To answer our research questions, we conducted a cross-sectional, online representative survey (in terms of age, sex, and German federal state) using an ISO-certified panel provider (Respondi, now called Bilendi, Germany). We conducted the survey across a stratified quota sample, interviewing an online panel of respondents in Germany from 20 December 2021 to 02 January 2022. In the first step, based on sociodemographic information, a random sample from the population of the online access panel is drawn. After this, a stratification or a quote module is used.

The interviews lasted for 26.8 min on average. Those who stated that they had not been vaccinated against COVID-19 were presented with a list of reasons for not being vaccinated. The whole study was conducted according to the Code of Ethics of the World Medical Association (30). Moreover, all subjects involved in the study gave their written informed consent. All information was collected anonymously.

In total, 1,456 persons (18 years and older) completed the survey. However, the sample was restricted to 1,310 observations, since further analyses only included those with valid responses other than “I do not know” or “Answer refused” in all variables. As can be seen in Table 1, 49% were women and 51% were men; the age ranged from 18 to 74 (M = 45.71, SD = 15.04 years). Of the sample, 85% were located in West Germany and 15% in East Germany. The last column in Table 1 shows the true population values of sex, age group, and German state in 2021. Differences in age groups are due to the online representativeness of our sample.

Approximately 51% of interviewees reported either a university or college entrance degree, while 48% achieved a middle-school or secondary-school diploma and 0.38% had no degree (yet). Most of the respondents reported their net household income to be between €2,600 and €4,999 (41%).

Of the sample, 83% had no so-called migration background, and 17% belonged either to the group of first-generation (6%) or second-generation immigrants (11%). A little over half of the respondents had children themselves (51%).

Notably, 36% stated that they belonged to a COVID-19 risk group, and 92% had not tested positive for COVID-19. However, 50% reported that a relative had tested positive for a COVID-19 infection. With 88%, our sample shows a slightly higher rate of vaccinated people aged 18 years or older than the official statistics in December 2021 with 84% (34).

To determine the factors related to COVID-19 vaccination status (RQ1), we measured sociodemographic variables as well as those related to COVID-19 status, trust in institutions, political views, and COVID-19 media usage. On a descriptive level, we measured trust in different types of vaccines by vaccination status (RQ2) and by the most important reason for not being vaccinated (RQ3). Table 1 depicts the detailed descriptive statistics for all measures.

We used the statistical software Stata 17 SE to prepare and analyze the data. After data cleaning and checks, we conducted univariate analyses for all variables related to RQ1 and RQ2 (see Table 1). To analyze factors related to COVID-19 vaccination status in RQ1, we showed bivariate relationships between vaccination status and independent variables before conducting various multiple logistic regressions with the criterion of COVID-19 vaccination status. Therefore, all the factors described above were analyzed separately, leading to a final model comprising all variables related to the respondents' vaccination status. Since the dependent variable is dichotomous (0 “not vaccinated against COVID-19” and 1 “vaccinated against COVID-19”), we used a logistic regression model rather than a linear probability model (OLS regression model). The logistic function means that the range of predicted values lies between 0 and 1, although the linear combination of independent variables has no limits. Furthermore, the odds ratios are displayed, that is, the ratio of the probability of getting vaccinated and not getting vaccinated.

Therefore, we focused on the odds ratios, the standardized coefficients based on the logit coefficients, their direction and significance, and on the pseudo-R² showing the goodness-of-fit. To analyze trust in vaccines as depicted in the responses to RQ2, we showed means of vaccine trust by vaccination status, using 95% coefficient plots and the unpaired t-tests to check statistical significance between unvaccinated and vaccinated persons. Finally, we addressed RQ3 by showing the univariate distribution of the most important reason for not being vaccinated.

