Contextualising planned behaviours to the vaccination against COVID-19 in the European Union

Abstract

The article targets the reasons that are behind behaviour orientation in the vaccination process against COVID-19. The data we are using come from the Flash Eurobarometer 494, collected in May 2021. The key dependent variable puts together vaccination intentions (soon, later on in 2021, undecided, later, never) and the fact of being vaccinated or not. A multivariate and multilevel analysis confirms the validity of an extended theory of planned behaviour in explaining the orientation to the vaccination against COVID-19. The space patterning of the behaviours is highly marked by differences among Old versus the New Member States of the European Union, clusters of countries, urban versus rural areas, and also by a function of trust in relevant institutions, and customs of using vaccination to cope with different diseases as an adult. New questions and hypotheses are generated by multiple comparisons.

Introduction

Why have some people been vaccinated against COVID-19 (C19) or planned the action soon, sometime during the year, later, or never? Is there a kind of motivational continuum between very soon and never? Do motivations differ from place to place? And if so, how much and for what reasons? Questions of this kind would deserve to be assumed because only by comparing the answers given in different contexts we could, very likely, advance in understanding the theme.

The sociocultural conditions in the C19 vaccination have been addressed through case studies at the national level (Cerda and García 2021) or by comparing the situations between neighbouring or similar countries by their cultural profile (Trent et al., 2021; Wollast et al., 2021). We will extend such an approach to 23 countries in the European Union (EU), using the data of a Flash Eurobarometer survey, from the perspective of planned behaviour theory (Wolff, 2021; Wollast et al., 2021). In line with the standard approach of the theory of planned behaviour (TPB), vaccination practices are well approximated by intentions, and these derive from attitude, pressure from the significant other and perceived control of targeted behaviour (Ajzen, 1991, 2011). In this kind of approach, the key issue is related to the right specification of the prediction model. The theoretical model became better specified in its development from the theory of reasoned action (Fishbein, 1967) which stipulated that attitudes and subjective norms could predict intentions that generate behaviours. Perceived behavioural control was added as a key predictor in the TPB.

The model of TPB is, in most applications, a universalist one, in which the context is frequently blurred. In addition, the intention of action and perceived behavioural control (Ajzen, 1991) tend to become unique intermediate variables through which all other factors act on the behaviour. The need for standardisation of context factors is especially emphasised in comparative research on different spatial units (countries, regions, communities). The extended TPB (Conner and Armitage, 1998) is a new sequence in the process of a better specification of the theory to explain planned behaviours. Initially, it targeted extending TPB by adding, as predictors of intentions to act, past behaviour/habits, belief salience, moral norms, self-identity, and affective beliefs. Different versions of extended TPB developed the function of action content and its context.

This analysis here supports a variant of extended TPB in the field to include the role of the context of place-time and social space in determining the behaviours of vaccination against C19. Highlighting such contexts at the level of the comparative territorial units can lead to questions and assumptions related to other factors that condition the planned behaviours. Like any contextualisation, it can also favour the substantiation of policies in the field of reference (Gaston, 2017).

We expand the TPB by proposing an integrated approach by (1) measuring the dependent variable, (2) the specification of the predictors, and (3) targeting levels of analysis. The dependent variable here is the orientation of the interviewees to vaccinate or not against C19. The usual approaches of TPB are focused on the factors that directly determine the intention for a specified planned behaviour or on the impact of intention on behaviour (see, for example, Buhmann and Brønn, 2018; Conner and Armitage, 1998). Our first integration is to consider behavioural intention and past behaviour of the uptake of the anti-C19 vaccination in the same variable. This is done in two ways, by measuring the latent variable of vaccination orientation as an ordinal or as a nominal/categorical variable (Babbie, 2020). In the first case, the values of the ordinal variable could be no intention to uptake vaccination against C19 (scored 1), sometime in the future (2), undecided (3), in a reasonable and specified time (4), soon (5), or already vaccinated (6). The nominal measure considers the same values of the variable by the six numerical codes but these do not indicate an ordering of the variable values.

The procedure of measuring the dependent variable allows for a modification of the research question. In the standard TPB, the questions are (1) on the factors of determination for the action intentions, and (2) on the impact of intention (and perceived control behaviour, in some cases) on the emergent action. In our integrated approach, the research questions are (1) on the specific causal profile that different predictors have on each of the values of the anti-C19 vaccination considered as a nominal variable, and (2) on the same dependent variable in its ordinal version of measurement. Different types of intentions of vaccination anti-C19 and the uptake of vaccine are simply stages in the process of vaccination. Such an approach is specially adjusted for situations when one does not have a clear time separation between structuring the intention and, later on, the act of adopting the planned behaviour as a result of previous intentions.

The new approach is also associated with considering specific societies or communities as involved in an innovation diffusion process (Rogers et al., 2005). Such societies are constituted as fields of adopters on non-adopters of vaccines as a health innovation. Those that are already vaccinated are early adopters. They are followed by several waves of later adopters, undecided, and non-adopters of the anti-C19 vaccination (Balas and Chapman 2018). The fields of adopters and non-adopters of the vaccination could be more or less aggregated functions of the available data and analytical interests.

The data we use in this analysis comes from the Flesh Eurobarometer 494, implemented in May 2021, in the countries of the EU. Four societies with very small subsamples (Estonia, Cyprus, Malta, and Luxembourg) were omitted from the analysis as we operate not only at the EU level but also country by country. Following the mentioned contextualisation premises, is not only the vaccination intention that counts but also the actual anti-C19 vaccination in the period before the survey moment. In other words, the dependent variable in the analysis includes past behaviour and future intentions in a single qualitative/nominal measured variable. In this way, it will be possible to determine specific influences of the different factors taken into account on some vaccination behaviours as realised or planned (no intention, some time, later in the year 2021, very soon, already vaccinated). Some errors could indeed affect the measurement because some people could declare that they did the anti-C19 vaccine but in fact, they did not. The great number of persons that answered the questionnaire could diminish the impact of such an error in the multivariate analysis that we adopted. Measurement errors in the practice of vaccination against C19 are also possible in the official statistics as an aggregation effect. The use of control variables and multilevel analysis in our case could contribute to diminishing such measurement errors.

The next section introduces some survey research to contextualise the data of the Eurobarometer we are using. The following section presents methods and data, results of the analysis and conclusions.

