheses. First, we expect that the severity of the pandemic, for which we use three different proxies, would be negatively correlated with hesitancy. Our first measure, in column (2) is the number of new cases of COVID-19 reported in the country. The higher the risk of infection and illness, as captured by the number of new cases, may make respondents more inclined to get vaccinated to protect against COVID-19 infections. At the same time, it should be recognized that testing capacity differed significantly across countries and over time, such that the measures of new COVID-19 cases per million is an imperfect proxy for pandemic severity. Nevertheless, this variable also captures the visibility of the pandemic to the overall population (even if true severity may be higher), and we would expect visibility / salience to be negatively correlated with hesitancy as well. Our second proxy for pandemic severity, in column (3), is the degree of stringency in the government’s policy measures, as captured by the Oxford policy stringency index. We expect ex-ante that a more stringent the policy response, signaling higher severity of the pandemic and more restrictions on movement, would be positively associated with willingness to be vaccinated. Our third proxy for pandemic severity, in column (4), is the number of excess deaths associated with the COVID-19 pandemic, as reported by the WHO. The excess mortality measure quantifies the direct and indirect impacts of COVID-19, measured in terms of the difference between the total number of deaths for a specific place and time period and the number of deaths that would have been expected in the absence of COVID-19. This difference is inclusive of both deaths directly attributable to COVID-19 as well as deaths indirectly associated with COVID-19 through its impact on health systems and society. As a prior, we would expect higher levels of excess COVID-19 deaths to be positively correlated with willingness to get vaccinated, although, at the same time, higher hesitancy can also bring about more excess deaths. 12 In our data, none of the three proxies for pandemic severity and/or visibility is statistically associated with vaccine hesitancy with the exception of COVID-19 excess deaths, where in countries in the second tercile of the excess deaths distribution vaccine hesitancy is, on average, lower, relative to the bottom tercile baseline. We also test whether trust in the government in general is associated with COVID-19 vaccine hesitancy, expecting a positive association due to widespread reporting of misinformation and antivax messages reported in the press, especially online. However, this association is not statistically significant in the regression model. Finally, we also disaggregate the vaccine hesitant population into those who say that they would not get a COVID-19 vaccine and those who are not sure (see Figure 5) and consider whether the underlying characteristics of these two subgroups differ. In order to do so, we consider vaccine hesitancy as being a spectrum that spans from a rejection of the vaccine to the acceptance of the vaccine, with uncertainty being an intermediate category. This allows us to model the vaccine decision as an ordinal model, keeping the extended specification from table 1 (see Table 5b in the Appendix). When we examine the correlates of the “No” and the “Not sure” groups separately, we observe similar patterns, but they are much weaker in terms of magnitudes for the “Not sure” group. For instance, lower levels of education are associated with a higher probability of answering No, and this is true for the Not sure group, but the magnitude of the marginal effects is much smaller for the latter. Likewise, the probability of answering “Not sure” is also higher among the working age and older population, relative to those under the age of 35, but the strength of these negative correlations is considerably smaller. The ECA, MNA and EAP regions have a higher probability of both No and Not sure answers relative to the LAC region baseline. 3d. Changes in vaccine hesitancy over time Estimates of vaccine hesitancy were available for two or more survey rounds in 14 countries in the sample. Seven countries had estimates from two survey rounds (Burkina Faso, Republic of Congo, Ethiopia, Indonesia, Nigeria, Philippines, and Uganda) and seven countries (The Gambia, Georgia, Iraq, Kazakhstan, Malawi, Tajikistan, and Uzbekistan) had results from three or more survey rounds. All surveys were collected between October 2020 and August 2021, although the specific months of data collection varied across countries. There are no clear patterns in terms of the changes in vaccine hesitancy in countries with multiple data points over the pandemic. Among the 14 countries with two or more surveys (collected between October 2020 and August 2021), the rates of COVID-19 vaccine hesitancy declined in half of the countries and increased in the other half, and changes in either direction were less than five percentage points in all but three countries (Figure 5 and see Appendix Table 6 for point estimates and confidence intervals). However, larger increases were observed in Iraq (30% increase), Malawi (36% increase), and Uzbekistan (18% increase). No patterns of change over time were observed across sub-populations either. It should be noted, however, that the period for which multiple rounds of the phone surveys are available is relatively short, and during this time very few countries experienced substantial increases in vaccination rates, such that individuals would have few in-person experiences from which to model behavior. 13 Figure 7: Changes in levels of vaccine hesitancy in 14 countries, October 2020 to April 2021 Notes: The shaded area around the line represents the confidence bands. 3e. Reasons for COVID-19 vaccine hesitancy Survey respondents in 45 countries who answered “no” or “not sure” about being vaccinated were asked the reason for their hesitancy. The countries include 23 in Latin America and the Caribbean, 14 in Sub- Saharan Africa, five in East Asia and the Pacific, two in the Middle East and North Africa (Djibouti and Lebanon), and one in Eastern Europe and Central Asia (Georgia). In the Republic of Congo, results on reasons for hesitancy were reported for two survey waves (December 2021 and March 2022). In 38 countries, the question solicited a single reason, whereas in seven countries the survey instrument allowed for multiple responses. Overall, the most common concern pertained to the safety of the vaccine, including concerns about side effects. This was true regardless of the number of responses collected. Figures 8a and 8b show the distribution of responses for countries with single and multiple response options, respectively. Among the 42 countries with the single answer option, an average of 43% of respondents cited safety concerns as the primary reason for not planning to be vaccinated, with a range from 15.4% in Somalia to 84.6% in the Philippines. Other commonly cited concerns included efficacy of the vaccine (19%), dislike of vaccines in general (9%), a preference for natural immunity (which includes perceiving self as low risk and already having had COVID-19 illness) (8%), lack of knowledge about or access to the vaccine (6%), lack of trust in the government/pharmaceutical industry/international community (5%), religious reasons (4%), perceiving oneself as not eligible (which includes counterindication) (4%) and other reasons, which included concerns about getting COVID-19 at the facility and other unspecified reasons (9%). These 14 averages mask significant variation across countries. Concerns about vaccine efficacy were greater or equal to concerns about safety in five Latin American countries (Argentina, Chile, Dominica, Dominican Republic, and Mexico). Respondents in the Democratic Republic of Congo and the Republic of Congo were hesitant because of dislike for vaccines in general (40% and 44% in March 2021 of hesitant respondents, respectively) and distrust for government, pharmaceutical industry, and the international community (40% and 22% in March 2021, respectively). Supply-side concerns for not getting vaccinated such as “health facility too far or too hard to get to,” “there is shortage of vaccines in the country,” and “I don’t know how to access the vaccine” (see Appendix Table 2b for full list) only accounted for 6% on average across all surveys with the single answer option. However, over 10% of respondents in several Latin