Determinants of voluntary compliance

There is ample evidence that trust, in particular institutional trust, is a powerful determinant of voluntary compliance with government policies (Devine Reference Devine2024). Higher levels of trust in government and its adjunct institutions are found to be positively related to voluntary compliance (Chanley, Rudolph and Rahn Reference Chanley, Rudolph and Rahn2000; Marien and Hooghe Reference Marien and Hooghe2011). High levels of trust in government and institutions have been associated with effectively implemented policies and the rule of law and indicate that a citizen is satisfied with government performance more broadly (Rothstein and Uslaner Reference Rothstein and Uslaner2005; Van der Meer and Hakhverdian Reference Van der Meer and Hakhverdian2017; Newton, Stolle and Zmerli Reference Newton, Stolle, Zmerli and Eric2018; Herreros Reference Herreros2023). Trust becomes particularly relevant in times of crisis and high levels of uncertainty (van Middelaar Reference van Middelaar2021; Lalot, Heering, Rullo et al. Reference Lalot, Heering, Rullo, Travaglino and Abrams2022), such as during the early stages of the pandemic, when many citizens felt threatened by the evolving (health) crisis. Studies show that in such situations, more trusting people are more likely to comply with mandated rules and regulations than citizens with lower levels of trust (Lalot, Heering, Rullo et al. Reference Lalot, Heering, Rullo, Travaglino and Abrams2022).

Recent contributions to the literature examining compliance with pandemic-related restrictions show that trust in government, and other relevant institutions, is positively correlated with the observance of mandated restrictions, such as social distancing rules or restrictions on people’s freedom to move, and vaccine uptake (Devine, Gaskell, Jennings et al. Reference Devine, Gaskell, Jennings and Stoker2021; Goldstein and Wiedemann Reference Goldstein and Wiedemann2022; Han, Zheng, Cristea et al. Reference Han, Zheng, Cristea, Agostini, Bélanger, Gt¨zkow, Kreienkamp and Leander2023; Devine, Valgardsson, Smith et al. Reference Devine, Valgardsson, Smith, Jennings, di Vettimo, Bunting and McKay2024; Green, Evans and Snow Reference Green, Evans and Snow2025). However, it remains unclear how trust interacts with policy measures and individual attitudes. We do not have a good understanding of whether those who trust more in institutions are also more likely to respond to other policy incentives, or whether policies can specifically target those with low trust. Similarly, we do not yet understand whether trust and other attitudes, such as ideology, amplify each other’s effects or not. Understanding these dynamics is crucial for designing effective policy interventions. If coercive policy only works on high-trusting individuals, who are already more likely to comply, then the policy would increase the gap between those with low and high trust.

We theorize two possible scenarios for how trust might interact with coercive policies and related individual attitudes such as ideology: trust as an ‘amplifier’ (‘amplification’ effect) and trust ‘compensating’ policies (‘compensation’ effect).Footnote ² For simplicity, for the remainder of this discussion, we will only refer to trust and coercive policies, but the same logic applies to other individual attitudes such as ideology. An amplification effect means that the effects of trust and policy coercion combine and add up on top of each other. Consider a hypothetical example of how an educational intervention, such as an after-school program, might interact with student IQ to influence future test scores. In an amplification scenario, those with high IQs would benefit the most from the program and increase their test scores more than students with lower IQs. The gap between the two groups would widen with the intervention. In a compensation scenario, the intervention would have the highest impact for those with lower IQs, as they have more room to make gains. In this scenario, the gap between the two groups would narrow after the intervention.

In our case, if trust ‘amplifies’ the effects of coercive policy, we expect to find stronger positive effects for highly trusting individuals. Under this scenario, individuals with a higher level of trust may be more likely to get vaccinated under a coercive policy because they trust that the chosen policy instrument is necessary, and its use is legitimate. Compliance, in this scenario, would require both trust and coercion, while those with lower trust would not see much change. As a consequence, the gap between the two groups would widen. If, instead, trust and coercive policy instruments are ‘compensatory’, we expect to see stronger positive effects of policies for low-trusting individuals as they react to the prospect of sanctions or fines in cases of non-compliance. High-trusting individuals should be highly likely to get vaccinated even in the absence of the policy.Footnote ³ We remain agnostic as to which element is the compensatory one. The ‘compensation’ scenario simply claims that either element (trust or policy instrument) is sufficient to increase policy compliance and fill in for the missing part.