Comparative approaches in surveys on C19 vaccination

International comparative surveys of vaccination behaviours differ, methodologically speaking, mainly depending on how the dependent variable is measured and how predictors are chosen, with or without a predetermined model. The measurement of the dependent variable may be with or without the inclusion of past C19 vaccination behaviours. Only vaccination intentions, only past behaviours or past behaviours and intentions can be considered in the construction of the dependent variable. Also, in the measurement model, the nominal, ordinal or quantitative estimations of the behaviours or intentions of the anti-C19 vaccination can be taken into account. In the series of predetermined theoretical models used to choose predictors, the most common examples are the TPBs or the Health Beliefs Model. Another criterion in structuring the research methodology on the topic was the use of comparative versus non-comparative approaches.

Before the main variants of vaccination intention were identified, surveys investigated attitudes towards a possible vaccine. The first, non-comparative survey was conducted online in Australia, in March 2020, on a representative sample of 1420 Australian adults. The survey tested the attitude of the people towards getting vaccinated, by asking for agreement with the statement that ‘getting myself vaccinated for C19 would be a good way to protect me against infection’ (Seale et al., 2021), on a Likert-type 5-point scale. Significant predictors of this attitude were identified by logistic regression after recording the attitude measure into a dummy variable. Caeteris paribus, the highest probabilities for engaging in the anti-C19 vaccination were for females, 70 or more years old people, those having chronic diseases and private health insurance. The same study mentioned also the share of people intending to be vaccinated in the first wave of pandemic in other countries but without doing systematic comparisons by subsamples, referring to survey results in Denmark, France, Germany, Indonesia, Italy, Portugal, the Netherlands, and the United Kingdom.

A comparative survey (Trent et al., 2021) was also conducted online, between July and September 2020, in five major cities in Australia (Sydney, Melbourne), the United States (New York, Phoenix), and the United Kingdom (London). The prediction of the intention to vaccinate, as a dummy variable, was determined by variables related to trust in government/government information about C19, experiences of infection with C19 in personal communities of family or friends, and the perception of risks of infection with the New covid. In the series of socio-demographic predictors, variables related to age categories, gender, income, health status, smoking, and vaccination against influenza are included. Age is the only predictor with a significant impact on the will to vaccinate against the New COVID for each of the five cities in the sense that young people have a lower vaccination propensity than the elderly. Vaccination against influenza in the last 12 months significantly increases the likelihood of vaccination against C19 in four of the five cities. The exception of the lack of impact appears only for the London sample. It is not entirely clear whether the exception is given by reality or by the characteristics of sampling.

Another model, close to the previous one, is the one in which the intention to vaccinate anti-C19 is measured not dichotomously but by a continuous variable and predictors are chosen according to predetermined theoretical models. The analysis is based on a survey conducted in five Anglophone countries, namely Australia, the United States, Great Britain, New Zealand, and Canada (Burke et al., 2021). The dependent variable is a factor score that multiplicatively aggregates three indicators related to vaccination intent for protection from C19. The choice of predictors was based on the Health Behaviour Model and the TPB. With the survey conducted before having government C19 vaccination decisions, the questionnaire applied did not request information about behaviours to protect against C19 through vaccination.

The research argues that the probability of vaccination is higher for those who trust the information transmitted by the national government on the subject and have a high-level perception of the risks associated with non-vaccination of anti-C19 for themselves and others. The socio-demographic categories that favour a higher probability of vaccination are the elderly, the unemployed but seeking a job, and those who have already been vaccinated for other diseases in the past. Those who were unemployed but were not seeking a job had lower intentions of vaccination. It results that looking for a job stimulated the unemployed to traying also to get vaccinated against C19 as a precondition for employment. Altruistic and collectivist attitudes have also proved favourable to vaccination against C19. Variations in gender, income, religion, religiosity or the knowledge of people who have been ill with C19 do not appear to be significant predictors in the total sample.

The TPB is also adopted in the structuring of surveys in which the analysis is made not on vaccination in pandemic conditions but about two behaviours to prevent contamination through C19, namely the limitation of social contacts and repeated hand washing (Wollast et al., 2021). For each of these behaviours, questions about the frequency of behaviours, the degree of structuring of intentions, social norms, attitudes, and the perception of control in carrying out health actions are included in the survey questionnaire. The same questionnaire form is applied online for countries of close culture, that is, France and Belgium. In path analysis, used for data processing, the distinction between behaviours, intentions, social norms, attitudes, and perception of control over actions is maintained. Endogenous variables measured by multiple items are the result of the summative aggregation of some Likert-type scales.

Research by representative country surveys that did not use any theory to grounding it is in the case of Robertson et al. (2021) for the Republic of Ireland. The survey on 1600 adult persons, accomplished in January 2021, asked about the probability of deciding to get vaccinated against C19 – definitely yes, probably yes, probably no, definitely no. A logistic regression considered intentions to accept anti-C19 vaccination versus the hesitants as a dependent variable. The hesitants to anti-C19 vaccination were mainly younger women with lower education, having children, and those in the category of ethnic minorities. This was valid in the model running only with demographics as predictors. Another model adding a knowledge test on anti-C19 vaccination and perceived severity of the infection was more efficient to explain the variation of the dependent variable. In this new model, all the demographic predictors lost their statistical explanation (Burke et al., 2021). This is relevant to the fact that demographic variables influence vaccination/hesitancy intentions through the medium of knowledge on the vaccination and perceived severity of C19 infections. The finding could also be relevant to the fact that the knowledge test on anti-C19 vaccination is a very powerful methodological tool. The authors of the study mentioned as a key limitation the fact that the approach considered only intentions, not behaviours as dependent variables. Other possible relevant predictors such as rural-urban residency, past experiences of vaccination against other illnesses, and trust in relevant institutions were not recorded by the survey. Such predictors proved to be efficient in the approach of better-specified models of analysis as is the case with the comparative Anglophone survey presented above (Burke et al., 2021).

The analytical construction we present retains basic ideas from the TPB, but it is also dependent on the survey data (Eurobarometer 494 of 2021) with which we work and, inevitably, on different contexts from country to country. It is construction-oriented, simultaneously, by the extensive TPB (Sommer, 2011) but also by the data. One of the basic features of the extensive TPB, in the 2011 Sommer version, includes past behaviour as a predictor of future intentions and behaviour. In the theoretical model that we propose in the analysis, we will resume the idea but with a different approach, to which we will refer in the next section. Past behaviour is considered from two perspectives, generic and particular. Past vaccination behaviour, regardless of the field of disease prevention, appears as a determinant of behavioural intentions. But specific past behaviour, related to the anti-C19 vaccination, is here considered as a constituent value of the nominal variable relating to the intention to vaccinate anti-C19.

In the next section, we enter the survey data that we use and the analysis model. We then present the results of the analysis and, finally, the conclusions.