The second column in Table 2 displays the bivariate relationship between each independent variable and the criterion of vaccination status, using a simple logistic regression model and the unadjusted odds ratios of every single predictor based on our data. For the sociodemographic variables, both age and higher income relate positively to being vaccinated, while respondents living in East Germany (compared with West Germany) and those with no school degree (compared with college/university degree) are less likely to be vaccinated. By contrast, there is no relationship between the respondents' vaccination status and either their migration status or whether or not they have children. All COVID-19 status variables correlate significantly with the participants' vaccination status. While belonging to a COVID-19 risk group and knowing someone infected with COVID-19 are factors that relate positively to being vaccinated, a respondent's own COVID-19 infection relates negatively to being vaccinated. While all variables depicting trust in institutions increase the likelihood of a COVID-19 vaccination, the level of right-wing political attitude decreases this probability. The likelihood of getting the COVID-19 vaccination also depends on the type of media a respondent consumes. While those who consume newspapers, public and private media, as well as official sources are significantly more likely to be vaccinated, the likelihood is low for those consuming more social and alternative media.

Based on our first research question, we performed five logistic regression models to investigate to what extent the individual dimensions of sociodemographic background, COVID-19 status variables, trust in institutions and political views, and COVID-19 media usage related to the respondents' COVID-19 vaccination status. Table 2 shows all models of the logistic regression analyses.

Model 1 comprises all sociodemographic variables and demonstrates that age and a household income >€5,000 increase the likelihood of being vaccinated, while residence in East Germany and a lack of school degrees decrease the likelihood. All other factors, such as sex, migration background, and having children, show no relationship to the respondents' vaccination status. However, the explanatory power of all the sociodemographic variables is rather low (pseudo-R² = 0.06).

Model 2 comprises variables directly related to COVID-19. Belonging to a risk group as well as knowing people with a COVID-19 infection increase the likelihood of being vaccinated, whereas one's own infection decreases it. With a pseudo-R² of 0.04, the COVID-19 status variables are correlated with vaccination status very weakly.

Trust in institutions as well as political views, as shown in model 3, have a rather high explanatory power (pseudo-R² = 0.29). Trust in medical experts and authorities as well as in hospitals, rescue workers, and aid organizations increases the likelihood of being vaccinated, whereas trust in companies makes vaccination is less likely. Political views also do not relate to vaccination status compared with the unadjusted model.

According to model 4, those who turn to public and official media to inform themselves about COVID-19 are more likely to be vaccinated. By contrast, those using social media and alternative information channels are less likely to be vaccinated. With a pseudo-R² of 0.21, COVID-19 media usage has the second highest explanatory power after trust in institutions.

Finally, model 5 comprises all variables related to vaccination status and shows a high model fit of pseudo-R² = 0.39. In terms of sociodemographic background, sex, age, and migration status do not correlate significantly with vaccination status. However, respondents from East Germany and people without a school degree are less likely to be vaccinated, while those with a household income of €5,000 or more are more likely to be vaccinated. Furthermore, belonging to a risk group increases the likelihood of being vaccinated against COVID-19, while both the respondents' own and a relative's COVID-19 infection status show no effect. Concerning attitudes toward institutions, trust in political institutions, medical experts and authorities, hospitals, rescue workers, and aid organizations increases the likelihood of being vaccinated, while trust in companies is negatively related. However, trust in state and legal authorities and political views show no correlation with being vaccinated against COVID-19. Where COVID-19 media usage is concerned, consuming information from social media and alternative media corresponds negatively with being vaccinated against COVID-19, while consuming information from newspapers, tabloids, public and private media, and official sources or science does not relate to vaccination status.

Overall, vaccinated people have a significantly higher trust in all types of vaccines than unvaccinated people [BioNTech: t₍₁₂₃₄₎ = −27.65, p < 0.001; Moderna: t₍₁₂₃₃₎ = −25.11, p < 0.001; Astra-Zeneca: t₍₁₂₂₀₎ = −12.97, p < 0.001; Johnson & Johnson: t₍₁₁₉₃₎ = −10.95, p < 0.001; Novavax: t₍₆₈₉₎ = −8.02, p < 0.001; Valneva: t₍₄₆₂₎ = −4.48, p < 0.001].

Vaccinated people have the highest trust in BioNTech, whereas unvaccinated people trust Novavax the most. By contrast, vaccinated people express the lowest trust in Valneva, while those who are not vaccinated trust Astra-Zeneca the least.

All in all, those who have been vaccinated trust mRNA-based vaccines the most, whereas those who have not been vaccinated put their highest trust in the recently developed protein-based vaccines, which use the same technology as influenza vaccines. Regarding unvaccinated people, our data reveal no significant differences between trust in traditional vaccines and other types of vaccines, except for Astra-Zeneca. Figure 1 depicts trust in different types of COVID-19 vaccines.