American countries reported the concerns “Health center too far or hard to reach” or “I don't have time to go to get vaccinated.” These included Antigua and Barbuda (22%), St. Lucia (18%), Jamaica (16%), Guyana (13%), Belize and Nicaragua (both 12%). The relatively incidence of supply-side concerns in the HFPS data may be influenced by the framing of the question, however, such as the access issues being queried in the context of other concerns, or because respondents were asked if they would get vaccinated “when the vaccine becomes available”, and access constraints may be equated in people’s minds with vaccines not being available presently, and thus vaccines becoming available signifying the alleviation of existing supply constraints. We are not able to test this formally, however. Among the seven countries with multiple answer options, a similar pattern was observed: concerns about safety were the most common concern cited in five countries (Mali, Nigeria, Kenya, Sudan, Uganda). In The Gambia, dislike of vaccines in general was the most cited reason, followed by preference for natural immunity (which includes perceiving self as low risk and already having had COVID-19 illness). In Georgia, the only country in Europe and Central Asia with information on reasons for hesitancy, the most cited reason was “not eligible,” followed by lack of knowledge or access and preference for natural immunity. Only one country, the Republic of Congo, reported reasons for two waves. Between December 2020 and March 2021, only small changes in reasons were seen: the share of respondents who cited distrust of the international community declined from 30% to 22%, and the share who cited concerns about safety and dislike of vaccines in general increased (from 27% to 31% and 42 to 44%, respectively). 15 Figure 8a: Reason for vaccine hesitancy, countries with single response Philippines Laos Mongolia Malawi Burkina Faso Sierra Leone Lebanon Bolivia Djibouti Colombia Ecuador Ethiopia Panama Peru Belice Paraguay El Salvador Jamaica Santa Lucia Nicaragua Uruguay Guatemala Dominican Republic Indonesia Honduras Guyana Costa Rica Congo, Rep (Mar 2021) Antigua & Barbuda Argentina Guinea Congo, Rep (Dec 2020) Malaysia Mexico Chile Haiti Dominica DRC Somolia 0 20 40 60 80 100 Share of vaccine hesitant households Safety (includes side effects) Efficacy Distrust (of government/pharmaceutical industry/international community) Dislike vaccines in general Prefer natural immunity Lack of knowledge or access Religious Not eligible Other Note: For the construction of harmonized response categories, see Appendix Table 2b. 16 Figure 8b: Reason for vaccine hesitancy, countries with multiple responses Kenya Uganda Sudan Mali Nigeria Gambia Georgia 0 20 40 60 80 100 120 140 160 Safety (includes side effects) Efficacy Distrust (of government/pharmaceutical industry/international community) Dislike vaccines in general Prefer natural immunity Lack of knowledge/access Religious Not eligible Notes: The X axis represents share of vaccine hesitant households. In some cases, the share is larger than 100% because multiple responses were allowed For the construction of harmonized response categories, see Appendix Table 2b. 4. Discussion 4a. Principal findings Data from the HFPS were used to investigate COVID-19 vaccine hesitancy. On average across 53 countries, one in five adults are hesitant about getting a COVID-19 vaccine. The highest levels were observed in Eastern European and Central Asian countries and the lowest levels in Latin American and Caribbean countries. Among population groups, female respondents, younger adults, and those with less formal education reported higher levels of COVID-19 vaccine hesitancy than their respective counterparts. Between October 2020 and August 2021, there tended to be little change in levels of hesitancy except in Iraq, Malawi, and Uzbekistan, where hesitancy increased. The main self-reported reason for being hesitant was concerns over safety, especially worries about side effects, and to a lesser extent, concerns over vaccine efficacy and dislike of vaccines in general. 4b. Strengths and comparisons with other studies This study has several advantages over the existing literature. It is based on a very large sample of respondents from national surveys in 53 developing countries, a part of the world that is under- represented in the literature. Over half of the surveys used random digit dialing, typically in upper middle- income countries in which a large share of the population uses mobile phones. In lower income contexts, most surveys, 35 percent overall, sampled based on previous face-to-face surveys which were in turn drawn from a census frame. In the remaining 10 percent of cases, sampling was carried out from lists provided by mobile phone operators. In each case, surveys drawn from pre-existing face-to-face surveys were reweighted using the baseline data to become more representative, while many random digit dialing surveys were also reweighted to make those samples more representative of the universe of phone numbers. While it is impossible to eliminate issues of representativeness in phone surveys, the high- 17 frequency phone surveys were typically carried out by National Statistics Offices and are more plausibly representative than convenience web surveys. Overall, the estimates of COVID-19 vaccine hesitancy from the HFPSs were more conservative than other published estimates (Figures 9 and 10). There are various possible explanations for the differences. The first relates to the timing of data collection. Most of the existing literature is based on surveys carried out in mid to late 2020 (e.g., de Figueiredo and Larson 2021, Gallup 2021, Lazarus et al. 2021, Solis Arce et al. 2021, and Wouters et. A. 2021), whereas the HFPS data included in our sample were collected from end- 2020 through August 2021. It is plausible to believe COVID-19 vaccine hesitancy generally declined during this period as multiple vaccines became available and widespread vaccination was safely rolled out in high-income countries. Another possible reason for the differences is the variation in how the vaccine hesitancy question was framed and the response options available. For example, the HFPS question in countries without general access to the vaccine at the time of data collection was “When a vaccine to protect you from COVID-19 is available to you, are you planning to be vaccinated?” whereas other studies included the phase “at no cost to you” or descriptors about the quality of the vaccine such as “COVID-19 vaccine proven safe and effective”) (Gallup 2021, Lazarus et al. 2021). There was even more variation across the studies in the answer options available to the respondents. One study relied on a 5-point Likert scale of agreement with the statement “If a COVID-19 vaccine is proven safe and effective and is available, I will take it”: ‘completely disagree’, ‘somewhat disagree’, ‘neutral/no opinion’, ‘somewhat agree’ and ‘completely agree” (Lazarus 2021). Others relied on a 4-point Likert scale to the question such as “yes, definitely,” “yes, probably,’ “no, probably not,” and “no, definitely not” (Africa CDC, Facebook 2021, de Figueiredo and Larson 2021, Wouters et al. 2021), whereas the HFPS mostly relied on three answer options (yes, not sure, no). It is possible that “probably not” is not the same as “unsure,” although it is not known how this might bias the respondents’ answers and cross-study comparisons. Finally, variations in survey modality may have created biases. Among the published studies included for comparison, many studies were based on data from commercial online sample providers, often using quotas sampling to ensure an appropriate distribution in terms of gender, age, and region (Africa CDC, de Figueiredo and Larson, 2021, Facebook 2021, Gallup 2021, Lazarus et al. 2021, and Wouters et. A. 2021), while others were based on convenience samples (Anjorin et al. 2021, Asadi Faezi et al 2021, Bono 2021, Wong 2021), or a mix of methods (Solis-Arce et al. 2021 and Sallam 2021). There was also wide variety in the survey mode, including online, computer-assisted telephone, and face-to-face surveys. It is not possible to know the extent to which each of the above-mentioned potential sources of bias affect the comparisons of the findings from the HFPS with the others. For example, it was possible to match the Facebook survey results for the month and year of the HFPS for 20 countries, thus taking away any differences in the timing of the survey. Nevertheless, the HFPS estimates were still lower than those reported by Facebook (Figure 9, Panel D), and this could be due to differences in question-and-answer wording and/or the sample frame. 