Data and theoretical model

The data we analyse comes from the Flash Eurobarometer 494, conducted in May 2021 in EU countries. We used the weighted version for 23 countries, with the elimination of very small subsamples, for Luxembourg, Malta, Estonia, and Cyprus (25,912 people of adulthood) to ensure greater stability of the calculation results at the country level.

The analysis model operates with the orientation of the anti-C19 vaccination as both an ordinal and nominal variable. In the ordinal version, we considered that the minimum pro-vaccination orientation is in those who declare that they will never get vaccinated to prevent this disease (score 1) and the score for the maximum orientation is 6, granted to those who have already been vaccinated. Intermediate values are expressed in terms of later (2), I do not know (3), sometime in 2021 (4), and soon (5). In this way, between the intention to vaccinate never and I have already been vaccinated, there is a polarity. In the nominal type measurement, the six values are only different, not ordered between them. The option was supported by the fact that past behaviours of anti-C19 vaccination were recorded at the same moment with the intent of being vaccinated. Past behaviour and intentions are considered as values of the same variable measuring the accomplished or announced practice of vaccination. Both of them had specific motivations that were not recorded in a non-panel survey. This is a procedure to integrate vaccination into a time context by reference to the past, near, uncertain, and long-time future. The space integration of the approach is done here by considering institutional and physical space. The contextualisation of the vaccination against C19 is accomplished by a time-space integration of the practice.

The first hypothesis (H1) that structures the theoretical model we are working with argues, in line with the planned behaviour theory, that the pro-vaccination behaviour is all the more intense as the anti-C19 pro-vaccination attitude is stronger (Figure 1).

Figure 1.

Theoretical model of predicting vaccination behaviours against C19.

The pro-vaccination orientation index (IPVO) is constructed as a factor score (multiplied by 100), from three indicators related to the agreement of the interviewees with formulations that claim that the anti-C19 vaccination is a civic duty, has more advantages than disadvantages and that vaccination, in general, has contributed to the disappearance of many diseases (Sandu, 2021). To measure the perception of the significant other in terms of anti-C19 vaccination we used as proxies, two confidence indices. The first of these is constructed as a factor score from indicators relating to trust in the government, local authorities and health authorities as sources of information about the anti-C19 vaccination. The second is about trust in online networks used as a source of information about the anti-C19 vaccination. The second hypothesis (H2) argues that increased trust in public administration institutions favours anti-C19 vaccination and the third hypothesis (H3) supports the expectation that increased trust in information on social networks will favour behaviours of refusal or hesitation in the anti-C19 vaccination (Wang and Liu, 2021).

The fourth hypothesis (H4) states an increased probability of vaccination/intention of anti-C19 vaccination for people who have been vaccinated at least once before as an adult to prevent another disease (Sandu, 2021). Otherwise formulated, the hypothesis claims that the increased likelihood of anti-C19 vaccination is higher for those who already have the personal habit of being vaccinated as adults. The hypothesis helps to a better specification of the explanatory model. TPB includes this factor as a proxy for perceived behavioural control (Ajzen, 1991). Some of its applications include the mentioned experience in the model (Burke et al., 2021; Conner and Armitage 1998). Some other prediction models omit the factor (Robertson et al., 2021). When one has to compare different countries of the EU, as in our case, the experience is a very useful predictor as it can allow for separating the country from personal experiences about past vaccination experiences. The questionnaire for the flash Eurobarometer 494 that we are using here allows for a differentiation between being or not vaccinated against other diseases as a child or as an adult. We used in this analysis only the information referring to the practice of vaccination as an adult.

Finally, in line with other previous research at the European level (Sandu, 2021) focused on the attitude towards the anti-C19 vaccination, it is expected that pro-vaccination intentions and behaviours of vaccination about the same infection by C19 will have a higher probability in more developed territorial units (H5): People from societies of the Old EU tend to accept vaccination anti-C19 more than people from the societies of the New Member States of the EU (H5a); living in urban areas of European society would bring higher rates of vaccination anti-C19 than in rural areas (H5b).

We present below the results of the analysis from a descriptive and explanatory point of view. In the strictly descriptive part, we present the way of grouping countries in terms of past behaviour patterns – intentions of vaccination against C19. The similarities that occur between populations in different countries allow the identification of criteria and favour similar attitudes and behaviours in the field of analysis. After identifying the frequency distributions for the six values of the main analysis variable related to the behaviours in the anti-C19 vaccination, we move on to the grouping of countries in terms of similarities between them, depending on the respective dominant behavioural patterns. Subsequently, the analysis returns to the individual level to distinguish models of prediction of vaccination behaviours/intentions.

National models for reporting to C19 vaccination

The 23 countries that remain under analysis after dropping those for which the samples are very small, in the survey, are far from unique in terms of anti-C19 vaccination behaviours. The first differentiation that appears very clearly is that between the Old and the New EU Member States (Figure 2). Except for Greece, the population of all the other countries of the Old EU is mostly oriented, following H5a, in favour of anti-C19 vaccination to a greater extent than that of the New Member States of the Union. Even if one computes the index of dominant orientation (see details of computation in the footnote of Figure 2) by country levels, not by rural and urban areas, people from Greece have a lower pro-vaccination orientation against C19, closer to that of the people from the New Member States (figures not shown here). For both groups of New and Old Member States, it was found that the pro-vaccination orientation was higher in the big cities compared with that in rural communities. The only exception is the people from Bulgaria, with a lower dominant opinion in favour of anti-C19 vaccination in urban areas compared with rural areas. The trend is in line with H5b and the exception should be addressed by country studies. In major EU cities, for example, the pro-vaccination dominant orientation index was 71% in the May 2021 survey data. For rural communities in the Union, the dominant pro-vaccination orientation was 66%. For all Central and Eastern European countries, these indices were lower than the EU average. Overall, the data in Figure 1 confirm H5 in its variants H5a and H5b. Only in the case of Belgium and Slovakia are situations different from the European trend in the sense that at the level of these countries, the index of the anti-C19 pro-vaccination orientation is higher in rural than in urban areas.

Figure 2.

Dominant orientation on anti-C19 vaccination by rural and urban areas in EU countries (%).

Data source: Flash Eurobarometer 494. The figures in the diagram represent the index of dominant orientation on C19 vaccination = (% positive orientation−% negative orientation) × (100−DK)/100. The positive orientation sums up those that are vaccinated or intend to do it soon or sometime in 2021. The negative orientation is given by those that said that they will never vaccinate or will do it an indefinite later. DK – non-answers. Urban here means living in large cities and rural is the category including villagers and dwellers of towns. For example, the index of dominant orientation on C19 vaccination in urban areas of Spain is 92%, 10 percentage points larger than in rural areas. Countries with very small samples (Malta, Cyprus, Estonia, and Luxembourg) were not included in the analysis. The index of dominant orientation on C19 vaccination for rural Bulgaria is −3.