Among unvaccinated individuals, the wish to “make my own decisions about my body” plays a major role (N = 33), followed by safety concerns (“Vaccines are not safe,” N = 25) and “lack of trust in government” (N = 10). To explain the last three reasons, only three people in the total state that they had a “bad experience with other vaccinations” (N = 1), that “other means help better” (against COVID-19, N = 1), or that they are “fundamentally opposed to vaccinations” (N = 1). Figure 2 shows the most important reason for not being vaccinated.

Our first research question addresses individual factors that may explain the COVID-19 vaccination status. In many regards, our results (based on model 5) identify previous findings on factors related to vaccination willingness. For instance, our results document a moderate correlation between socioeconomic factors, such as income, and vaccination status (11–13). Interestingly, these correlations are especially prevalent in margin categories (e.g., low income is related to low vaccination status). Current scholarship often explains these relationships by pointing to a lack of understanding of vaccine importance and insufficient access to medical care (39). It remains unclear to what extent the latter applies to a sample in Germany, where access to healthcare is comparatively secure. Nevertheless, it seems essential to educate the public about the advantages of vaccination in general and to emphasize that vaccinations are accessible to all and free of charge.

Contrary to previous research on vaccination willingness, our analyses do not show that sociodemographic background, including factors such as age, sex, and migration status, plays a significant role in predicting vaccination status (11–14, 39). Regarding age and vaccination status, it seems plausible that the risk of the severe course of disease eliminates this relationship. Hence, we assume that belonging to a risk group relates to both vaccination willingness and status more than age does. For sex and migration status, none of our models documented a link to vaccination status. As the SAGE vaccine hesitancy determinants matrix (9) suggests, individual and group factors, such as perceived risk and benefits, influence vaccination status, which may explain the non-significant correlation in the present study. Thus, future research should focus on these contradicting results by considering additional factors (e.g., risk perception and language barriers).

Apart from socioeconomic and sociodemographic factors, our study partly replicates results from previous surveys. Thus, it corresponds to previous findings [e.g., (15)] that people who belong to a COVID-19 risk group have a higher vaccination rate. This finding is not surprising, since those in a risk group have a high likelihood of a severe course of infection, hospitalization, and death. This raises the concern that people over 60 years of age are at higher risk of serious infection, due to older age and related comorbidity (40) alone. However, in this age group, there is a massive immunization gap of 1.9 million people in Germany [July 2022, (7)]. Although these people do not see themselves as being at risk, we cannot deny the objective risk of advanced age. Thus, it is important to address unvaccinated people in this risk group both by encouraging general practitioners to educate them individually and by launching public campaigns that inform older people about the immediate risks caused by an infection.

In contrast to belonging to a risk group, which seems to be a reliable factor for vaccination status and willingness, the results regarding respondents' own infection status and knowing someone who was or is infected are inconsistent [e.g., (11, 15)]. Future research should investigate this potential relationship in greater detail, for instance, by focusing on the role of the severity of the course taken by the disease, the impact of infection waves (Delta vs. Omicron variant), the number of subsequent infections, and the strength of social tie to the infected person.

Corresponding with the official COVID-19 statistics, people from East Germany are less likely to be vaccinated than those from West Germany (6). This finding depicts a new phenomenon and differs from the correlation between residence and other disease-related vaccinations (e.g., tetanus or diphtheria), which shows higher vaccination rates for East Germans than for West Germans (41). One explanation may be that in East Germany, there is widespread support of the right-wing populist party AfD (Alternative for Germany), which is skeptical of or refuses COVID-19 measures, such as wearing masks (42), a fact associated with a lower trust in public and state institutions. Thus, fostering trust in public institutions, debunking populist parties' false claims about the COVID-19 measures, and locally targeted health campaigns may help further increase the vaccination rate in the regions concerned.