18 Figure 9: Comparison of share of population COVID-19 that is hesitant to take the COVID-19 vaccine, HFPS with other studies Notes: The y-axes indicate the share of the sample who reported not being willing or not sure if they will take the COVID-19 vaccine, as reported in the HFPS. The x-axes indicate the share of households in the survey that will definitely or probably not be vaccinated against COVID-19 from the following sources. Panel A: Figueiredo and Larson (2021). Panel B: Wouters et al. (2021). Panel C: Gallup (2021). Panel D: Facebook (2021). Even though the HFPS estimates are lower than other sources, the levels of COVID-19 vaccine hesitancy reported here are nonetheless higher than the levels of vaccine hesitancy reported for other vaccines. We compared the average country levels of COVID-19 vaccine hesitancy from this study with results from 2019 Global Monitor for 45 countries with observations in both studies. Overall, the levels of COVID-19 vaccine hesitancy reported here are higher than disagreement about the importance of vaccines for children (Figure 10, Panel A), concerns about the safety of vaccines (Figure 10, Panel B) and concerns about the efficacy of vaccines (Figure 10, Panel C). However, there are also similarities in the pattern: countries in Europe and Central Asia and some Sub-Sahara African countries reported the highest levels of vaccine hesitancy even before the pandemic, and the biggest reason for pre-COVID-19 vaccine hesitancy was around safety concerns; both findings are consistent with the COVID-19 vaccine hesitancy results derived from the HFPSs. Nevertheless, the higher levels of hesitancy reported for the COVID-19 vaccine suggest that respondents are more concerned about the COVID-19 vaccine than about childhood vaccines. The result is even more striking considering that the COVID-19 vaccine hesitancy is estimated rather conservatively in the HFPS, when compared against estimates from other sources. For example, a comparison between the Gallup (2021) estimates of COVID-19 vaccine hesitancy with 2019 Global Monitor estimates of general vaccine hesitancy (also collected by Gallup) for 38 countries included in this study showed that 34.7% of respondents on average across countries were 19 hesitant to be vaccinated against COVID-19, whereas from the Global Monitor only 4.4% did not agree vaccines were important for children, 17.2% did not agree vaccines are safe, and 15.2% did not agree they are effective. Possible reasons for the higher rates of COVID-19 vaccine hesitancy include that it is a very new vaccine that uses an innovative technology (mRNA), it was rapidly developed with a streamlined approval process, and may have unknown long-term effects. This is consistent with the reasons given by respondents, which are mostly around safety, and to a lesser extent efficacy. Figure 10: Comparison of COVID-19 Vaccine Hesitancy with pre-COVID sentiment about vaccines from 2019 Global Monitor, 45 countries Notes: The y-axes indicate the share of the sample who report either not willing or not sure if they will vaccinate, as reported in the HFPS. The x-axes indicate the share of the 2019 Global Monitor Report sample that reported they do not agree vaccines are important for children (Panel A), do not agree that vaccines are safe (Panel B), and do not agree vaccines are effective. Our findings that women, younger adults, and those with less education are more vaccine hesitant are largely consistent with results presented elsewhere. De Figueiredo and Larson (2021) and Solis Arce et al. (2021) reported that women were more vaccine hesitant than men. While Lazarus and colleagues found that men were slightly more hesitant than women, the gender difference was small. The findings with respect to age were also similar—the youngest adults were significantly less willing to be vaccinated than adults over 65 years of age (Lazarus et al. 2021, de Figuereido and Larson 2021). The low levels of vaccine hesitancy reported among the oldest adults provides an important opportunity for vaccine campaigns to target this demographic, especially given recent research that targeting vaccines to older age groups saves the most lives and is highly cost-effective (Cheikh, Spitz and Wilson 2022, Orangi et al. 2022). The findings on education are also consistent with other studies (Lazarus et al. 2021 and de Figueiredo and Larson 2021), although Solis-Arce and colleagues (2021) reported mixed results. Knowing that citizens who are younger and with less formal education are the most likely to be vaccine hesitant can help vaccination campaigns target these population groups. Although we hypothesized that confidence in the government in general would be negatively associated with vaccine hesitancy, this association was not statistically significant in the regression analysis. This is largely because trust indicators are only available at the national level, making the estimates imprecise, but it is also consistent with other studies. Kanyanda and colleagues (2021) considered a measure of trust in the government’s management of the COVID-19 crisis for Malawi and did not find a significant association after controlling for other factors (although it was significant in bivariate regressions). De 20 Figueiredo and Larson (2021) found a positive association between belief the government was handling the crisis well and willingness to be vaccinated in a sample of 32 countries with various income levels. Lazarus and colleagues reported the same, but only in bivariate odds ratios. Future analysis is needed to investigate this relationship, preferably using an individual level measure of trust in government rather than the national-level measure used in this analysis. With respect to our results on changes over time, which showed that changes were minimal in all countries except Iraq, Malawi, and Uzbekistan, where we report relatively large increases. It is difficult to identify the factors that led to increased hesitancy in these cases. The larger body on changes over time in vaccine hesitancy during the same timeframe as the HFPS data were collected is mostly from high- income countries and shows mixed results: increased rates of vaccine hesitancy reported in Australia, Canada and the United States between April and December 2020 (To et al. 2021, Lavoie et al. 2022, Szilagyi et al. 2020), rates remained stable in Italy between June 2020 and January 2021 (Basio et al. 2022) and decreased in Greece between November 2020 and May 2021 (Sypsa et al. 2022). The main reasons reported in the HFPS regarding COVID-19 vaccine hesitancy, which mainly revolve around safety and to a lesser degree efficacy, are consistent with the few studies that report reasons for COVID-19 vaccine hesitancy in low- and middle-income countries and with the main reasons for vaccine hesitancy historically. Solis Arce and colleagues (2021) also reported concerns about side effects to be the most frequently expressed reason in their sample of low- and middle-income countries. Facebook survey results for the same month were available for 21 countries, although reasons for not getting vaccinated were collected separately for respondents who would “definitely not” and “probably not” take the COVID- 19 vaccine and multiple reasons were collected. Nevertheless, they show that concerns about safety (worded as “concern about side effects” or “wait and see if it’s safe”) were the primary concern, followed in most countries by either concerns about efficacy or lack of trust in the government (Figure 11). Concerns have been raised about supply constraints – both lack of vaccine supply and lack of infrastructure to deliver the vaccines -- being the greatest barrier to COVID-19 vaccination scale-up in developing countries. This study did not find this to be a major concern for respondents (6% average across all surveys with the single answer option). As we noted above, however, the relative incidence of supply-side concerns in the HFPS data may be influenced by the framing of the question, as respondents may have interpreted the question whether they would get vaccinated “when the vaccine becomes available,” as indicating a time when current access constraints get resolved. Furthermore, there was a set of small countries in Latin America