Figure 1 illustrates the difference in a 2 × 2 heat map. Panel (a) shows the compensatory effect. Moving from low coercion to high coercion has an effect for those not trusting (0.5 increase moving from the bottom left cell to the upper left cell), but no additional effect for those who trust (an effect of 0 moving from the top left cell to the top right cell). Conversely, the effect of increasing trust is negligible for those already affected by coercion, but important for those without coercion. Panel (b) shows the amplification effect. The effect of policy coercion for those not trusting is now 0.35, moving from the bottom left cell to the upper left cell. Now there is an additional effect for those who trust, of 0.45 when moving from the top left cell to the top right cell. The two effects now amplify each other, and it is those with high trust and subject to coercion who are most likely to comply. Notice that the overall effect is the same. It is the distribution of the effect among the cells that is captured by our two scenarios.

Figure 1.

Illustration of compensation vs. amplification effects.

We propose that trust in the government, health institutions, and the health industry helps citizens to better handle situations in which information to assess the course of action proposed by the government and health experts is limited. In the case of COVID-19, the level of uncertainty was very high, first due to the novelty of the threat (new type of disease) and later due to the novel type of vaccination that became available to citizens in early spring 2021. Because institutional trust also extends to policies mandated and supported by the relevant institutions (Levi and Stoker Reference Levi and Stoker2000; Rothstein and Stolle Reference Rothstein and Stolle2008), we expect people with higher levels of trust – on average – to have a significantly higher likelihood of getting vaccinated than citizens that have a lower level of trust. Furthermore, we propose that higher levels of trust among non-vaccinated citizens increase the likelihood to eventually get vaccinated. We therefore expect trust to be ‘compensatory’ with policies and individual characteristics, rather than ‘amplify’ them. However, we form contrasting hypotheses to test the two divergent mechanisms.

In addition to trust (whether in its amplifying or compensating form), a citizen’s ideological disposition also influences how they perceive and evaluate their governments’ policy advice. We expect that citizens who situate themselves more to the left are more likely to comply with government policy to get vaccinated due to their – on average – higher level of prosocial behavior (Schubach and Thielmann Reference Schubach and Thielmann2025), while citizens who are more on the right of the political spectrum are less likely to comply with the vaccination mandate. The underlying mechanism that informs this proposition is that citizens who are more conservative and economically liberal are – on average – less supportive of policies that conflict with an individual’s self-determination and personal freedom. Citizens on the conservative spectrum are also less supportive of state interventions than citizens who position themselves more on the left of the political spectrum (Feldman Reference Feldman and Leonie2013).

Furthermore, compliance often comes with (immaterial) personal costs and unclear benefits (Scholz and Lubell Reference Scholz and Lubell1998). This was in particular the case with the vaccination mandate as the virus was not perceived as a threat by some groups within the society, while their cooperation was still needed to achieve immunity levels high enough for vulnerable groups to be protected. There is evidence that citizens who are more on the left of the political spectrum are more solidaristic and socio-tropic and therefore more concerned with the overall public health situation than citizens who position themselves more on the right side of the political spectrum (see Piurko, Schwartz and Davidov (Reference Piurko, Schwartz and Davidov2011); Evans and Neundorf (Reference Evans and Neundorf2020)). Based on this, we expect respondents who place themselves more to the left of the political spectrum to comply – on average – more with the vaccination mandate than respondents on the right of the spectrum.

During COVID, in particular during its early stages, decisions affecting people’s everyday life were taken under severe time pressure, with limited knowledge and deliberation. Many of the policies to contain the virus were suggested by health experts and implemented in a top-down manner. Decision-making processes took place outside of regular legislative processes, or legislative processes were significantly cut short. This instigated criticism and opposition from populist parties and their supporters (Rovny, Bakker, Hooghe et al. Reference Rovny, Bakker, Hooghe, Jolly, Marks, Polk, Steenbergen and Anna Vachudova2022; Schwörer and Fernández-García Reference Schwörer and Fernández-García2022). The criticism voiced by these parties was rooted in their critical view of experts and the ruling elites, and the fact that decisions taken would ignore the ‘will of the common people’.