The second level of differentiation can be identified at the level of country clusters (Figure 3). The graph allows the identification of three similarity nuclei, in the sense of groupings of countries with the utmost similarity between the anti-C19 vaccination behavioural profiles, assessed in terms of past intentions and behaviours: Bulgaria–Slovakia, Denmark–Portugal, and Ireland–Sweden–Netherlands–Spain. The largest grouping, in terms of the number of countries, is Ireland–Sweden–Netherlands–Spain. Italy also revolves around this nucleus of similarity. Together, the five countries recorded, in mid-2021, form the European grouping with the strongest intentions of rapid vaccination against C19 (Table 2). Three of the five countries are part of the developed North-Global (Ireland–Sweden–Netherlands), and two, Spain and Italy are in the southern part of the continent. We also keep in mind, for the time being, the information on the possible relevance of the territorial proximity between some of the countries that make up the grouping.

Figure 3.

Dendrogram of similarities among European countries by intentions and practices of vaccination against C19.

Data source: Eurobarometer 494. Hierarchical cluster analysis, furthest neighbour, Pearson correlations as measures of similarity. Profile of each country determined by the share of the country respondents to the question on practices and intention of vaccination against C19: vaccinated already, as soon as possible, later on in 2021, undecided, later, never. Countries with very small samples (Malta, Cyprus, Estonia, and Luxembourg) were not included in the analysis. The six clustering variables are standardised with z scores. The vertical line in the diagram is a marker of the most homogeneous clusters of similarity.

Table 2.

Predictors of intentions to get vaccinated at the European Union level by categories of practices, May 2021.

Predictors	‘When would you like to get vaccinated against COVID-19?’ (reference category ‘do not know’)
	Never		Later		Sometime in 2021		Soon		Already vaccinated
Index of pro-vaccination orientation (IPVO)	−0.009	***	−0.001		0.006	***	0.015	***	0.016	***
Vaccinated as adult*	0.053		−0.023		0.174		0.411	***	0.744	***
Index of trust in institutions	−0.271	**	0.262	**	0.476	***	0.478	***	0.442	***
Index of trust in online networks and web	0.065		0.054		−0.004		0.024		−0.143	**
Large city*	−0.056		−0.099		0.083		0.250	*	0.112
Age 15–29 years old (yo)*	0.124		0.378	*	0.451	**	0.264	+	−0.485	**
Age 60+ yo* (reference age 30–59 yo)	0.055		0.473	**	0.016		0.098		1.609	***
Man*	0.250	*	0.342	**	0.332	**	0.458	***	0.274	**
Tertiary education*	0.049		0.106		0.179		0.094		0.024
Still studying*	−0.117		−0.163		−0.060		0.053		−0.242
Employee*	0.056		0.402	**	0.448	***	0.532	***	0.684	***
Having with children under 15 years old*	0.022		0.118		0.254	*	0.130		−0.432	***
Constant	−1.167	***	−0.270		0.585	**	0.906	***	0.486	*
Pseudo R2	0.236
N	23,560

Data source: Eurobarometer 494, May 2021. Multinomial logistic regression. The full model that is not presented here included also 23 residence countries (with Finland as a reference category). The coefficients in the table are computed by controlling for the residence country. Pseudo R² for the model without the country predictors is 0.210. Regression in STATA, using the pweight option.

^*p < 0.05; **p < 0.01; ***p < 0.001.

The data in Figure 3 and Table 1 make it possible to highlight the fact that most of the countries in the Old EU belong to pro-vaccination groups, in intent and behaviour. To the mentioned grouping of the five countries are added the groupings of similarity Belgium–Germany and Denmark–Finland–Portugal. The only countries under analysis that are part of the Old EU Member States but do not show a strong similarity to that model are Austria, France, and Greece. Austria and France are closer in terms of the behavioural-attitudinal model of anti-C19 vaccination to the countries of the East and Centre of the EU, Hungary, in particular. Greece, for its part, has a behavioural profile closer to Bulgaria–Croatia–Slovakia than to that of the countries of the Old EU. For now, we note, at a descriptive level, again, a confirmation of H5 that formulated the expectation that the intentions and pro-vaccination behaviours-C19 will be much stronger in the Old, compared with the New EU.

Table 1.

Profiles of European Union countries by patterns of orientation to the vaccination against C19.

Types	Sub-types	Clusters of similar countries by vaccination behaviours	‘When would you like to get vaccinated against COVID-19?’
			Never	Later	Do not know	Sometime in 2021	Soon	Already vaccinated
Controversy societies	Polarised between vaccinated and antivaxxers	FR AT HU	6.0	3.1	3.7	−2.2	−19.1	12.8
	Large share of antivaxxers	LV SI	5.4	3.8	3.0	.3	−5.8	−1.2
	Between never and undecided	LT PL	9.6	.5	10.0	.1	−9.4	−1.9
	A high controversy society	RO	3.4	3.7	3.8	3.3	−14.2	5.9
	Controversy with high share of antivaxxers	SK HR BG	12.3	10.3	3.7	5.8	−11.5	−6.8
Antivaxxers orientation	Prevalent antivaxxers orientation	GR	2.5	4.6	−.9	3.0	−2.9	−2.4
	High antivaxxer orientation	CZ	5.7	3.9	−.6	−.5	−1.4	−3.3
Provaccination	Provaccination orientation	DK FI PT	−5.4	−.9	−2.3	7.5	4.2	−3.9
	High intentions to provaccination	IE NL SE ES IT	−14.6	−6.7	−6.4	1.3	26.1	−10.9
	Unconditional provaccination	BE DE	−4.7	−5.8	−5.8	−8.7	8.2	6.1

Data source: Eurobarometer 494. The clusters of similar countries by vaccination orientation are derived from the dendrogram in Figure 1. Figures are adjusted standardised residuals in a table crossing cluster of countries with the values of the variable referring to vaccination intentions and past behaviours. A positive sign indicates a positive association between column and row value and the negative one is significant for a disassociation. For a probability of 95%, the threshold for significance is 1.96. For example, the highest positive association to getting vaccinated soon is for the subsample of the survey from Ireland–The Netherlands–Sweden–Italy–Spain as indicated by the value of 26.1.

Shaded figures indicate significat positive associations.