where over 10% of respondents cited supply constraints as their main concern, specifically “health center too far or hard to reach” or “I don't have time to go to get vaccinated.” This finding reinforces the importance citizens place on having easy and quick access, especially in Caribbean countries that are highly dependent on tourism. Counterindication – that is, believing one should not be vaccinated due to pre-existing health conditions – has been identified has a significant barrier to vaccination in Eastern Europe and Central Asia (World Bank, UNIFEF and JICA 2021). In Georgia, Tajikistan, Uzbekistan, and Kazakhstan, the most common reason for hesitancy was not being eligible (World Bank, UNIFEF and JICA 2021). Indeed, in many countries in that region, medical regulations recommend against vaccination for anyone with a preexisting medical condition. In Uzbekistan it was found that medical personnel interpreted the regulation even more strictly than it was written, regularly refusing to vaccinate persons with a pre-existing condition. For these countries, two policy responses may help to address this challenge: revising standard regulations (which 21 up until now have recommended against vaccination for people with a wide range pre-existing health conditions, including having had a recent surgery) about who is eligible to be vaccinated, and conducting awareness campaigns for both health care professionals and the public at large to stress the message that vaccination is recommended in nearly all cases, including among people with pre-existing health conditions (World Bank, UNIFEF and JICA 2021). Figure 11: Reason for “definitely not” (upper panel) or “probably not” (lower panel) taking the COVID-19 vaccine, Facebook data for month of HFPS) Respondent who will "definitely not" take the COVID-19 vaccine Belize Brazil Colombia Nigeria Costa Rica Lebanon Panama Honduras Peru El Salvador Guyana Respondent who will "probably not" take the COVID-19 vaccine Panama Costa Rica Uruguay Nicaragua Nigeria Iraq Argentina Guatemala Honduras Guyana Indonesia 0 50 100 150 200 250 300 Safety Efficacy Don't trust the government Dislike vaccines Don't need a COVID-19 vaccine Other people need it more than me Against my religious beliefs Concerned about the cost Other Notes: Y-axes indicate the share of survey respondents; multiple answers were allowed. Data source: Facebook (2021) for same month as reported in the HFPS. The month for each country is reported in Appendix Table 1. 22 4c. Limitations of this study One potential limitation of this study is that it is based on phone survey data which exclude respondents who did not have access to a phone. This method of data collection was necessary to collect information quickly during the early months of the COVID-19 pandemic while respecting local movement restrictions and minimizing the risk of COVID-19 transmission. Given that having a phone may be non-random, this might have created a risk that the results are representative only of the population with access to a phone. However, in the countries in this sample, access to mobile phones was high, and in countries where the sample was based on a previous survey, sampling weights were used in the analysis to correct for the biases resulting from non-random access to phones. A recent analysis using the same phone survey samples from four African countries demonstrated that the weighting procedures successfully minimized the selection bias in the phone surveys for a wide range of indicators (Ambel, Mcgee, and Tsegay 2021). Nevertheless, the variation in hesitancy estimates from different data sources, such as the phone surveys vs Facebook, also points to the need for methodological research that would explore further the biases introduced by phone surveys, or by online surveys, vis-à-vis face-to-face surveys. These types of experiments would aid the interpretation of findings in environments where data collection modalities are constrained, as was the case with the COVID-19 pandemic, and would also help guide researchers on issues related to sample corrections. A second limitation is that country surveys were collected at different times, and some surveys were collected before a vaccine was available in the country while others were collected after the vaccine was available. During data collection (and the pandemic more generally), influential events occurred, including adverse news about side effects of certain vaccines, which could have changed respondents’ intentions. Furthermore, there are multiple vaccine manufacturers and types, and the type of vaccine available or anticipated to become available in each country may influence levels of hesitancy. 4d. Conclusions and policy implications The main findings of this paper are that on average across countries, the level of COVID-19 vaccine hesitancy is approximately 20 percent and remained unchanged in most countries between late 2020 and the first half of 2021. These results suggest that COVID-19 vaccine hesitancy in developing countries, while less prevalent than previously thought, will be an important and enduring obstacle to recovery from the pandemic. The sizeable discrepancies with other studies of COVID-19 vaccine hesitancy suggest that measures of hesitancy depend greatly on the framing of the hesitancy questions and the nature of the sample. Knowing that people over the age of 65 years are the least likely to be hesitant provides an important opportunity to scale up vaccine rollout in this population group, which is at highest risk of severe disease and mortality from COVID-19. Although supply constraints have long been thought to be the main barrier to vaccination rollout in developing countries, these constraints have abated over time, particularly in middle- and high-income countries. Our results showed only a very small proportion of respondents reported barriers in access or lack of supply as a reason for not getting vaccinated, albeit this may be due, in part, to the framing of the questions. The most cited reasons for hesitancy were concerns about the safety of the vaccine, followed by concerns about its efficacy. At the same time, in many low-income countries, access issues remain important – the estimates in this study showed that about 20 percent of the population was hesitant to get a COVID-19 vaccine, but with only 1 in 5 individuals in low-income countries vaccinated to date, there is still a large share of the population who would get vaccinated if vaccines were made available to them, and alleviating access constraints will remain a policy priority in such contexts. 23 The estimates of COVID-19 vaccine hesitancy are higher than levels of hesitancy reported towards other vaccines, indicating the challenges in scaling up COVID-19 vaccination campaigns may be even greater than for other diseases. Given the high level, and the persistence of COVID-19 vaccine hesitancy, designing effective vaccination campaigns that address the key concerns underlying vaccine hesitancy will be very important. The results in this study aimed at helping guide policy makers in developing countries in their efforts to scale up national vaccination efforts. In particular, the findings point to the fact that it will be important to design vaccination campaigns that address concerns about safety, especially about side- effects, and that effectively reach younger adults. This will likely call for multi-pronged strategies, given the heterogeneity in information sources across different population groups. For example, in Belize, the government has deployed several initiatives that were successful in increasing vaccine uptake, including door-to-door educational sessions in remote villages via mobile units that were deployed prior to the arrival of vaccination teams, as well as information-sharing via social media and using radio to reach those who lack internet access (Margolies et al., 2022). Providing accurate information about vaccines to the population is very important, but so are questions related to who provides this information, and how it is provided. A World Bank program aimed at supporting developing countries in understanding and reducing vaccine hesitancy using behavioral science has found that personalized messaging is more effective than generic messaging, while certain messengers, such as health care workers and friends/family members, can be more effective than others, pointing to the importance of identifying such messengers within the social networks in groups with high hesitancy rates (Bidani et al., 2022). 