The relationship between populism, institutional distrust, and policy compliance is theoretically complex and potentially circular. Institutional distrust has been shown to be a determinant of support for populist parties (Geurkink, Zaslove and Jacobs Reference Geurkink, Zaslove and Jacobs2020), with populist rhetoric that deepens distrust over time. This dynamic can create a self-reinforcing cycle (Hooghe and Dassonneville Reference Hooghe and Dassonneville2018). This dynamic extends specifically to (scientific) experts and health institutions through what scholars term ‘science-related populism’ (Mede and Schäfer Reference Mede and Schäfer2020): the belief that academic elites ignore ordinary people’s common sense when determining what constitutes legitimate knowledge. Given this, citizens who support populist parties may be particularly resistant to both trust-based appeals (or voluntary compliance with government-mandated policies and expert-driven recommendations). Whether coercive policies compensate for this distrust or amplify populist resistance remains an open empirical question that our design allows us to test. Given this, we also expect supporters of populist parties to be more skeptical of the vaccination mandate and more likely to not be vaccinated.Footnote ⁴

Hypotheses:

Coercive policy to increase vaccine uptake

If voluntary compliance with the policies implemented is not sufficient, governments can raise the costs of non-compliance for citizens. Such costs can take different forms. During the pandemic, the majority of countries, for example, fined people when violating social distancing rules, lockdowns, or other measures taken to contain the pandemic.Footnote ⁵ When it comes to vaccination, however, fining people for not complying with the mandate is incompatible with citizens’ right to physical integrity. This is why a variety of other strategies to incentivize vaccination take-up were implemented. In line with already existing knowledge (see ie Betsch, Schmid, Heinemeier et al. (Reference Betsch, Schmid, Heinemeier, Korn, Holtmann and Böhm2018)), COVID-19 vaccines were administered free of charge, and public authorities aimed to make it as easy as possible for citizens to obtain vaccination appointments. Given the novelty of the disease and the vaccine, governments and relevant other actors also engaged in large information campaigns to build confidence in the vaccine and educate citizens.Footnote ⁶

However, governments realized that free and easy access and the effort to build confidence in the newly developed vaccines did not yield sufficiently high numbers of vaccinated citizens. To avoid that the public health systems would be overburdened again during the fall and winter of 2021, governments across Europe changed strategy and switched to more coercive tools.Footnote ⁷ On the one hand, public authorities tried to call on citizens’ sense of responsibility and duty by moralizing non-compliance with the vaccination mandate (Bor, Jörgensen, Lindholt et al. Reference Bor, Jörgensen, Lindholt and Bang Petersen2023) and also emphasized again the risks of an infection.Footnote ⁸ Results from recent studies indicate that these mechanisms did have the expected effect. Schraff (Reference Schraff2021) shows that anxiety or fear was indeed a significant determinant of compliance. Similar results were found by Jörgensen, Bor and Bang Petersen (Reference Jörgensen, Bor and Bang Petersen2021) who show that fear motivated compliance with protective behavior.

On the other hand, and going beyond mere talk, governments announced measures that would significantly restrict the individual freedoms of unvaccinated people by, for example, banning them from visiting restaurants, social events, and participating in leisure activities, or traveling abroad. Such activities and venues would only be open to people presenting so-called ‘vaccination passports’ or ‘green passes’. The requirement to present a vaccination passport for a variety of activities was expected to serve as a strong incentive for those citizens still undecided and not necessarily those with very strong anti-vaccination convictions.

Among those who are already vaccinated, we expect that learning about the upcoming restrictions, on average, will increase the willingness to get a booster shot. As with individual characteristics, we form contrasting hypotheses for the role of trust. If trust ‘compensates’ the policy, then people whose level of institutional trust is high already are very likely to get a second shot with or without the policy. People with low levels of institutional trust, however, should react positively to the policy. If trust ‘amplifies’ the effects of the policy, then we should see higher levels of compliance for trusting individuals under the policy, with lower or negative effects for low-trusting individuals.

We do not expect that the implementation of a vaccination passport affects the behavior of citizens with strong anti-vaccination convictions. On the contrary, such policies may lead to an even stronger resentment of the vaccination mandate (Jörgensen, Bor and Bang Petersen Reference Jörgensen, Bor and Bang Petersen2024).