The most specific profiles of behaviour-intention are for Romania, the Czech Republic, and Greece. Romania is in the same cluster as Latvia, Lithuania, Poland, and Slovenia but as a marginal member, with a rather lower similar profile to the rest of the grouping. The Czech Republic is also in the same cluster as other Eastern European countries (Slovakia, Croatia, and Bulgaria) but, also, as a marginal member, with a lower coefficient of similarity. The same marginal position in the same cluster is specific to Greece which is part of the Old Member States of the EU.

Of course, the 10 groupings of countries generated by cluster analysis are with diffuse limits. At this stage of the analysis, it is not clear whether they differ from each other because of their socio-demographic compositions or certain cultural or economic characteristics. We will have an answer to the question when we introduce the survey respondents belonging to those groupings as independent variables in multivariate analysis and multilevel models.

Previous findings have operated with country-wide survey data. If the analysis goes down to the individual level, can we talk about country models or a European model of reporting on anti-C19 vaccination? What is the socio-demographic profile of those who said they would get vaccinated soon or never or opted for one of the intermediate variants? What about those who have already been vaccinated? These questions we will answer in the section that follows.

Socio-cultural orientations in the anti-C19 vaccination

The pro-vaccination attitude leads, as expected according to the first hypothesis, to the adoption of vaccination intentions soon or in the foreseeable future of the survey year (Table 2). A low level of this attitude favours, at the European level, the adoption of the refusal to vaccinate. Only in the grouping of those who vaguely intend to adopt that vaccine the level of that attitude does not matter significantly.

Why some people are more pro-vaccination oriented, against C19, than others, recording higher IPVO values? It is a question of status, experience, and place. We will summarise shortly the answer to this question, based on the results of multiple regression (that is not presented here) having IPVO as a dependent variable. High pro-vaccination orientation, at the EU level, is specific for older men from cities, being of higher or middle-level education, with high trust in relevant institutions for anti-C19 vaccination, and that lived the experience of being vaccinated before to avoid other illnesses. All these factors have specific, net effects to favour higher values for IPVO. Irrespective of these factors, there are country effects. Living in Spain, Italy, Sweden, and Ireland had the highest positive impact on pro-vaccination against C19. At the other extreme are the countries with the highest negative effects on the same attitude. Here are included mainly central-European countries like Slovenia, Slovakia, and Austria. Living in Latvia had the highest negative effect on the same attitude.

Trust in the government, health administration, and local government appears to be one of the strongest factors in favouring the anti-C19 vaccination, in line with the expectations formulated by H2. Symmetrically, distrust in these institutions leads to the refusal of vaccination. Surprisingly, trust in online social networks does act by H3, only for those who have already been vaccinated. Specifically, those who have already been vaccinated tend to have a low level of trust in online networks. For the remaining categories analysed, that factor no longer matters significantly. Such a finding may also be related to the fact that the analysis model was not sufficiently well specified. We do not know, for example, from the basic data of the survey used, which of those interviewed were high-frequency Internet users, associated with social networks.

In H4, the expectation is formulated that people who have been vaccinated in the past as adults, for another disease, will be more oriented in favour of anti-C19 vaccination. We do not find that relationship as significant except for those who have already been vaccinated or are going to get vaccinated shortly. That factor appears to be, by presence, a favourable condition for rapid vaccination. His absence, however, does not automatically lead to the blocking of the intention to vaccinate anti-C19.

A poorly structured culture of vaccination as a preventive measure is especially specific for Romania, Lithuania, Poland, and Hungary, in the New Member States of the EU, and Italy and Greece from the Old EU. It is in these countries that there is a lower probability of being before vaccinated as an adult for other illnesses, controlling for socio-demographic factors (gender, age, education urban, or rural residence). On the positive extreme, with a high effect on being vaccinated as an adult, irrespective of disease is Portugal. (Results of regression analysis that are not presented here.)

The H5a assumption holds that residency in the Old EU Member States favoured a higher likelihood of adopting a positive attitude towards the vaccination against C19. The descriptive data, to which we have already referred, confirmed this expectation.

The gap between Old and New Member States of the EU, in terms of pro-vaccination orientation, has multiple sources. Trust in institutions, for example, that are relevant for the vaccination against C19 is systematically higher in the Old Member States (Table 4), except in France. The quality of the institutions that provide the context for anti-C19 vaccination comes, very likely, from the long history of the survival of the low-quality of institutions for the former communist states of the EU (Mishler and Rose, 2001). Social development as measured by life expectancy at birth is also, systematically higher in the Old Member States, with a variation between 81 and 83 years old, in 2021, versus the New Member States with a life expectancy between 71 and 77 years old, according to EUROSTAT sources. The only exception from this point of view is Slovenia, a New Member State, which has a life expectancy that is equal to the value of the index for Germany (81 years old), in 2021. Low life expectancy goes together with lower quality of the health system and higher morbidity rates. Economically, all the New Member States had a gross domestic product (GDP) per capita below the EU average (EUROSTAT data for 2019, not mentioned here). By contrast, 10 out of 14 Old Member States had a GDP per capita higher than the EU average. These are only examples of development gaps between New and Old Member States that could be relevant to population attitudes and behaviours related to vaccination against C19.

H5a is also supported by data at the individual level, in multilevel analysis. In Table 3 we present, in simplified form, a picture of the countries for which we have recorded, with data at an individual level, a significant impact of the country of residence on the personal orientation in terms of anti-C19 vaccination. All the eight countries for which there is a positive and significant effect on the intention to vaccinate soon are in the category of the Old EU Member States. The model of adopting the vaccine in the past, until mid-2021, was specific for Germany–Ireland–Belgium, Austria–France and Hungary–Romania. The cultural model of refusing C-19 vaccination is specific to three Eastern European countries (the Czech Republic, Croatia, and Bulgaria) and Greece.

Table 3.

Specific patterns of vaccination behaviours and intentions.

Countries as significant predictors	‘When would you like to get vaccinated against COVID-19 ? ‘ (reference category ‘do not know’)
	Never	Later	Sometime in 2021	Soon	Already vaccinated
Germany				+	+
Ireland				+	+
Belgium				+	+
The Netherlands				+
Portugal				+
Italy				+
Spain				+
Denmark				+
Czech Republic	+	+
Greece	+	+
Croatia	+
Bulgaria	+
Hungary					+
Austria					+
France					+
Romania					+

Data source: Eurobarometer 494, May 2021. Simplified table with significant positive regression coefficients (+) using the full model of multinomial regression from Table 2. Example: Living in the Netherlands favours significantly the option of declaring that the person will vaccinate soon, caeteris paribus.