24 References Africa CDC. 2021. COVID-19 vaccine perceptions: a 15-country study, 2021. Available: https://africacdc.org/download/covid-19-vaccineperceptions-a-15-country-study/ [Accessed 18 May 2021]. Anjorin, A.A., Odetokun, I.A., Abioye, A.I., Elnadi, H., Umoren, M.V., Damaris, B.F., Eyedo, J., Umar, H.I., Nyandwi, J.B., Abdalla, M.M. and Tijani, S.O., 2021. Will Africans take COVID-19 vaccination? PloS one, 16(12), p.e0260575. Ambel AA, Mcgee KR, Tsegay AH. 2021. Reducing Bias in Phone Survey Samples: Effectiveness of Reweighting Techniques using Face-to-Face Surveys as Frames. World Bank: Washington, DC. Faezi, N.A., Gholizadeh, P., Sanogo, M., Oumarou, A., Mohamed, M.N., Cissoko, Y., Sow, M.S., Keita, B.S., Baye, Y.A.M., Pagliano, P. and Akouda, P., 2021. Peoples’ attitude toward COVID-19 vaccine, acceptance, and social trust among African and Middle East countries. Health Promotion Perspectives, 11(2), p.171. Biasio LR, Bonaccorsi G, Lorini C, Mazzini D, Pecorelli S. Italian Adults' Likelihood of Getting COVID-19 Vaccine: A Second Online Survey. Vaccines (Basel). 2021 Mar 17;9(3):268. doi: 10.3390/vaccines9030268. PMID: 33802873; PMCID: PMC8002675. Bidani, B., Menon, R., Nguyen, S.N. Vakis, R., and Afif, Z. 2022. “Vaccine hesitancy: 10 lessons from chatbotting about COVID-19 in 17 countries.” Investing in Health Blog, Washington, DC: The World Bank. https://blogs.worldbank.org/health/vaccine-hesitancy-10-lessons-chatbotting-about-covid-19-17- countries Brewer, N.T., Chapman, G.B., Rothman, A.J., Leask, J. and Kempe, A., 2017. Increasing vaccination: putting psychological science into action. Psychological Science in the Public Interest, 18(3), pp.149-207. Bucci EM, Berkhof J, Gillibert A, Gopalakrishna G, Calogero RA, Bouter LM, Andreev K, Naudet F, Vlassov V. Data discrepancies and substandard reporting of interim data of Sputnik V phase 3 trial. Lancet. 2021 May 22;397(10288):1881-1883. Chapman, G.B., Li, M., Colby, H. and Yoon, H., 2010. Opting in vs opting out of influenza vaccination. Jama, 304(1), pp.43-44. Cheikh, N., J. Spitz and D. Wilson. 2022. Increase Vaccine Targeting and Uptake: lessons learned from Global COVID-19 Vaccine Implementation. Considerations for World Bank Task Teams Managing COVID- 19 Fast Track Facility Operations. Washington, D.C.: World Bank Group. Dong E, Du H, Gardner L. An interactive web-based dashboard to track COVID-19 in real time. Lancet Infect Dis 2020; 20(5): 533-4. https://coronavirus.jhu.edu/ accessed March 1, 2022. Facebook. 2021. COVID-19 Trends and Impact Survey. de Figueiredo A, Larson HJ. Exploratory study of the global intent to accept COVID-19 vaccinations. Communications medicine. 2021 Sep 9;1(1):1-0. Forman, R., Shah, S., Jeurissen, P., Jit, M. and Mossialos, E., 2021. COVID-19 vaccine challenges: What have we learned so far and what remains to be done? Health Policy, 125(5), pp.553-567. 25 Gallup. 2019. Wellcome Global Monitor – First Wave Findings. https://wellcome.org/reports/wellcome- global-monitor/2018#downloads-4d1c. Gallup. 2021. Gallup World Poll. https://www.gallup.com/178667/gallup-world-poll-work.aspx. Glassman, A., Kenny, C. and Yang, G., 2022. COVID-19 vaccine development and rollout in historical perspective. Center for Global Development. Feb.Gramacho WG, Turgeon M. 2021. When politics collides with public health: COVID-19 vaccine country of origin and vaccination acceptance in Brazil. Vaccine. May 6;39(19):2608-2612. doi: 10.1016/j.vaccine.2021.03.080. Guevara, S. Facebook COVID Symptoms Surveys - Hesitancy Rates (May). https://public.tableau.com/app/profile/stefani.guevara4262/viz/FacebookCOVIDSymptomsSurveys- HesitancyRatesMay/GlobalHesitancyRateMay2021. Accessed October 29, 2021. Hale T, Angrist N, Goldszmidt R, et al. 2022. A global panel database of pandemic policies (Oxford COVID- 19 government response Tracker). Nat Hum Behav 2021;5:529–38. Database accessed Feb 28, 2022. Hoogeveen, JG and G. Lopez-Acevedo, eds. 2021. Distributional Impacts of COVID-19 in the Middle East and North Africa Region. MENA Development Report;. Washington, DC: World Bank. © World Bank. https://openknowledge.worldbank.org/handle/10986/36618 License: CC BY 3.0 IGO. International Monetary Fund. 2021. World Economic Outlook Update. July 2021. Fault Lines Widen in the Global Recovery. Inglehart, R, C. Haerpfer, A. Moreno et al. 2022. Would Values Survey: Round Seven – Country Pooled Datafile Version: https://www.worldvaluessurvey.org/WVSOnline.jsp. (accessed April 21, 2022) Madrid: JD Systems Institute. Jarrett, C., Wilson, R., O’Leary, M., Eckersberger, E. and Larson, H.J., 2015. Strategies for addressing vaccine hesitancy–A systematic review. Vaccine, 33(34), pp.4180-4190. Lavoie K, Gosselin-Boucher V, Stojanovic J, Gupta S, Gagné M, Joyal-Desmarais K, Séguin K, Gorin SS, Ribeiro P, Voisard B, Vallis M, Corace K, Presseau J, Bacon S; iCARE Study Team. Understanding national trends in COVID-19 vaccine hesitancy in Canada: results from five sequential cross-sectional representative surveys spanning April 2020-March 2021. BMJ Open. 2022 Apr 5;12(4):e059411. doi: 10.1136/bmjopen-2021-059411. PMID: 35383087; PMCID: PMC8983402. Lazarus, JV, Ratzan SC , and Palayew A et al. 2021. A global survey of potential acceptance of a COVID-19 vaccine. Brief Communication. Nature Medicine 27: 225–228. https://doi.org/10.1038/s41591-020-1124- 9. MacDonald, NE, and SAGE Working Group on Vaccine Hesitancy. Vaccine hesitancy: Definition, Scope, and determinants. Vaccine. Volume 33, Issue 34, 2015, Pages 4161-4164 https://doi.org/10.1016/j.vaccine.2015.04.036 Mallapaty, S. 2022. Researchers Fear Vaccine Hesitancy in Developing Nations. Nature 601: 174-175. Margolies, A., de Hoop, J., Kim, P., Paffhausen, A.L. and Di Giorgio, L. 2022. “A shot in the arm: How to increase COVID-19 vaccine acceptance and uptake in the Caribbean?” Latin American and Caribbean: 26 Opportunities for All Blog, Washington, DC: The World Bank. https://blogs.worldbank.org/latinamerica/shot-arm-how-increase-covid-19-vaccine-acceptance-and- uptake-caribbean Masia, NA, Smerling, J. Kapfidze, T. et al. 2018. Vaccination and GDP Growth Rates: Exploring the Links in a Conditional Convergence Framework, World Development, Volume 103: 88-99. Omer, S.B., Benjamin, R.M., Brewer, N.T., Buttenheim, A.M., Callaghan, T., Caplan, A., Carpiano, R.M., Clinton, C., DiResta, R., Elharake, J.A. and Flowers, L.C., 2021. Promoting COVID-19 vaccine acceptance: recommendations from the Lancet Commission on Vaccine Refusal, Acceptance, and Demand in the USA. The Lancet, 398(10317), pp.2186-2192. Orangi, S., et al. 2022 (preprint). Epidemiological impact and cost-effectiveness analysis of COVID-19 vaccination in Kenya. https://efaidnbmnnnibpcajpcglclefindmkaj/https://www.medrxiv.org/content/10.1101/2022.04.21.222 74150v1.full.pdf Ritchie, H., E. Mathieu, L. Rodés-Guirao, et al.2020. - "Coronavirus Pandemic (COVID-19)". Published online at OurWorldInData.org. Retrieved from: 'https://ourworldindata.org/coronavirus' on March 8, 2022. Robinson, E., Jones, A., Lesser, I. and Daly, M., 2021. International estimates of intended uptake and refusal of COVID-19 vaccines: A rapid systematic review and meta-analysis of large nationally representative samples. Vaccine. Sallam M. COVID-19 Vaccine Hesitancy Worldwide: A Concise Systematic Review of Vaccine Acceptance Rates. Vaccines (Basel). 2021 Feb 16;9(2):160. doi: 10.3390/vaccines9020160. PMID: 33669441; PMCID: PMC7920465. Sanjay, PR, and Bloomberg. 2022. “The world’s biggest vaccine manufacturer has stopped making COVID jabs amid 200 million dose glut.” Fortune. April 22, 2022. https://fortune.com/2022/04/22/serum- institute-world-biggest-vaccine-manufacturer-stopped-covid-jabs-200-million-dose-glut/ Solís Arce, J.S., Warren, S.S., Meriggi, N.F. et al. COVID-19 vaccine acceptance and hesitancy in low- and middle-income countries. Nat Med 27, 1385–1394 (2021). https://doi.org/10.1038/s41591-021-01454-y. Sypsa V, Roussos S, Engeli V, Paraskevis D, Tsiodras S, Hatzakis A. Trends in COVID-19 Vaccination Intent, Determinants and Reasons for Vaccine Hesitancy: Results from Repeated Cross-Sectional Surveys in the Adult General Population of Greece during November 2020-June 2021. Vaccines (Basel). 