Instrument-Specific Hypotheses:

Rolling it out: COVID-19 vaccination in the United Kingdom and Germany

In this section, we provide relevant background information surrounding the timing of our surveys. Specifically, we provide insight into the period between January 2021 and January 2022. This covers the time when the first vaccines were administered to vulnerable citizens to the time when the majority of citizens in Western countries had the chance to get two (or three) shots of the available COVID-19 vaccines. With a little more than one year into the pandemic, the prospect of an effective vaccine was welcomed with a lot of enthusiasm among citizens in the two countries that are part of this study and in Western Europe more generally. While we cannot draw firm conclusions regarding the impact of these factors on our findings, we nonetheless argue that it is important to understand what the general public health situation in each of the two countries was at the time.

The development of an effective vaccine was seen as the key to overcoming the public health crisis and saving lives. Indeed, vaccine development began as soon as the genetic sequence of COVID-19 was released on 10 January 2020 (Fauci Reference Fauci2021). Less than a year later, ‘Phase 3’ trials showed several vaccines with more than 90% efficacy. As with many health-related prevention measures, vaccines work best if a majority of the target population complies with the policy. The prospect that the vaccine would be on offer to the mass public (in the Western hemisphere) shortly (with first vaccines applied to vulnerable groups of people already late 2020) was greeted with enthusiasm and a sense of relief shared by many, indicating that achieving a substantively high proportion of vaccinated citizens should be reached with ease. In some countries, however, the initial enthusiasm faded rather quickly and gave way to a variety of controversies surrounding the application of some vaccines (ie AstraZeneca) on the one hand, and criticism about how the vaccination roll-out was managed, on the other hand. It became clear that the ambitious goal of reaching ‘herd immunity’ through a very high proportion of citizens who comply with the government’s call to get vaccinated was likely too ambitious.

Parallel to these vaccine-specific discussions, debates on how public health can be maintained and overcrowding of public health institutions avoided continued, fueled by a new variant of the virus (Delta-variant at that time) that proved to be more transmissible than earlier types. That also meant that there was a continued need to protect vulnerable groups within societies without restricting individual freedoms too much. In this situation, policymakers started to discuss new policy schemes (so-called ‘green passes’) that would restrict the range of activities accessible to unvaccinated people with the aim to a) protect the vaccinated and vulnerable groups and b) increase the pressure on unvaccinated citizens to comply with the vaccination mandate during this crucial period.

Figures 2 and 3 show how the pandemic evolved over these 12 months and indicate political decisions relevant for our paper. Specifically, we show the development of the caseload (in number of new cases/100,000 inhabitants), the number of fully vaccinated citizens (as percentage points), the number of people who got their booster shot (CSSE 2023), and the level of restrictions as captured by the Oxford Stringency Index (OxCGRT 2020). The vertical lines visualize the timing of the survey and relevant political decisions related to restrictions that were imposed to manage the pandemic and alleviate the pressure on public health institutions during that year.

Figure 2.

UK timeline of COVID cases, vaccinations, and boosters.

Note: Solid black line: Cases per 100,000; Dotted green line: Stringency Index; Dot-dash blue line: Fully vaccinated citizens per 100; Dashed red line: Boosters per 100 (sources: (OxCGRT 2020; CSSE 2023)).

Figure 3.

German timeline of COVID cases, vaccinations, and boosters.

Note: Solid black line: Cases per 100,000; Dotted green line: Stringency Index; Dot-dash blue line: Fully vaccinated citizens per 100; Dashed red line: Boosters per 100 (sources: OxCGRT 2020; CSSE 2023).

Comparing the timelines between the United Kingdom and Germany shows differences in the evolution of the pandemic in late March. At the time, the vaccination roll-out had gained a significant pace in both countries. While in the United Kingdom, the caseload during spring 2021 was relatively low, Germany lived through a small wave of infections between March and May 2021. The United Kingdom started to administer the first vaccinations at the end of 2020 prioritizing front-line workers, particularly in the health care and educational sector. The political discourse during the first phase of the vaccination roll-out in the United Kingdom was marked by a great level of enthusiasm and the way the vaccination campaign was organized was generally well received. During the first months of the mass roll-out, the uptake by the general public was quick, and by mid-July almost half of the adult population was considered fully vaccinated. This coincided with the lifting of all restrictions at the peak of the COVID-19 wave caused by the newly emerging Delta-variant. By the time we fielded our survey, more than 70% of the overall population was fully vaccinated and the curve (representing the fully vaccinated population) flattened out. The booster campaign started in the fall, shortly after the collection of our data.

