Do Social Norms Influence Young People’s Willingness to Take the COVID-19 Vaccine?

ABSTRACT

Although young adults are not at great risk of becoming severely ill with COVID-19, their willingness to get vaccinated affects the whole community. Vaccine hesitancy has increased during recent years, and more research is needed on its situational determinants. This paper reports a preregistered experiment (N = 654) that examined whether communicating descriptive social norms – information about what most people do – is an effective way of influencing young people’s intentions and reducing their hesitancy to take the COVID-19 vaccine. We found weak support for our main hypothesis that conveying strong (compared to weak) norms leads to reduced hesitancy and stronger intentions. Furthermore, norms did not produce significantly different effects compared to standard vaccine information from the authorities. Moreover, no support was found for the hypothesis that young people are more strongly influenced by norms when the norm reference group consists of other young individuals rather than people in general. These findings suggest that the practical usefulness of signaling descriptive norms is rather limited, and may not be more effective than standard appeals in the quest of encouraging young adults to trust and accept a new vaccine.

The COVID-19 pandemic has resulted in millions of deaths, poorer physical and mental health, lockdowns, and impoverished economies. Fortunately, new vaccines have been produced at an unusually rapid pace and is being distributed across the world. The World Health Organization has estimated that vaccination prevents approximately 2.5 million deaths each year (World Health Organization, 2013), and it is considered one of the most successful public health measures (Dubé et al., 2015). However, vaccine accessibility is not by itself sufficient to ensure vaccine coverage – vaccine acceptance is also necessary.

Vaccine hesitancy and acceptance

Vaccine hesitancy refers to a delay in acceptance or refusal of vaccination (Dubé et al., 2015). The contributing factors vary depending on the specific vaccine or disease, and individual as well as social influences and contextual circumstances (MacDonald & The SAGE Working Group on Vaccine Hesitancy, 2015; Shapiro et al., 2018). It seems plausible, however, that vaccine hesitancy can be socially contagious. Recently, the anti-vaccine movement has gained more ground and contributed to a lowered vaccine acceptance rate (Lewandowsky et al., 2012) and therefore, to an increase in outbreaks of diseases that could have been prevented, such as measles (Olive et al., 2018). Vaccine-related misinformation and conspiracy theories flourish on the internet and social media (e.g., Van Prooijen & Douglas, 2018), and continues to do so during the Covid-19 pandemic. The spreading of misinformation is likely to affect perceived risks with the vaccine and encourages vaccine hesitancy (Betsch et al., 2010). Indeed, Huynh (2020) showed that during the pandemic, concerns about potential side effects from vaccines and about profiteering pharmaceutical companies have increased, whereas concerns about vaccine effectiveness have not. Young people may be particularly susceptible to social influences on vaccine attitudes, considering that they frequently use social media and tend to go online to seek health information (Rideout et al., 2018).

Social norms and health-related decisions

Because of our inherently social nature, human beings tend to be attentive to cues about which behaviors, attitudes, and values that are acceptable among most people. This attention is so fundamental that it often occurs automatically, and people tend to underestimate the extent to which their actions are influenced by social norms (Nolan et al., 2008). The research literature typically distinguishes between two broad classes of social norms: descriptive and injunctive norms (Cialdini et al., 1990). Whereas injunctive norms advise us about what we ought to do; for example, that we should try our best to protect other people by preventing the spread of the coronavirus, descriptive norms tell us how most people behave, for example, that most people want to get vaccinated against the disease. Observing what others do can often provide clues about an appropriate course of action, as behaviors that are common signal that they are accurate, effective, and/or that they elicit social approval (Mollen et al., 2010). Apart from producing disapproval from others (Cialdini et al., 1990), deviance from social norms can stir negative affect and lead to reduced well-being (Sassenberg et al., 2011). Indeed, being able to adjust one’s behavior to fit in with one’s group and in the community is crucial for social functioning in society. These tendencies are for better or worse, as social norms can encourage healthy behaviors (e.g., exercising, Okun et al., 2002; healthy eating, Mollen et al., 2013; hand washing, Curtis et al., 2009), as well as unhealthy or harmful ones (e.g., smoking cigarettes or sharing needles, Reid et al., 2010; drinking alcohol; Bewick et al., 2010). If people learn that the majority want to take the coronavirus vaccine, this may reassure them that this behavior is suitable and effective and that it is a good idea to follow suit (Chung & Rimal, 2016; Legros & Cislaghi, 2020). Conversely, if people learn that others are skeptical or unwilling to take the vaccine, they too may start having doubts.