2022 Mar 18;10(3):470. doi: 10.3390/vaccines10030470. PMID: 35335102; PMCID: PMC8950863. Szilagyi PG, Thomas K, Shah MD, Vizueta N, Cui Y, Vangala S, Kapteyn A. National Trends in the US Public's Likelihood of Getting a COVID-19 Vaccine-April 1 to December 8, 2020. JAMA. 2020 Dec 29;325(4):396–8. doi: 10.1001/jama.2020.26419. Epub ahead of print. PMID: 33372943; PMCID: PMC7772743. To QG, Stanton R, Khalesi S, Williams SL, Alley SJ, Thwaite TL, Fenning AS, Vandelanotte C. Willingness to Vaccinate against COVID-19 Declines in Australia, Except in Lockdown Areas. Vaccines (Basel). 2021 May 10;9(5):479. doi: 10.3390/vaccines9050479. PMID: 34068479; PMCID: PMC8150580. 27 Utazi CE, Pannell O, Aheto JMK, Wigley A, Tejedor-Garavito N, et al. (2022) Assessing the characteristics of un- and under-vaccinated children in low- and middle-income countries: A multi-level cross-sectional study. PLOS Global Public Health 2(4): e0000244. https://doi.org/10.1371/journal.pgph.0000244 Wong, L.P., Alias, H., Danaee, M., Ahmed, J., Lachyan, A., Cai, C.Z., Lin, Y., Hu, Z., Tan, S.Y., Lu, Y. and Cai, G., 2021. COVID-19 vaccination intention and vaccine characteristics influencing vaccination acceptance: a global survey of 17 countries. Infectious diseases of poverty, 10(1), pp.1-14. World Bank, UNIFEF and JICA 2021. Counterindication: now the key driver of COVID-19 vaccine hesitancy in Central Asia. World Bank. 2020. High Frequency Mobile Phone Surveys of Households to Assess the Impacts of COVID- 19: Overview (English). Washington, D.C.: World Bank Group. http://documents.worldbank.org/curated/en/703571588695361920/Overview World Bank. 2021. World Development Indicators. Washington, D.C.: The World Bank (producer and distributor). http://data.worldbank.org/data-catalog/world-development-indicators. Accessed March 12, 2022. World Bank. 2021. Contraindications: now the key driver of COVID-19 vaccine hesitancy in Central Asia. World Bank. 2022. Global Economic Prospects, January 2022. Washington, DC: World Bank. doi: 10.1596/978-1-4648-1758-8 World Health Organization. 2019. Ten threats to global health in 2019. https://www.who.int/news- room/spotlight/ten-threats-to-global-health-in-2019. World Health Organization. 2022. Global Excess Deaths Associated with COVID-19 (Modelled Estimates). Downloaded on September 8, 2022 at https://www.who.int/data/sets/global-excess-deaths-associated- with-covid-19-modelled-estimates. Wouters, O.J., Shadlen, K.C., Salcher-Konrad, M., Pollard, A.J., Larson, H.J., Teerawattananon, Y. and Jit, M., 2021. Challenges in ensuring global access to COVID-19 vaccines: production, affordability, allocation, and deployment. The Lancet, 397(10278), pp.1023-1034. 28 List of Appendix Tables Appendix Table 1: Description of 53 HFPS surveys Appendix Table 2a: Survey response options for COVID-19 vaccine hesitancy: distribution of countries by questionnaire type Appendix Table 2b: Survey response options for COVID-19 vaccine hesitancy: Harmonized answers by questionnaire type Appendix Table 3: Percent of population hesitant to take the COVID-19 vaccine, HFPS Appendix Table 4: Tukey’s tests of significance of differences across groups in percent of population hesitant to take the COVID-19 vaccine Appendix Table 5a: Correlates of vaccine hesitancy (Logit model, marginal effects) Appendix Table 5b: Correlates of vaccine hesitancy (Ordinal logit regression, marginal effects) Appendix Table 6: Percent of population hesitant to take the COVID-19 vaccine, 14 countries with multiple rounds of survey results (95% confidence intervals in parentheses). Appendix Table 7: Comparison of results on vaccine hesitancy from the HFPS and other studies 29 Appendix Table 1: Description of 53 HFPS surveys Country Survey Months Total Number of households Sample frame Antigua and Barbuda 06/2021 790 RDD Argentina 06/2021 1216 RDD Belize 06/2021 816 RDD Bolivia 05/2021 1272 RDD Brazil 08/2021 2166 RDD Bulgaria 07/2021 1000 RDD Burkina Faso 12/2020 1944 Previous survey Chile 06/2021 1212 RDD Colombia 06/2021 1221 RDD Congo, Dem. Rep. 12/2020 986 Previous Survey Congo, Rep. 12/2020, 03/2021 578, 1495 Previous Survey Costa Rica 06/2021 802 RDD Croatia 03/2021 1217 Non-survey list Dominica 06/2021 861 RDD Dominican Republic 06/2021 1205 RDD Ecuador 05/2021 1352 RDD El Salvador 06/2021 816 RDD Ethiopia 10/2020, 02/2021 2704, 2178 Previous Survey The Gambia 12/2020, 04/2021, 08/2021 1334, 1287, 1059 Previous Survey Georgia 01/2021, 03/2021, 06/2021 2033, 2100, 1936 RDD Guinea 11/2020 1334 Previous Survey Guatemala 06/2021 1206 RDD Guyana 06/2021 785 RDD Haiti 07/2021 2813 Non-Survey list Honduras 07/2021 1021 RDD Indonesia 11/2020, 03/2021 3953, 3555 Previous Survey Iraq 12/2020, 1614,1651,1378, Non-Survey list 01/2021,06/2021, 1297,1141 07/2021, 08/2021 Jamaica 06/2021 828 RDD Kazakhstan 02/2021, 05/2021, 06/2021 917, 1732, 1610 Previous Survey Kenya 03/2021 6730 Previous Survey 30 Country Survey Months Total Number of households Sample frame Lao PDR 03/2021 2153 RDD Lebanon 03/2021 5113 RDD Mali 01/2021 1884 Previous Survey Malawi 11/2020, 03/2021, 04/2021 1589, 1549,1338 Previous Survey Malaysia 06/2021 2210 RDD Mexico 06/2021 2624 RDD Mongolia 12/2020 1147 Previous Survey Nicaragua 06/2021 833 RDD Nigeria 02/2021, 10/2020 1699, 1762 Previous Survey Panama 06/2021 815 RDD Paraguay 06/2021 1076 RDD Peru 06/2021 1210 RDD Philippines 12/2020, 05/2021 1805, 2122 Non-survey list Saint Lucia 06/2021 835 RDD Sierra Leone 11/2020 1198 Previous Survey Sudan 03/2021 2545 Non-Survey list Tajikistan 05-08/2021 232 Previous Survey Thailand 05/2021 1786 RDD Uganda 11/2020, 02/2021 2135, 2121 Previous Survey Uruguay 06/2021 816 RDD Uzbekistan 04-06/2021 1496, 1356, 1300 Previous Survey Vietnam 01/2021 3940 Previous Survey Zimbabwe 12/2020 1227 Previous Survey 31 Appendix Table 2a: Survey response options for COVID-19 vaccine hesitancy: distribution of countries by questionnaire type Questionnai Questionnaire Questionnai Questionnaire Questionnaire Questionnaire Questionnaire Questionnaire Questionnaire Questionnaire Questionnaire Questionnaire Questionnaire re type 12 re type 1 type 2 type 3 type 4 type 6 type 7 type 8 type 9 type 10 LAC type 11 type 5 Congo, Croatia Kenya Philippines Lebanon Democratic Guinea Malaysia Mali Gambia, The Lao PDR Cambodia Argentina Kazakhstan Republic Republic of Antigua and Thailand Indonesia Sudan Burkina Faso Mongolia Kyrgyzstan Congo Barbuda Sierra Djibouti Uganda Belize Uzbekistan Leone Ethiopia Bolivia Malawi Brazil Nigeria Chile Colombia Costa Rica Dominica Dominican Republic Ecuador Guatemala Guyana Honduras Haiti Jamaica Mexico Nicaragua Panama Peru Paraguay El Salvador Uruguay 32 Appendix Table 2b: Survey response options for COVID-19 vaccine hesitancy: Harmonized answers by questionnaire type Answer Quest type 12 Quest type 1 Quest type 2 Quest type 3 Quest type 4 Quest type 5 Quest type 6 Quest type 7 Quest type 8 Quest type 9 Quest type 10 Quest LAC Quest type 11 categories 1. Efficacy I do not I do not It may not I don’t think I don’t think think the I don’t think think they work COVID-19 coronavirus vaccine COVID-19 are vaccines vaccines are I don’t think would work vaccines effective, I don’t think effective/wo effective/wo vaccines I don’t think I don’t think against effective/wo that they it will work rk rk work it will work it will work COVID rk work I don’t think the vaccines available in my country are effective 2 Safety I fear the Health risks (includes unforeseen side future effects) I am worried I am worried negative side I am worried The risk of about the about the effects from about the vaccinating is I don’t think safety of the Safety of the vaccine I don’t think the COVID- safety of the I don’t think it higher than it will be safe vaccine vaccine safety it will be safe 19 vaccine vaccine is safe the benefits I don’t I am worried think it’s about the I have heard safe, I am worried I am worried side effects I am worried I’m worried the vaccine I am worried because of about the about the of the about the about side has negative about the the side side effects side effects vaccine side effects effects side effects side effects effects 3. Perceive I am not I am not I am not I am not at I am not self as low enough at I am strong, I enough at enough at enough risk I believe I do I am not at worried about risk risk of never got risk of risk of of not need to risk of COVID-19 contracting any disease contracting contracting contracting be getting COVID-19 in the past coronavirus COVID-19 covid-19 vaccinated covid-19 4.Dislike I am I am against vaccines in I am against I am against I am against General I don’t trust In general, I I am against I am against against Negative past vaccines in general vaccines in vaccines in vaccines in distrust of the vaccines don’t