In Germany, the vaccination roll-out was considerably slower. The reasons for the slightly later start of the vaccination campaign were related, on the one hand, to the coordinated approach by EU member states during the negotiation and purchase period and the distribution of the vaccines among the EU member states. On the other hand, the pace was slowed down by complicated domestic distribution processes and constraints, such as the fact that general practitioners (at the start of the vaccination drive) were not allowed to administer COVID-19 vaccines, chaotic scheduling procedures for vaccination appointments, and malfunctioning vaccination centers. These struggles also had a negative impact on how the vaccination campaign set off. Indeed, Germans were considerably less satisfied with the vaccination campaign compared to other Europeans (Truchlewski, Kriesi, Moise et al. Reference Truchlewski, Kriesi, Moise and Oana2025). The slow start was further complicated by controversial decisions and discussions related to the AstraZeneca vaccine which caused confusion among part of the population and led to an increase in vaccine hesitancy.Footnote ⁹

We see a similar pattern emerging with the uptake of the booster shot. While in the United Kingdom, a large number of citizens readily accepted to get a booster shot, likely also incentivized by the prospect of the re-introduction of restrictions, in particular the necessity to have a COVID passport to participate in certain activities, the uptake of booster shots in Germany was slow and the curve shows signs of flattening toward the end of the period visualized in the graph. This is in line with other research which showed that UK citizens were among the most enthusiastic to receive vaccinations for COVID-19, while enthusiasm was much lower in Germany (Lazarus, Ratzan, Palayew et al. Reference Lazarus, Ratzan, Palayew, Gostin, Larson, Rabin, Kimball and El-Mohandes2020). Other research, which came after our data collection, confirmed that Germans were more likely to be skeptical of the development of vaccines, their safety, and have higher distrust of authorities (Desson, Kauer, Otten et al. Reference Desson, Kauer, Otten, Peters and Paolucci2022; Reis et al. Reference Reis, Michalski, Bartig, Wulkotte, Poethko-Müller, Graeber, Schaffrath Rosario, Hövener and Hoebel2024). We conclude that German respondents have more room to be influenced by our coercive policy instrument.

Two of the timeline factors that we discuss are crucial for the design and the timing of our survey. First, our survey came after the initial vaccination campaign was completed, in the sense that all adults had had a chance to get their vaccination, and before the start of the booster campaign. We utilize the timing of the fielding phase to perform two analyses: one on recall of the initial vaccination (observational) and another on the self-reported likelihood of getting the booster shot using a vignette experiment (or the initial vaccine for respondents that were still unvaccinated).

The second factor is related to the timing of the introduction of restrictive green pass policies. In the United Kingdom, general restrictions, such as limitations for group gatherings and shop closures, were lifted in July 2021, and partially reimposed only in December, when the COVID Pass was introduced.Footnote ¹⁰ In Germany, the 3G rule had been in place since 23 August 2021,Footnote ¹¹ which meant that to access public spaces, Germans needed to be vaccinated, recently healed from COVID, or have a recent test (geimpft, geheilt, and getested). Unvaccinated individuals therefore had the option to continually test to avoid restrictions. This changed when a new wave, fueled by the Omicron variant, hit the country. On 3 December 2021, Germany imposed more severe restrictions on the unvaccinatedFootnote ¹² which limited their access to public spaces and social gatherings. Accordingly, neither our respondents in the United Kingdom nor those in Germany were ‘pre-treated’, in the sense of living under a policy regime where a policy similar to our ‘green pass’ experimental treatment already existed.

Data

We conducted an experiment embedded in a population-based survey. The surveys were fielded by Respondi (now ‘Bilendi’) in late summer 2021 (between September 8 and 23), making use of different country-specific online access panels.Footnote ¹³ Voting-age nationals were selected from these access panels using gender- and age-based quotas. In each country, we surveyed around 1200 individuals. The survey consisted of three sections.