Several correlational studies have examined the association between perceived norms and vaccine attitudes and intentions (e.g., Allen et al., 2009; Dillard, 2011; Nyhan et al., 2014). Most of these studies have found positive associations between norms and vaccine attitudes. However, although this relationship could be due to subjective norms influencing vaccine intentions, another possibility is that those who want to take a certain vaccine are motivated to think that other people also plan to take it, as this could serve to validate their beliefs. Correlational studies on social norms are thus unable to rule out the problem of reverse causality. Only one study (that we know of) has previously examined the experimental effects of signaling descriptive norms on vaccine intentions or hesitancy. Recently, Xiao and Borah (2020) examined in an explorative manner the effects of descriptive and injunctive norms on college students’ intentions to take the human papillomavirus (HPV) vaccine. They found no effect on intentions to take the vaccine, but mixed effects on intentions to seek information and on perceived risks with the vaccine. However, a methodological weakness was that the majority of their participants had already completed (60% of the sample) or initiated (73%) the HPV vaccine series. Thus, participants’ reported intentions may have been based on previous vaccination behavior.

Because the COVID-19 pandemic is new and ongoing, information about the virus is constantly being updated and different experts are sometimes disagreeing with each other. People may therefore find themselves unsure about of how to behave, suggesting that descriptive norms could serve as a powerful source of information in this context. Indeed, whether norms are formed through direct observation (such as descriptive norms) or imaginatively (more likely for injunctive norms) during a pandemic, is likely to affect their resilience and their power to influence behaviors (Rimal & Storey, 2020), suggesting that descriptive norms should be of particular interest in the context of a pandemic. To the best of our knowledge, this study is the first to test the causal effect of communicating descriptive norms on intentions to be vaccinated against COVID-19. Examining this is of practical importance, as signaling information about what other people do could be a straightforward and inexpensive way to aid in the pursuit of minimizing vaccine hesitancy.

The importance of similarity to the reference group

People conform to norms not only for guidance on a correct course of action; behaving the same way as other people can also be a way to obtain social approval and bond with others (Mollen et al., 2010). Indeed, Young (2015) proposed that one of the most important compliance mechanisms that explain why people conform to social norms is what he labeled Signaling and Symbolism: People want to signal their membership in a given group to themselves and/or to others, and they therefore follow what they perceive to be the rules specific to that group. From a social identity perspective (Tajfel & Turner, 1979), adherence to ingroup norms is not only externally but also internally motivated because feeling similar to ingroup members leads to a positive self-image (Sassenberg et al., 2011). This implies that within the social identity framework, the distinction between descriptive and injunctive norms is less clear: Behaviors that are common among ingroup members not only describe behavior but also prescribe it, because they tell members how they should behave to fit in (Hogg & Reid, 2006). For example, a recent study by Goldring and Heiphetz (2020) found that the more common people believe that certain behaviors are in the ingroup, the more moral they perceive these behaviors to be (and this effect was not as prominent for the outgroup). The tendency to be influenced by perceptions of what others do should therefore be especially likely when one identifies with, or feels similar to, the (norm) reference group (Christensen et al., 2004). In line with this theorizing, White et al. (2009) found that subjective norms predicted recycling intentions, particularly for individuals who identified strongly with the group; Louis et al. (2007) found with a longitudinal design that perceived norms for students’ eating behavior interacted with participants’ identification as students to predict intentions of healthy eating; and Lee and Su (2020) found that perceived similarity to a blogger who wrote about the HPV vaccine was positively associated with subjective norms. Finally, in an experimental study testing the effects of descriptive norms on charity donations, Agerström et al. (2016) found that when the reference group was students at their own university rather than students in general, people donated more money.