Challenges of socio-demographic and institutional contexts

Beyond the effects of attitude, trust in institutions relevant to the anti-C19 vaccination and personal health experiences, socio-demographic factors also matter significantly when regression analyses are done country by country.

The continental region and country context have an important role to play in differentiating causal or conditioning relationships (Tables 5 and 6 in the Annex containing patterns of multiple regression with the ordinal-dependent variable measuring orientation towards the C19 vaccination). In each of the countries included in the analysis, being over 60 years of age significantly increases the likelihood of C19 vaccination, but more in the countries of the Old EU than in the New Member States. Why is this so? A possible explanation would be the higher average age in the West and the North than in the East. Further analysis is needed. The only exception is Latvia, where being a person over the age of 60 years does not have a significant impact on behavioural orientation towards vaccination.

The residential environment itself, controlling the other predictors, only matters to Finland, Latvia, and Romania. In these countries, people in urban areas, in large cities, in particular, tend to support anti-C19 vaccination more than the population in rural communities. It is likely that in these countries social interaction favourable to C19 infection is much stronger in urban than in rural areas compared with the situation in the rest of the EU. It’s a hypothesis generated from data analysis. To be tested, though.

The presence of children under the age of 15 in families tends to discourage adult vaccination in 10 of the 23 analysed societies. This effect has a maximum intensity in Poland, the Czech Republic, and Ireland and Finland. We do not know why that is the case. Sampling mode effect or country conditionings?

In the total European sample, the status of employees considerably increases the probability of vaccination in fact or as an intention (Table 2). When the analysis is made country by country and with a dependent variable of the ordinal type, the results are different (Table 5 and Table 6 in annexe). Among the countries of the Old EU, only in Italy do we see a significant favouring of the anti-C19 vaccination for those who work. In Eastern Europe, that relationship is more common. It occurs, at a significant level, in descending order of intensity, in Romania, Hungary, Lithuania, and Slovakia. Why only here? Was it only here a higher institutional pressure for the C19 vaccination at the level of employees? Likely, it is not only a supplementary pressure for employees to vaccinate coming from their institutional environment but, also, a higher density of interactions at work, stimulating the employees to be more cautious for their and their family’s health care. Vaccination for employees could be, also, a way to preserve their job in pandemic conditions when finding a job is not so easy.

Conclusion and discussion

The integrated approach of the vaccination against C19 was promoted, in this analysis, by including past personal vaccination, along with the values of the intention of anti-C19 vaccination, in the same variable, by promoting multilevel analysis, and by expanding the list of predictors. The differentiating lines are not only those between vaccinated–unvaccinated and I will get vaccinated soon versus never. In addition, it was also useful to measure the vaccination orientation both by one nominal and one ordinal variable, allowing for triangulation and sensitivity analysis (Treiman, 2014). Ordinal measurement considered the stages between being vaccinated (score 6) and clear refusal of vaccination against C19 (score 1) as extreme states having between them the ordered states in the series I will vaccinate soon (5), sometime in 2021 (4), do not know (3), later (2). Nominal measurement considered also the same six possible categories of the variable vaccination but the coding numbers here indicate no order. The same technique of regression analysis was used on nominal and ordinal measures of the same measures of vaccine orientation and gave consistent results.

The research question was no more, as in the classical TPB, ‘what are the determinants of the intention to adopt certain behaviours?’. Intentions to adopt vaccination against C19 were considered as stages leading to vaccination. The target of the research was to identify significant predictors for each value of the vaccination orientation, using a large representative sample for 23 out of the 27 countries of the EU. The new integrated approach allowed for an understanding of anti-C19 vaccination as an innovation diffusion process (Rogers et al., 2005), distinguishing between early adopters, late adopters, and opponents of vaccination against C19.

At the individual level, all five hypotheses are supported by data. Pro-vaccination orientations against C19 tend to be stronger for those with vaccination-friendly attitudes (IPVO), with confidence in the relevant institutions for vaccination against C19, who have the experience of being vaccinated as an adult, against other diseases and reside especially in major cities in the EU Member States. The culture of vaccination in general (as measured by IPVO and prior experiences of vaccination as an adult) and the attitude of the interviewee on relevant institutions towards anti-C19 vaccination are by far more important than education on vaccination behaviour. Education per se, controlling for other factors, has no significant effect on vaccination orientation (see Table 2). This could be an indication that education has mostly an indirect effect on anti-C19 vaccination, through the medium of attitudes, perceptions and trust. Higher education, for example, favours more positive attitudes on vaccination against C19 (higher values of IPVO), and, implicitly or indirectly, more structured behaviours against C-119 vaccination.

The fact that the experience of vaccination, for whatever illnesses, favours significantly the belonging to the category of the already vaccinated and to those that are decided to be soon vaccinated is not only a confirmation of the fourth hypothesis. It has, also, a methodological relevance to the fact that the practice of past vaccination is part of a continuum as the same type of causal conditioning appears only for those that are decided to be vaccinated soon. No other categories on the measure of vaccination intention connect significantly to the predictor measuring the culture of vaccination.

At a more aggregated level, national contexts, in turn, bring significant differentiation in pro-vaccination guidelines. Germany, Ireland, and Belgium, for example, were, in mid-2021, the countries with the strongest orientation in favour of anti-C19 vaccination both in terms of the share of people already vaccinated and in terms of rapid vaccination intentions. The strongest nucleus of antivaccine against C19 was, also for mid-2021, in Greece, Bulgaria, Croatia, and the Czech Republic.

County effects on behaviour-intentions of vaccination against C19 were significantly conditioned by the culture of vaccination as an adult against other illnesses, previous pandemic period. Easter European countries like Romania, Poland, Hungary, and Lithuania, recorded a significantly poorer culture of vaccination to prevent other diseases than C19. Portugal, from the grouping of the Old EU states, is an opposite example of a highly structured culture of vaccination against other illnesses, favouring pro-vaccination against C19.

At the space level, we found that the pro-vaccination orientation against C19 is stronger in the Old than in the New Member States of the Union, and the urban area compared with the rural area, with few exceptions discussed in the text. We do not know exactly what mostly counts for the higher anti-C19 vaccination in the Old compared with the New EU. There should be some other factors that were not included in the analysis models. Maybe the quality of the health system or the social contexts with a higher density of population favour more social interactions in the Old compared with the New Member States.