In the first section, we collect information about respondents’ political views and preferences as well as their levels of trust in different types of organizations and institutions: trust in private pharmaceutical companies, in medical research centers, in government, and in the national health service. For the analysis, we create a factor to capture the underlying level of trust in institutions.Footnote ¹⁴ We also asked respondents a number of questions regarding their satisfaction with government policy toward COVID: satisfaction with business economic support, vaccination roll-out, economic support for individuals, protection of vulnerable groups, restrictions, and personal freedom. We combine these attitudes into a factor capturing underlying satisfaction with COVID policies.Footnote ¹⁵

In addition, we also leveraged their attitudes toward a number of economic policies, and their understanding of how responsibilities should be distributed between politics and experts. All respondents were asked to indicate the extent to which they were worried about COVID-19 and whether or not they ever tested positive for COVID-19. Furthermore, we also asked respondents about their vaccination status.Footnote ¹⁶ Respondents who indicated that they did not have the intention to get vaccinated were asked to indicate the reason that best explains their objection from a list of commonly cited reasons for vaccinating hesitancy.

The second part of the survey consisted of a ‘vaccination’ experiment, which we will discuss in more detail below. The third section collected information about participants’ socioeconomic background, such as the level of education, income, and family status. In what follows, we present key information about the distribution of vaccinated vs. unvaccinated people in our sample and provide insight about respondents’ objections to get vaccinated.

Figure 4 shows that at the time of our survey, 83.2% of our German respondents were already vaccinated.Footnote ¹⁷ Official data at the time show that in Germany at that point in time, 73.4% of the adult population was vaccinated (ECDC 2023). This indicates that our survey slightly over-sampled vaccinated people in Germany.Footnote ¹⁸ At the same time, the proportion of respondents who indicated that they are vaccinated is slightly higher in the United Kingdom (88%), in line with the vaccination rate reported by the NHS at the time (86%) (NHS 2025). 22.5% of the non-vaccinated respondents in the United Kingdom and 29.9% of the non-vaccinated German respondents indicated that they did not intend to get vaccinated.Footnote ¹⁹ A very small share of our respondents in both countries preferred not to answer this question.

Figure 4.

Vaccination status in sample.

Figure 5 shows that a plurality of respondents in both countries who were not vaccinated when fielding our survey indicated that they thought that the vaccine was unsafe (40% of British and 30% of German respondents, respectively).Footnote ²⁰ The second most important factor to objecting vaccination for German respondents was the fast development of the vaccine (roughly 22%) followed by the fear of potential side effects (19%). In the United Kingdom, the second largest group of non-vaccinated respondents deemed it unnecessary to get vaccinated (19%). All other categories received a similar share of responses. The survey also allowed respondents to indicate other factors that contributed to their decision. One reason indicated by several respondents was the feeling that the state/government was forcing them to get vaccinated, which – in the end – led them to abstain.Footnote ²¹ Another reason that respondents offered was related to a general sense of mistrust and fear that the state would fool and harm its citizens.Footnote ²²

Figure 5.

Reasons for vaccine hesitancy.

Experimental design

For the experimental part, participants were first divided into two groups, which differ in their vaccination status. The first group consists of respondents who indicated that they are already vaccinated. The second group consists of respondents who indicated that they were not vaccinated. Depending on whether a respondent belongs to the group of vaccinated or unvaccinated people, they were exposed to different scenarios. Prior to the group-specific scenario, all respondents read the following disclaimer: ‘We are now going to present you with a scenario regarding the vaccination campaign in [the United Kingdom/Germany]’. Table 1 details the experimental setup.

Table 1.

Experimental setup

After having read the vignette, respondents were asked to answer the following outcome question: Considering the above scenario, how likely are you to get vaccinated against COVID-19, for the unvaccinated group, and Considering the above scenario, how likely are you to get vaccinated with the ‘booster shot’ against COVID-19?, for the vaccinated group.Footnote ²³

Figure 6 shows the distribution of our first outcome variable, the one that captures respondents’ reaction to the vignette experiment. Unsurprisingly, the proportion of respondents among the already vaccinated who indicate that they will also get a COVID-19 booster shot is very high in both countries. Combining the categories capturing citizens who responded that they are ‘likely’ or ‘extremely likely’ to get the booster shot amounts to more than 80% in both countries. Turning to the group of unvaccinated respondents shows that the majority (between 60% and 70%) in this group indicate that they are unlikely to get vaccinated. Only roughly 15% in both countries are undecided.

Figure 6.

Distribution of dependent variable.