Furthermore, trust in institutions shows the most significant correlation with vaccination status. A novel finding of our study is that the more people put trust in companies (e.g., food supply), the less likely they are to get vaccinated. One explanation might be that as part of the private sector in the free market, companies tend to symbolize a libertarian point of view, suggesting that the decision to get vaccinated is related to an individualistic, liberal attitude fostering vaccination. This means it is more a personal, private choice rather than a matter of public health. To resolve this association, private companies should be actively involved in a vaccination campaign to increase the percentage of the vaccinated population. Therefore, Dhama et al. (43) proposed a multisectoral approach, defined as an alliance between various agencies from the public and private sectors, to build long-lasting trust in vaccines. In Germany, more than 150 companies advertised a vaccination against COVID-19 together and, therefore, temporarily changed their branding [e.g., chocolate manufacturer Rittersport or supermarkets like LIDL; (44)]. Such strategies could be implemented more extensively to raise vaccination rates in the future.

In contrast to previous research [e.g., (17)], political views as measured on a left-right political scale have no significant relationship to vaccination status. This may be so because trust in public institutions eliminates that relationship. Thus, being politically right-wing decreases institutional trust, which, in turn, relates to being vaccinated negatively. Consequently, building trust in public institutions is more important to the objective of reaching a high vaccination rate, as the next section explains in further detail.

In line with previous research, trust in politics, in hospitals/rescue workers, and, above all, in medical experts and authorities (11, 16, 17, 45) increases the likelihood of getting vaccinated. It is evident that trust plays an important role in ensuring compliance with public health measures in general (45), is related to prevention measures, such as getting vaccinated, in particular, and thus constitutes a highly relevant resource that must be maintained by public institutions. For instance, to cope with a crisis and maintain trust, these institutions must have a stringent communication strategy, which was lacking during the pandemic in public perception (46) (p. 55). Hence, building trust in public health institutions is vital to the objective of increasing the overall vaccination rate. It could be achieved by clear, target-oriented, and effective communication addressing those not fully convinced by the vaccination yet.

As Yang and Huang (47) hold, health communication that combines high-quality information with a traditional communication style, e.g., through banners and posters, leads older people in particular to develop greater trust in science and health professionals. However, social media as a means of communication can decrease trust, demonstrating that the quick spread of misinformation on social media negatively influences people's opinions (48). Corroborating the findings of previous research [e.g., (45)], our study shows that those who get COVID-19 information from social and alternative media are less likely to be vaccinated. There is an especially high negative correlation to vaccination status for those who consume the alternative media on COVID-19 that spread on social media channels such as Telegram. This is due to the fact that fake news, scientific misinformation, and conspiracy theories about health risks are widely distributed through alternative media (49) and shared on social media such as Telegram (50), which then raises these media consumers' concerns about public health measures such as vaccinations. By contrast, our final model suggests that traditional media consumption has no impact on vaccination status. Given that social media consumption is likely to have a negative impact on vaccination status, it is advisable that content from traditional media be disseminated on social media. Although this is already partly the case, future studies should test the strategies for communicating traditional media on social media so as to generate the greatest possible benefit for reaching sufficiently high vaccination rates.

The implications of this study are 2-fold. First, countermeasures, such as online fact checks, marking false statements, and social and alternative media surveillance and regulations, should be increased to both debunk false claims about the COVID-19 vaccinations and alleged side effects and communicate clearly and openly with people who are hesitant to get a COVID-19 vaccine. For instance, Facebook uses tools to alert users that they have read a post containing incorrect information and to make these posts less visible (51). Second, since our study shows that trust in medical experts and authorities is most closely associated with a higher likelihood of getting vaccinated, relevant institutions should reclaim the high ground of COVID-19 coverage on social media to “turn around” the relationship between social media and vaccination status as shown in our results above.

Since we can assume that trust in vaccines is related to vaccination status, our second research question examines the difference of trust in the COVID-19 vaccines between vaccinated and unvaccinated individuals. Our results demonstrate that those who are vaccinated show higher trust in all vaccines, especially in mRNA-based vaccines, than those who are not vaccinated. This is because most people in Germany received mRNA-based vaccines (6) and so place the greatest trust in these vaccines in order to, for instance, reduce cognitive dissonance (52). In line with the rather negative media coverage (53), Astra-Zeneca is rated badly compared with other vaccines. Such a correlation is not a new phenomenon and can also be observed when persons are exposed to health media in mass media, decreasing patients' belief in the benefits of medication (54). This phenomenon also emerged for other vaccines, such as that against human papillomavirus (HPV) (55). Furthermore, people tend to place greater weight on negative media reports (56), which could lead to a more elevated risk perception of possible side effects and so to decreased trust.