Apart from the distinction between the Old and the New Member States, in terms of the index of the dominant orientation of the population concerning the anti-C19 vaccination, the differences in the similarity between groups of countries also operate a lot. The 10 groupings of countries also settle into three broad categories with dominant pro-vaccination, and anti-vaccination orientations and strong controversies on the issue within the countries. The largest group of countries with predominantly pro-vaccination-oriented populations consists of Ireland–Netherlands–Sweden–Spain–Italy. Anti-vaccination guidelines were recorded, with maximum intensities, in mid-2021, in Greece and the Czech Republic. The groups of maximum controversy between the pros and cons of vaccination were in Latvia–Slovenia, Lithuania–Poland, and Romania. The finding is significant for the fact that at the national level there are not only pros and cons in societies but there are also controversial places. Several factors that are unmeasured here could count. The institutional spaces of vaccination with its early or later, good or less satisfactory communication and knowledge processes seem to be essential ones. The ability of a society to learn from population reactions to the vaccination measures and to correct public policies, as a function of these reactions were, very likely, important aspects.

There are several uncertainties or weak points in the analysis of these data. The attitude towards anti-C19 vaccination is a logical predictor of the intention to vaccinate against C19, in line with the TPB. What about including, as we did in the article, the attitude on the vaccination against C19 as a predictor of past vaccination against C19 in the multinomial regression? It is possible, at least in theory, to have a successful vaccination not only as a consequence of the previous attitude but, also, the current attitude regarding anti-C19 vaccination as determined by the former practice of ani-C19 vaccination. Even, if possible, such a reverse or circular effect, is rather unlikely because the anti-C19 vaccine was rather new. So, a certain circular loop could function between the two variables but the probability is higher of having a structured attitude favouring anti-C19 vaccination even before the past vaccination against C19.

The Eurobarometer survey we are using here was done in May 2021. Unfortunately, we could not have a dynamic image including changes that took place, at the individual level, in terms of anti-C19 vaccination. There is no other survey similar to the flash Eurobarometer 494, including the same key variables for a replication of the analysis.

It is, also, a weak point of this analysis the fact that survey data do not allow us to know what kind of vaccination was adopted by the interviewee. It is very likely that not only the culture of vaccination counts but, also, the type of vaccination.

Author biography

Dumitru Sandu is emeritus professor of sociology at the University of Bucharest. His main publications are on transnational migration, transition sociology, community and regional development, and measures of social capital.

Appendix

Table 4.

Descriptives of key predictors for vaccination against C19.

Typology of vaccination orientations in the society		Country	Vaccinated as adulta	Residence in large citya	Employeea	% 15–29 years old	% 60+ years old	Index of provaccine orientation		Index of trust in institutions relevant for vaccination
			Mean	Mean	Mean	Row N %	Row N %	Mean	SD	Mean	SD
Controversy societies	Polarised between vaccinated and antivaxxers	France	.78	.22	.39	18.8	33.4	−14.32	101.37	−.11	.94
		Hungary	.58	.38	.52	15.0	32.8	−36.36	109.81	−.37	.78
		Austria	.81	.33	.49	26.8	23.8	−28.96	109.47	.04	1.04
	Large shares of antivaxxers	Latvia	.64	.48	.49	25.8	15.0	−69.65	102.77	−.36	.77
		Slovenia	.67	.19	.52	20.0	21.6	−56.95	106.65	−.39	.69
	Between never and undecided	Lithuania	.45	.50	.49	27.7	18.3	−14.99	105.47	−.20	.87
		Poland	.55	.37	.51	23.8	22.3	−21.13	101.84	−.37	.79
	A high controversy society	Romania	.42	.45	.57	21.4	16.4	−13.41	106.08	−.07	1.00
	Controversy with high share of antivaxxers	Slovakia	.76	.23	.49	20.3	24.3	−46.75	111.33	−.21	.86
		Bulgaria	.68	.60	.39	21.1	21.0	−47.35	108.39	−.26	.87
		Croatia	.66	.37	.47	22.1	26.3	−39.26	98.85	−.40	.70
Antivaxxers orientation	Prevalent antivaxxers orientation	Greece	.60	.52	.41	25.5	12.9	3.62	95.85	−.08	.86
	High antivaxxers orientation	Czech Rep.	.87	.29	.52	18.5	26.8	−20.11	97.41	−.19	.76
Provaccination orientation	Provaccination orientation	Denmark	.71	.35	.48	24.6	27.8	7.02	91.81	.16	1.03
		Portugal	.90	.36	.46	21.2	24.7	3.80	84.14	.37	1.07
		Finland	.84	.30	.30	20.6	31.5	10.50	95.55	.44	1.13
	High share of soon provaccination	Ireland	.64	.39	.58	26.1	13.2	13.90	93.21	.21	1.03
		Sweden	.74	.33	.49	17.3	32.7	40.64	89.59	.56	1.22
		The Netherlands	.67	.22	.56	17.6	33.8	−7.17	100.10	.31	1.09
		Spain	.69	.37	.52	17.3	28.1	35.01	82.44	−.02	.91
		Italy	.56	.30	.33	16.3	29.0	23.58	90.82	.07	.96
	Unconditional provaccination	Belgium	.80	.23	.40	18.2	28.2	−4.82	101.76	.02	1.01
		Germany	.87	.29	.45	12.9	34.0	−.15	100.07	.09	1.08

Data source: Eurobarometer 494, May 2021.

All the descriptive statistics and multivariate analyses are computed by the weighting variable specified in the Flash Eurobarometer EB494 by the data provider. Malta, Cyprus, Luxembourg and Estonia having very small samples of less than 100 persons, were omitted from all the computations.

^aDummy variable.

Maximum values on columns are marked by shadow.

Minimal values on columns are in bold.

Shadow mark significant coefficients for p < 0.05.

Table 5.

Predicting vaccination against C19 in Old Member States of the European Union.