trust the vaccine vaccines in vaccines in experiences general general general general vaccines in general vaccines in general general general with vaccines 5 Religious I have religious Personal or It is against Religious Religious reasons/con It is against It’s against Religious Religious religious my religion reasons reasons cerns my religion. my religion reasons reasons beliefs I am concerned about its 33 Answer Quest type 12 Quest type 1 Quest type 2 Quest type 3 Quest type 4 Quest type 5 Quest type 6 Quest type 7 Quest type 8 Quest type 9 Quest type 10 Quest LAC Quest type 11 categories halal certification 6 I am Concerned concerned about I’m worried I am worried I am worried I am worried I am worried of being getting to get of getting of getting to get to get infected covid at infected with infected with infected with infected with infected with with covid- facility COVID-19 at COVID-19 at coronavirus COVID-19 at COVID-19 at 19 at the the health the health at the health the health the health health facility facility facility facility. facility center 7 Supply It does not /access Health Health suit me (will Health barrier facility is too facility too be given too facility too There is Health far or too far or too far or far or too shortage of I don’t know center too hard to get hard to get difficult to hard to get vaccines in how to access far or hard to to find) to the country the vaccine to reach I don’t have It will take time to get too long to vaccinated/ get It will take vaccinated/ I There is no I am not the I don’t have too long to don’t have vaccination priority group time to go get time to get center near to get the to get vaccinated vaccinated my place vaccine vaccinated 8 Prefer I prefer to I already had natural build COVID-19 and immunity immunity do’'t need a (includes against I already vaccine already I believe that COVID-19 had covid- had COVID- the remedies naturally by 19 / I no 19) I already had natural or I already had having the longer coronavirus traditional COVID-19 disease need it It is better to leave nature take its course; the COVID symptoms are mostly light 34 Answer Quest type 12 Quest type 1 Quest type 2 Quest type 3 Quest type 4 Quest type 5 Quest type 6 Quest type 7 Quest type 8 Quest type 9 Quest type 10 Quest LAC Quest type 11 categories 9 Distrust I do’'t trust I do’'t trust (of35overn the the mentt/pha government/ pharmaceutic rmaceutica to the I do not trust al industry l Be wary of organization pharmaceuti Lack of trust industry/in international who gives cal to producers ternational community the vaccine companies of vaccines community Lack of trust I do’'t trust ) to heath the I do not trust system or government the healthcare government providers I do not trust the COVID- 19 vaccines I heard the vaccine is meant to control population growth I do not believe in COVID-19 10 Not I have Counter- eligible / underlying indication or conterindic health prolonged ation conditions medical and I believe discharge taking the vaccine will I am not make it eligible to get worse the vaccine 11 Lack of I do not have knowledge enough about information COVID 19 about the Vaccine vaccine 35 Answer Quest type 12 Quest type 1 Quest type 2 Quest type 3 Quest type 4 Quest type 5 Quest type 6 Quest type 7 Quest type 8 Quest type 9 Quest type 10 Quest LAC Quest type 11 categories I did not know that a vaccine exists against COVID-19 12 Other Everyday stresses are overwhelmin g to think about Other Other Other Some other Other other, getting Others Other (specify) (specify) reason other Others (Specify) specify) vaccinated (specify) Other, specify (specify) I will wait till more people are vaccinated No one in I am not sure my I will get the neighborhoo vaccine I d got it want 36 Appendix Table 3: Percent of population hesitant to take the COVID-19 vaccine, HFPS Average across countries Percent (%) hesitant to take the COVID-19 Confidence Intervals vaccine All countries 20.0 [ 17.24 , 22.67 ] East Asia & Pacific 26.2 [ 21.38 , 31.02 ] Europe & Central Asia 58.8 [ 55.04 , 62.56 ] Latin America & Caribbean 8.0 [ 6.45 , 9.47 ] Middle East & North Africa 47.4 [ 38.80 , 55.96 ] Sub-Saharan Africa 15.5 [ 11.81 , 19.19 ] Low income countries 14.6 [ 7.82 , 21.39 ] Lower middle income countries 27.7 [ 23.78 , 31.65 ] Upper middle income countries 12.7 [ 9.77 , 15.61 ] High income countries 5.9 [ 3.40 , 8.36 ] Urban 17.6 [ 14.60 , 20.75 ] Rural 23.2 [ 19.83 , 26.64 ] Male 17.3 [ 15.02 , 19.63 ] Female 22.5 [ 18.77 , 26.21 ] Head of household 10.8 [ 12.82 , 18.53 ] Non-Head of household 9.4 [ 17.37 , 26.54 ] No education 22.8 [ 15.36 , 30.31 ] Any Primary 19.8 [ 15.80 , 23.78 ] Any Secondary 19.0 [ 14.64 , 23.42 ] Any Tertiary 13.7 [ 10.56 , 16.83 ] Age 34 and younger 20.3 [ 17.29 , 23.34 ] Age 35–- 64 20.1 [ 17.34 , 22.84 ] Ages 65 and older 17.8 [ 13.70 , 21.80 ] Antigua and Barbuda 24.5 [ 22.51 , 26.49 ] Argentina 10.1 [ 8.66 , 11.53 ] Belize 28.4 [ 26.10 , 30.64 ] Bolivia 24.3 [ 22.47 , 26.04 ] Brazil 3.1 [ 2.15 , 4.04 ] Bulgaria 66.2 [ 59.48 , 72.90 ] Burkina Faso 23.6 [ 16.60 , 30.53 ] Chile 3.4 [ 2.22 , 4.57 ] Colombia 11.2 [ 8.91 , 13.51 ] Congo, Dem. Rep. 61.2 [ 61.23 , 61.23 ] Congo, Rep. 13.0 [ 11.20 , 14.79 ] 37 Average across countries Percent (%) hesitant to take the COVID-19 Confidence Intervals vaccine Costa Rica 11.9 [ 9.66 , 14.22 ] Croatia 33.2 [ 29.21 , 37.25 ] Dominica 35.9 [ 32.63 , 39.14 ] Dominican Republic 5.0 [ 2.25 , 7.81 ] Ecuador 19.4 [ 15.54 , 23.21 ] El Salvador 8.2 [ 6.91 , 9.44 ] Ethiopia (excludes Eritrea) 3.5 [ 1.81 , 5.09 ] Fm Sudan 23.7 [ 19.19 , 28.11 ] Gambia, The 33.1 [ 27.44 , 38.70 ] Georgia 65.2 [ 59.62 , 70.72 ] Guatemala 29.6 [ 25.09 , 34.05 ] Guinea 20.2 [ 15.74 , 24.61 ] Guyana 20.0 [ 15.75 , 24.23 ] Haiti 58.1 [ 55.31 , 60.86 ] Honduras 13.9 [ 10.31 , 17.51 ] Indonesia 21.4 [ 17.71 , 25.07 ] Iraq 47.4 [ 38.81 , 55.99 ] Jamaica 50.6 [ 47.11 , 54.03 ] Kazakhstan 75.3 [ 72.39 , 78.18 ] Kenya 17.9 [ 17.11 , 18.74 ] Lao PDR 13.0 [ 9.70 , 16.25 ] Lebanon 32.2 [ 24.56 , 39.84 ] Malawi 29.3 [ 25.03 , 33.54 ] Malaysia 25.7 [ 22.37 , 29.10 ] Mali 21.1 [ 13.14 , 29.03 ] Mexico 6.2 [ 4.85 , 7.55 ] Mongolia 19.3 [ 14.13 , 24.38 ] Nicaragua 18.7 [ 16.04 , 21.40 ] Nigeria 16.6 [ 12.27 , 20.84 ] Panama 13.3 [ 11.12 , 15.49 ] Paraguay 15.3 [ 10.71 , 19.79 ] Peru 10.6 [ 7.55 , 13.59 ] Philippines 53.6 [ 46.28 , 60.83 ] Sierra Leone 21.5 [ 17.31 , 25.66 ] St. Lucia 43.2 [ 38.46 , 47.93 ] Tajikistan 26.7 [ 18.32 , 35.03 ] Thailand 36.6 [ 31.25 , 42.00 ] Uganda 11.6 [ 6.51 , 16.61 ] Uruguay 9.1 [ 5.94 , 12.17 ] Uzbekistan 54.6 [ 50.02 , 59.18 ] Vietnam 15.9 [ 11.72 , 20.14 ] Zimbabwe 15.8 [ 10.42 , 21.14 ] 38 39 Appendix Table 4: Tukey’s tests of significance of differences across groups in percent of population hesitant to take the COVID-19 vaccine Group Comparison Difference Std. t-score p- Signi- Error value ficance Gender Female vs Male 0.052 0.022 2.316 0.021 ** Household Head vs Non-Head -0.063 0.026 -2.398 0.016 ** Status Age Ages 65 and older vs Age 35 - 64 -0.023 0.025 -0.937 0.349 Ages 65 and older vs Age 34 and -0.026 0.026 -0.994 0.320 younger Age 35 - 64 vs Age 34 and younger -0.002 0.021 -0.107 0.543 Education Any Primary vs Any Secondary 0.008 0.030 0.250 0.599 Any Primary vs Any Tertiary 0.061 0.026 2.355 0.019 ** Any Primary vs No Education -0.030 0.043 -0.705 0.481 Any Secondary vs Any Tertiary 0.053 0.028 1.941 0.052 * Any Secondary vs No Education -0.038 0.044 -0.860 0.390 Any Tertiary vs No Education -0.091 0.041 -2.211 0.027 ** Sector Rural vs Urban 0.056 0.023 2.375 0.018 ** Region EAP vs ECA -0.326 0.031 -10.450 0.000 *** EAP vs LAC 0.182 0.026 7.077 0.000 *** EAP vs MENA -0.212 0.050 -4.219 0.000 *** EAP vs SSA 0.107 0.031 3.454 0.001 *** ECA vs LAC 0.508 0.021 24.579 0.000 *** ECA vs MENA 0.114 0.048 2.390 0.017 ** ECA vs SSA 0.433 0.027 16.103 0.000 *** LAC vs MENA -0.394 0.044 -8.870 0.000 *** LAC vs SSA -0.075 0.020 -3.705 0.000 *** MENA vs SSA 0.319 0.048 6.691 0.000 *** 40 Income Group HIC vs LIC -0.087 0.037 -2.368 0.018 ** HIC vs LMIC -0.218 0.024 -9.202 0.000 *** HIC vs UMIC -0.068 0.020 -3.481 0.001 *** LIC vs LMIC -0.131 0.037 -3.559 0.000 *** LIC vs UMIC 0.019 0.024 0.808 0.419 LMIC vs UMIC 0.150 0.025 6.010 0.000 *** Significance levels: *** for p-values <= 0.01, ** for 0.01 < p-values <= 0.05, * for 0.05 < p-values <= 0.1 41 Appendix Table 5a: Correlates of vaccine hesitancy (Logit model, marginal effects) Dep. Var: Hesitancy (1) (2) (3) (4) (5) (6) Male -0.027* -0.028* -0.027* -0.027* -0.033** -0.040*** (0.013) (0.013) (0.013) (0.012) (0.012) (0.011) Head