The vaccines Novavax and Valneva, based on a “traditional” technology used for influenza vaccines, are often seen as an alternative for unvaccinated people, who trust these vaccines the most, and thus as a way to increase the vaccination rate (25). However, trust in these vaccines is still low and does not differ from other vaccines so much, which is also reflected in the fact that Novavax has not increased the demand for vaccinations significantly since its roll-out in February 2022 (57). According to the German Minister of Health, this might be due to fake news on social media claiming Novavax would cause cancerous tumors, among other things (58). Such claims spread faster than public vaccination campaigns for the use of Novavax. This demonstrates the need to implement a vaccination campaign for the other protein-based vaccine Valneva before its release so as to build trust long in advance and debunk false claims as soon as they start to spread even on a small scale.

Although the results of our survey demonstrate a relationship between trust and vaccination status, it remains unclear to what extent this trust is generated by knowledge about the vaccine and attitudes toward the manufacturer arising from brand awareness. Recapitulating the massive media presence of the manufacturers and the intensive coverage of side effects [e.g., Astra Zeneca; cerebral venous thrombosis (53)], it seems plausible that trust cannot be traced back solely to knowledge about the vaccines. Therefore, future studies should investigate factors (e.g., knowledge and brand awareness) that may explain the relationship between trust and vaccination status.

Our last research question aims to identify the most important reason for not being vaccinated against COVID-19. Our analyses revealed that the first reason is the respondents' wish to make their own decisions about their bodies, followed by doubts about vaccine safety. This result partly contradicts research on people's reasons for refusing vaccination when vaccines were not available yet. These studies cited concerns about safety, side effects, and the vaccines' fast development as people's most prominent reasons against vaccination (27–29), demonstrating a move from “general” safety reasons to highly individual reasons concerning the conditions of one's own body. Moreover, a lack of trust in the government fully corresponds to previous research (16) and findings in this study on factors related to vaccination status as described above.

In other words, the issue at hand mainly concerns decisions about and control of one's own body, which is relatively difficult to address in public vaccination campaigns compared with the safety worries expressed in former studies. Our study reveals a more affective and very individual statement about the desire to keep full control of one's own body and the freedom to do so. Because public vaccination campaigns on traditional or social media platforms may fail to address that highly individual and ethical reason, a person who knows the other's specific body condition well enough is central: Therefore, general practitioners play a key role since doctors know their patients' body condition best, while patients “[…] are likely to establish trust with known health professionals, as their experience of that person increases” (59) (p. 2). Hence, healthcare providers' recommendations have a strong positive relationship to vaccination willingness [e.g., (60) for hepatitis B vaccination]. At the same time, studies have shown that healthcare practitioners' training is essential in treatment to counteract existing fears about vaccination by informing and involving patients in the process of decision-making. Doing so increases the adherence to medical recommendations [e.g., (61)]. Conversations with the general practitioner, called “second-tier” or informal support, are essential to making an informed decision about whether to get vaccinated (62) and are thus an effective element in increasing the vaccination rate against COVID-19. According to Bartoš et al. (63), while ~90% of medical doctors trust COVID-19 vaccines, most respondents from the general public believe that only half of the doctors trust the vaccines. Future vaccination campaigns should address this misconception by actively integrating medical doctors.

Due to research and survey design, our study is subject to several methodological limitations. First, our data were generated using an online panel. Even though we aimed to fully represent the German population, our data are only online representative of sex, age groups from 18 to 74 years, and federal state. Moreover, our data did not gather specific groups, such as people younger than 18 years or older than 74 years or those with no online access at all, which is clear from the demographic differences between our sample and the actual German population in 2021 as shown in Table 1. Therefore, future studies should also conduct telephone (CATI) or personal interviews (CAPI), in order to minimize these restrictions. Furthermore, our study was designed as a cross-sectional survey to collect data. Consequently, interpretations of the data are linked to the time and location of data collection and are thus subject to a certain reactivity, since surveys concerning vaccination status depend on dynamic contextual factors. While the cross-sectional design prevents us from drawing other causal conclusions, repeating this survey using longitudinal data with the same sample could allow for more causal statements.