Predictors	Denmark		Finland		Portugal		Ireland		Netherlands		Sweden		Spain		Italy		Germany		Belgium		France		Austria		Greece
	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z
IPVO	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00
Index of trust in institutions	0.17	0.01	0.10	0.11	0.16	0.02	0.29	0.00	0.11	0.05	0.08	0.17	0.18	0.02	0.06	0.36	0.03	0.62	0.12	0.07	0.12	0.08	0.19	0.00	0.33	0.00
Index of trust in online networks and web	−0.13	0.05	−0.13	0.01	−0.13	0.16	−0.16	0.01	−0.07	0.29	−0.03	0.59	−0.16	0.06	−0.05	0.36	−0.07	0.24	−0.10	0.11	−0.16	0.05	−0.24	0.00	−0.20	0.00
Vaccinated as adult*	0.30	0.04	0.75	0.00	0.49	0.04	0.24	0.11	0.32	0.02	0.31	0.05	0.31	0.05	0.32	0.01	0.64	0.00	0.30	0.06	0.25	0.12	0.36	0.04	−0.02	0.88
Age 15–29 years old*	−0.21	0.20	−0.33	0.05	−0.41	0.01	−0.29	0.07	−0.43	0.02	−0.43	0.02	0.48	0.02	−0.11	0.59	0.16	0.46	−0.59	0.00	−0.67	0.00	−0.12	0.41	−0.69	0.00
Age 60+ yo*	2.32	0.00	0.85	0.00	2.29	0.00	1.52	0.00	1.90	0.00	1.53	0.00	2.26	0.00	1.31	0.00	1.45	0.00	1.04	0.00	1.25	0.00	1.52	0.00	0.81	0.00
Man*	−0.33	0.01	0.16	0.19	0.04	0.80	−0.22	0.13	−0.06	0.65	−0.24	0.07	0.03	0.86	0.00	0.99	−0.09	0.50	−0.14	0.26	−0.06	0.64	0.02	0.87	0.26	0.09
Tertiary education*	0.07	0.66	−0.07	0.66	−0.02	0.89	0.00	1.00	−0.12	0.43	−0.36	0.02	−0.07	0.65	0.02	0.91	0.04	0.79	−0.03	0.86	−0.23	0.19	−0.13	0.38	0.19	0.33
Still studying*	−0.26	0.21	0.07	0.73	0.09	0.67	−0.35	0.12	0.05	0.85	−0.75	0.00	−0.88	0.00	−0.13	0.61	−0.12	0.62	−0.06	0.80	0.00	1.00	0.02	0.92	−0.01	0.96
Employee*	−0.05	0.74	0.27	0.06	0.22	0.13	−0.05	0.73	−0.03	0.87	0.05	0.78	−0.03	0.84	0.38	0.01	0.30	0.04	−0.15	0.33	0.17	0.30	0.21	0.11	0.25	0.11
Having children under 15 years old*	−0.15	0.29	−0.42	0.00	−0.13	0.34	−0.54	0.00	−0.14	0.33	−0.30	0.03	−0.15	0.34	−0.26	0.05	−0.35	0.02	−0.32	0.03	−0.32	0.03	−0.25	0.06	0.01	0.94
Urban residence*	0.11	0.39	0.32	0.02	−0.01	0.97	−0.09	0.56	−0.01	0.97	0.02	0.91	0.17	0.24	0.03	0.80	0.00	0.98	−0.14	0.34	0.13	0.40	0.12	0.36	0.04	0.81
Pseudo R2	0.22		0.17		0.17		0.20		0.19		0.21		0.11		0.15		0.16		0.15		0.19		0.19		0.20
N	1004		1004		1014		1052		1008		1005		1008		1061		1052		1001		1001		1067		1040

Data source: Eurobarometer 494 May 2021. Ordinal logistic regression for each country. Countries of similar profiles by vaccine orientation (Figure 1) or in proximity are placed in proximity columns.

IPVO: Index of pro-vaccination orientation.

^*Dummy variable.

Shadow mark significant coefficients for p < 0.05.

Table 6.

Predicting vaccination against C19 in New Member States of the European Union.

Predictors	Latvia		Slovenia		Lithuania		Poland		Romania		Slovakia		Bulgaria		Croatia		Czech Rep		Hungary
	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z	Coef.	P > z
IPVO	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.01	0.00	0.02	0.00	0.01	0.00	0.01	0.00	0.01	0.00
Index of trust in institutions	0.40	0.00	0.25	0.03	0.07	0.46	0.17	0.05	0.17	0.02	0.15	0.05	0.18	0.04	0.27	0.00	0.33	0.00	0.26	0.08
Index of trust in online networks and web	−0.25	0.00	−0.10	0.27	−0.12	0.08	−0.02	0.77	−0.20	0.00	−0.19	0.00	−0.03	0.58	−0.13	0.03	−0.27	0.00	−0.26	0.00
Vaccinated as adult*	0.31	0.07	0.20	0.18	0.16	0.30	0.50	0.00	−0.05	0.74	−0.04	0.76	0.38	0.01	0.30	0.03	0.40	0.02	0.50	0.00
Age 15–29 years old *	0.03	0.89	−0.30	0.12	−0.28	0.09	−0.57	0.00	−0.19	0.33	−0.74	0.00	−0.16	0.38	−0.46	0.01	−0.61	0.00	−0.22	0.43
Age 60+ yo*	0.41	0.06	1.21	0.00	1.10	0.00	0.43	0.03	0.60	0.01	0.68	0.00	0.52	0.00	0.61	0.00	0.83	0.00	0.94	0.00
Man*	−0.08	0.60	0.05	0.74	−0.06	0.70	−0.10	0.44	−0.01	0.97	−0.18	0.16	0.04	0.76	0.00	0.97	−0.02	0.86	0.05	0.78
Tertiary education*	0.17	0.34	0.04	0.79	0.24	0.17	0.03	0.85	0.10	0.57	−0.18	0.21	0.12	0.42	0.20	0.19	−0.16	0.27	0.52	0.00
Still studying*	0.46	0.07	0.10	0.71	0.30	0.28	0.14	0.58	0.30	0.26	0.26	0.28	0.10	0.76	0.16	0.58	0.11	0.52	0.67	0.10
Employee*	0.28	0.07	0.24	0.10	0.53	0.00	0.28	0.06	0.61	0.00	0.29	0.05	0.16	0.24	0.15	0.29	−0.03	0.84	0.60	0.00
Having children under 15 years old*	−0.15	0.33	−0.23	0.14	0.07	0.65	−0.48	0.00	0.07	0.62	−0.14	0.30	−0.33	0.02	−0.15	0.27	−0.45	0.00	0.06	0.77
Urban residence*	0.36	0.02	−0.17	0.34	−0.01	0.95	−0.27	0.06	0.32	0.04	0.19	0.22	0.13	0.39	0.04	0.75	−0.03	0.84	0.10	0.54
Pseudo R2	0.18		0.20		0.19		0.16		0.19		0.19		0.19		0.17		0.17		0.23
N	1019		1012		1113		1020		1014		1005		1014		1043		1009		1001

Data source: Eurobarometer 494, May 2021. Ordinal logistic regression for each country. Countries of similar profiles by vaccine orientation (Figure 1) or in proximity are placed in proximity columns.

IPVO: Index of pro-vaccination orientation.

^*Dummy variable.

Shadow mark significant coefficients for p < 0.05.