of HH 0.013 0.009 0.021 -0.001 0.014 0.007 (0.014) (0.014) (0.013) (0.016) (0.015) (0.012) Education of respondent (ref – No education 0.130*** 0.131*** 0.122*** 0.142*** 0.133** 0.153*** (0.036) (0.037) (0.033) (0.040) (0.044) (0.040) Any primary 0.068*** 0.069** 0.067*** 0.073*** 0.059** 0.079*** (0.020) (0.021) (0.020) (0.020) (0.022) (0.023) Any secondary 0.069*** 0.070*** 0.068*** 0.070*** 0.064*** 0.067*** (0.011) (0.011) (0.011) (0.011) (0.011) (0.012) Age group (ref. -- 34 and younger) Working age (35-64) -0.047*** -0.048*** -0.042** -0.050*** -0.046** -0.049*** (0.014) (0.013) (0.015) (0.013) (0.015) (0.012) Retirement age (65+) -0.103*** -0.102*** -0.096*** -0.101*** -0.103*** -0.102*** (0.024) (0.025) (0.024) (0.023) (0.025) (0.024) Rural area 0.016 0.014 0.015 0.008 0.012 0.011 (0.019) (0.018) (0.017) (0.015) (0.015) (0.011) Region (ref – LAC) EAP 0.189** 0.192* 0.222** 0.095 0.264* 0.188 (0.067) (0.078) (0.077) (0.069) (0.130) (0.130) ECA 0.313*** 0.315*** 0.295*** 0.328*** 0.313*** 0.295*** (0.055) (0.057) (0.056) (0.059) (0.040) (0.062) MNA 0.208*** 0.212*** 0.245*** 0.216*** 0.282*** 0.369*** (0.037) (0.040) (0.062) (0.032) (0.059) (0.051) SSA 0.074 0.081 0.070 -0.022 0.047 0.022 (0.075) (0.077) (0.070) (0.105) (0.083) (0.104) Income group (ref. – LIC) LMIC -0.068 -0.065 -0.045 -0.088 -0.054 -0.071 (0.122) (0.119) (0.112) (0.123) (0.097) (0.069) UMIC -0.047 -0.038 -0.066 -0.095 -0.035 -0.137 (0.149) (0.148) (0.130) (0.150) (0.116) (0.087) HIC -0.173 -0.164 -0.190 -0.223 -0.202 -0.309*** (0.163) (0.160) (0.143) (0.164) (0.130) (0.090) Survey month (ref -- Nov 2020 - Jan March - May 2021 -0.069 -0.060 -0.094 -0.056 -0.095 -0.083 (0.095) (0.108) (0.103) (0.098) (0.104) (0.095) June - August 2021 -0.029 -0.025 -0.024 -0.052 -0.057 -0.055 (0.088) (0.098) (0.100) (0.087) (0.092) (0.075) New COVID-19 cases per million, terciles (ref. – bottom tercile) Cases (middle tercile) -0.017 0.026 (0.067) (0.060) Cases (top tercile) -0.019 -0.016 (0.057) (0.047) Oxford stringency index terciles (ref. – bottom tercile) Stringency (middle tercile) 0.065 0.134 (0.072) (0.070) Stringency (top tercile) -0.008 0.027 (0.076) (0.068) WHO excess deaths due to COVID-19 terciles (ref.- bottom tercile) Excess deaths (middle tercile) -0.117* -0.109 (0.052) (0.072) 42 Dep. Var: Hesitancy (1) (2) (3) (4) (5) (6) Excess deaths (top tercile) -0.069 0.040 (0.072) (0.108) Confidence in government index tercile (ref.- top tercile) Confidence in government 0.048 0.070 (0.053) (0.057) Confidence in government 0.024 0.053 (0.097) (0.103) Pseudo R-squared 0.0745 0.0748 0.0794 0.0813 0.0869 0.1067 N 65088 65088 65088 65088 65088 65088 Notes: Weighted logit regressions. Marginal effects reported. Standard errors clustered at country level. *, **, *** indicates significance at the 95%, 99%, and 99.9% level. 43 Appendix Table 5b: Correlates of vaccine hesitancy (Ordinal logit regression, marginal effects) Dep. Var.: Would you get the vaccine? No Not sure Yes Male -0.025** -0.007** 0.036** (0.01) (0.002) (0.011) Head of HH 0.011 0.001 -0.006 (0.012) (0.002) (0.011) Education of respondent (ref – Tertiary) No education 0.119*** 0.022** -0.156*** -0.028 (0.007) (0.038) Any primary 0.057** 0.013** -0.080*** -0.018 (0.004) (0.022) Any secondary 0.058*** 0.012*** -0.068*** (0.009) (0.003) (0.010) Age group (ref. -- 34 and younger) Working age (35-64) -0.040*** -0.009*** 0.049*** (0.010) (0.002) (0.012) Retirement age (65+) -0.079*** -0.018*** 0.097*** (0.020) (0.005) (0.025) Rural area 0.010 0.002 -0.012 (0.009) (0.002) (0.011) Region (ref – LAC) EAP 0.149 0.034 -0.183 (0.093) (0.021) (0.113) ECA 0.238*** 0.055*** -0.293*** (0.044) (0.016) (0.058) MNA 0.288*** 0.066*** -0.354*** (0.042) (0.011) (0.049) SSA 0.015 0.004 -0.019 (0.082) (0.019) (0.101) Income group (ref. – LIC) LMIC -0.074 -0.017 0.091 (0.055) (0.012) (0.067) UMIC -0.109 -0.025 0.134 (0.069) (0.017) (0.085) HIC -0.241*** -0.056** 0.297*** (0.071) (0.019) (0.088) Survey month (ref -- Nov 2020 - Jan 2021) March - May 2021 -0.054 -0.011 0.065 (0.081) (0.016) (0.098) June - August 2021 -0.046 -0.010 0.055 (0.065) (0.012) (0.077) New COVID-19 cases per million, terciles (ref. – top tercile) Cases (middle tercile) 0.016 0.003 -0.019 (0.048) (0.010) (0.058) Cases (bottom tercile) -0.028 -0.006 0.034 (0.037) (0.008) (0.045) Oxford stringency index terciles (ref. – top tercile) Stringency (middle tercile) 0.089 0.020 -0.110 (0.057) (0.013) (0.070) Stringency (bottom tercile) 0.012 0.003 -0.015 (0.041) (0.011) (0.052) 44 Dep. Var.: Would you get the vaccine? No Not sure Yes WHO excess deaths due to COVID-19 terciles (ref.- bottom tercile) Excess deaths (middle tercile) -0.075 -0.019 0.094 (0.057) (0.015) (0.071) Excess deaths (top tercile) 0.057 0.010 -0.067 (0.096) (0.015) (0.112) Confidence in government index tercile (ref.- top tercile) Confidence in government (middle tercile) 0.052 0.016 -0.069 (0.041) (0.013) (0.054) Confidence in government (bottom tercile) 0.040 0.013 -0.053 (0.070) (0.023) (0.093) R-squared 0.0866 N 65088 Notes: Weighted ordinal logit regressions. Categories ranked from no to not sure to yes. Marginal effects reported. Standard errors clustered at country level. *, **, *** indicates significance at the 95%, 99%, and 99.9% level. 45 Appendix Table 6: Percent of population non-hesitant to take the COVID-19 vaccine, 14 countries with multiple rounds of survey results (95% confidence intervals in parentheses). country Oct- Nov- Dec- Jan- Feb- Mar- Apr- May- Jun- Jul- Aug- 2020 2020 2020 2021 2021 2021 2021 2021 2021 2021 2021 Burkina Faso 0.79 0.76 (0.73, (0.67, 0.86) 0.86) Congo, Rep. 0.84 0.87 (0.75, (0.81, 0.94) 0.93) Ethiopia 0.98 0.97 (0.91, (0.89, 1.00) 1.00) Gambia, The 0.65 0.55 0.67 (0.58, (0.49, (0.60, 0.71) 0.60) 0.74) Georgia 0.36 0.29 0.35 (0.33, (0.27, (0.32, 0.38) 0.32) 0.38) Indonesia 0.79 0.79 (0.73, (0.72, 0.85) 0.85) Iraq 0.66 0.50 0.45 0.47 0.53 (0.61, (0.45, (0.40, (0.42, (0.47, 0.72) 0.54) 0.50) 0.52) 0.58) Kazakhstan 0.23 0.32 0.25 (0.17, (0.27, (0.21, 0.28) 0.37) 0.29) Malawi 0.83 0.52 0.71 (0.76, (0.47, (0.64, 0.89) 0.58) 0.78) Nigeria 0.86 0.83 (0.80, (0.77, 0.93) 0.90) Philippines 0.44 0.46 (0.40, (0.42, 0.48) 0.51) Tajikistan 0.73 0.79 0.80 0.73 (0.61, (0.67, (0.65, (0.57, 0.84) 0.92) 0.95) 0.90) Uganda 0.84 0.88 (0.79, (0.83, 0.90) 0.94) Uzbekistan 0.55 0.50 0.45 (0.51, (0.46, (0.41, 0.60) 0.54) 0.49) 46 Appendix Table 7: Comparison of results on vaccine hesitancy from the HFPS and other studies Comparison with de Figueiredo and Larson (2021) HFPS de Figueiredo and Larson Argentina 10.1% 24.3% Brazil 3.1% 17.0% Chile 3.4% 27.9% Croatia 33.2% 58.5% Ecuador 19.4% 20.3% Indonesia 21.3% 17.1% Lebanon 32.2% 55.9% Malaysia 25.8% 13.9% Mexico 6.2% 18.0% Nigeria 15.2% 35.9% Paraguay 15.3% 48.5% Peru 10.5% 28.3% Vietnam 15.9% 3.2% Average 16.3% 28.4% Comparison with Wouters et al. 2021 HFPS Wouters et al. Argentina 10.1% 24.0% Brazil 3.1% 12.0% Chile 3.4% 28.0% Croatia 33.2% 59.0% Ecuador 19.4% 20.0% Indonesia 42.6% 17.0% Lebanon 32.2% 56.0% Mexico 6.2% 18.0% Nigeria 30.3% 36.0% Paraguay 15.3% 49.0% Peru 10.5% 28.0% Vietnam 15.9% 2.0% Average 18.5% 29.1% Comparison with Gallup (2021) Country HFPS Hesitancy Gallup Argentina 10.1% 37.0% Bolivia 24.2% 35.0% Brazil 3.1% 30.0% Burkina Faso 22.2% 44.0% Chile 3.4% 40.0% Colombia 11.2% 30.0% Congo, Dem. Rep. 61.2% 61.2% Congo, Rep. 14.4% 48.0% 47 Costa Rica 11.9% 26.0% Croatia 33.2% 57.0% Dominican Republic 5.0% 35.0% Ecuador 19.4% 28.0% El Salvador 8.1% 25.0% Ethiopia 2.8% 16.0% Georgia 66.8% 44.0% Guatemala 29.6% 29.0% Guinea 20.1% 46.0% Honduras 13.9% 30.0% Indonesia 21.3% 30.0% Iraq 47.8% 39.0% Jamaica 50.6% 68.0% Kenya 17.9% 27.0% Lao PDR 12.9% 16.0% Lebanon 32.2% 57.0% Malaysia 25.8% 28.0% Mali 21.1% 51.0% Mexico 6.2% 25.0% Mongolia 19.2% 39.0% Nicaragua 18.8% 13.0% Nigeria 15.2% 42.0% Paraguay 15.3% 47.0% Peru 10.5% 27.0% Philippines 54.9% 49.0% Tajikistan 23.7% 36.0% Thailand 36.6% 39.0% Uganda 13.6% 38.0% Uruguay 9.1% 39.0% Uzbekistan 49.8% 30.0% Vietnam 15.9% 19.0% Zimbabwe 15.8% 27.0% Average 22.4% 36.2% Comparison with Facebook (2021) Country HFPS FB Argentina 10.1% 21.5% Belize 28.4% 48.0% Brazil 3.1% 25.8% Chile 3.4% 43.0% Colombia 11.2% 15.7% Costa Rica 11.9% 17.1% Ecuador 19.4% 17.1% 48 El Salvador 8.1% 15.8% Guatemala 29.6% 18.6% Guyana 20.0% 54.4% Honduras 13.9% 13.7% Indonesia 21.3% 19.2% Iraq 47.8% 52.8% Lebanon 32.2% 22.3% Mexico 6.2% 13.0% Nicaragua 18.8% 37.4% Panama 13.3% 24.9% Paraguay 15.3% 16.2% Peru 10.5% 12.2% Uruguay 9.1% 50.8% Average 16.7% 27.0% 49