Perceived similarity can be enhanced by mere situational circumstances, such as when people find out that they have stayed in the same hotel room (Goldstein et al., 2008), but it can also arise from sharing similar values, or from sharing objective characteristics such as gender or age (Hoffner & Buchanan, 2005). Apart from a general effect of norms, we therefore also expect that young people are particularly influenced by descriptive norms when the reference group consists of their own age group.

The present study

After more than a year into the COVID-19 pandemic, it is clear that age is the most important risk factor for becoming severely ill from the coronavirus. A recent study by Fridman et al. (2021) found that individuals with less favorable attitudes toward a COVID-19 vaccination also perceived the virus to be less threatening. As young adults are generally not at risk of becoming severely ill or dying from the virus, they may be less motivated to take the vaccine. Nevertheless, their willingness to get vaccinated affects the whole community: As the virus spreads among young people, it continues to burden health care systems and entire societies. If people are hesitant to take a vaccine, they may unknowingly be influenced by how others are behaving. Moreover, this tendency is theoretically expected, but empirically unverified, to be particularly prominent when these “others” are a group of similar others, such as one’s own age group. We therefore sought to examine the effects of conveying descriptive social norms on young people’s intentions and hesitations to get vaccinated against the coronavirus, hypothesizing that:

H1. The stronger the social norm, the stronger the intentions to take the vaccine, and the weaker the hesitancy.

H2. Young people will be especially influenced by norms when the norm reference group consists of young people rather than people in general.

H3. Signaling a strong social norm (that 85% plan to take the vaccine) will be more effective in increasing intentions and reducing hesitancy to take the vaccine compared to standard appeals from National Health Services and compared to a baseline condition.

H4. Signaling a weak social norm (that 45% plan to take the vaccine) will result in weaker intentions and increased hesitancy to take the vaccine compared with standard appeals from National Health Services and compared with the baseline condition.

By including a strong vs weak norm comparison, as well as comparisons with two different control groups, we thoroughly assess the power of descriptive norms in this context. First, the experimental design allows us not only to examine whether there is a difference between communicating strong relative to weak norms, but also whether there is a difference between communicating a strong norm compared to baseline levels, and whether there is a difference between communicating a weak norm compared to baseline. This is of theoretical as well as applied value as it informs us whether people are particularly affected by others being pro-vaccine, or reversely, if people are more affected by others being vaccine hesitant. Furthermore, we want to evaluate the practical usefulness of communicating norms in the context of willingness to take a new vaccine in an ongoing pandemic, which is why we not only study the effects of norm strength compared to baseline levels, but also include a “standard message” condition (which is common in norm experiments; e.g., Agerström et al., 2016; Goldstein et al., 2008). This design allows us to examine both whether communicating strong and/or weak norms produces changes in vaccine intentions compared to baseline, and further, whether norm messages produce effects beyond those of the reassuring messages that are provided on behalf of the health authorities. We consider the inclusion of both types of control condition a methodological strength, as most norm experiments have included only one of these.

We conducted the study in the UK, which has suffered greatly from the pandemic. Importantly, the data were collected at a time (end of December, 2020) when the first vaccine had just started to be administered (to people aged 80 years and over, people who live or work in care homes, and health/social care workers at high risk). During this time, a survey with a nationally representative sample indicated that 86% of British adults would be likely to take the COVID-19 vaccine if offered (Opinions and Lifestyle Survey, 2021). Shortly before, it was unclear whether a new vaccine would be effective and approved. This means that the decision to take or refrain from the vaccine was imminent at the time, and that it concerned everyone.

Method

For reasons of transparency, we report how we determined the sample size, all data exclusions, all manipulations, and all measures in the study. The preregistration can be found at: https://osf.io/86qe7/?view_only=2f78e70b258340f99897228036589b1e

Participants, design, and procedure

The data were collected on Prolific. The study was introduced as examining people’s perceptions of the vaccine against the coronavirus. We aimed for a sample size of 660 cases, as this would result in about 90% power to detect small-to-moderate effects of f = 0.20 according to G*Power 3.1.9.2, even if a few participants would have to be excluded due to failed attention checks. The participants (N = 661) were screened for being 18–30 years old, residing in the UK, and being native English speakers. They were paid an hourly rate of 7.05 British pounds. After giving their informed consent, they were randomly assigned to one of six experimental conditions, with information saying that:

(1) it is now estimated that among people in general, 85% plan to take the vaccine against the coronavirus,

(2) it is now estimated that among people in general, 45% plan to take the vaccine,

(3) it is now estimated that among people who are 18– 30 years old, 85% plan to take the vaccine,

(4) it is now estimated that among people who are 18– 30 years old, 45% plan to take the vaccine,

(5) “The National Health Service (NHS) declares that the coronavirus vaccine is safe and effective and that it gives you the best protection against the coronavirus” (copied from the NHS website), and

(6) a baseline condition where nothing was mentioned about other people’s plans, and no information about the vaccine was provided.

This experimental manipulation closely resembles those used in several previous norm experiments (e.g., Agerström et al., 2016; Goldstein et al., 2008; Sinclair & Agerström, 2021). In order to encourage participants to pay attention to the manipulation, they had to wait 10 seconds before they could proceed to the next page, and they were also exposed to a reminder of the importance of reading the text thoroughly. The participants proceeded to answer vaccine intentions and vaccine hesitancy scales, followed by an attention check (“It’s important that you pay attention to this study. Please tick ‘Strongly agree’”), demographic questions, perceived purpose of the study, and a manipulation check with six response options (asking them to identify which information that had been present in the beginning of the study). Finally, all participants were debriefed about the full purpose of the study and provided contact information should they have any questions.

In the pre-registered plan, we had specified that participants who could guess the purpose of the study, or who failed to answer both the attention check and the manipulation check correctly, would be excluded from the experiment. Only one participant failed both checks and was therefore excluded. An additional six participants were excluded for reporting suspicion that the study purpose was to examine how attitudes were influenced by the information provided. In total, seven participants were thus excluded, leaving a final sample of 654 participants, of which 444 (67.9. %) were women, 205 (31.9%) were men, three individuals selected “other” and two selected the option “prefer not to say”. Age ranged from 18 to 30 (median = 25 years, M = 24.60, SD = 3.61). When asked to describe their current employment status, 59.2% selected “working”, 26.5% “student”, 11.5% “unemployed”, and 2.9% “other”.

Measures

Intentions

We measured intentions to take the coronavirus vaccine with three items (Cronbach’s α = .97): “Please indicate how willing you would be to take the vaccine against the coronavirus” (1 = Extremely unwilling – 5 = Extremely willing); “How likely is it that you will take the vaccine against the coronavirus?” (1 = Extremely unlikely – 5 = Extremely likely); and “Do you plan to take the vaccine against the coronavirus?” (1 = No, definitely not – 5 = Yes, definitely). These items were similar to those used in several studies on vaccine intentions (e.g., Xiao & Borah, 2020).

Vaccine hesitancy

The second dependent variable was a modified version of the vaccine hesitancy scale (VHS), developed by the SAGE Working Group on Vaccine Hesitancy, and validated by Shapiro et al. (2018). We modified seven of the nine items so that they referred specifically to the vaccine against coronavirus instead of referring to vaccines in general. We also adapted some items so that all referred to one’s own willingness to get vaccinated, instead of willingness to vaccinate one’s children. The scale is comprised of two subscales. Two items measured perceived risks with the vaccine (Cronbach’s α = .76): “New vaccines against the coronavirus carry more risks than older vaccines” and “I am concerned about serious adverse effects of vaccines against the coronavirus”. The other subscale measures lack of confidence in the vaccine with nine items (Cronbach’s α = .93), e.g.; “Vaccines against the coronavirus are effective” (R), “The information I receive about vaccines against the coronavirus from the vaccine program is reliable and trustworthy” (R). All vaccine hesitancy items were measured on a five-point scale (1 = Strongly disagree – 5 = Strongly agree).

Results

Descriptive statistics

In line with previous research (Shapiro et al., 2018), there was greater endorsement of the risks subscale compared to the lack of confidence subscale. Intentions to take the coronavirus vaccine was generally high. Descriptive statistics are displayed in Table 1. As expected, the two subscales of the Vaccine Hesitancy Scale correlated strongly (r = .65, p < .001). The intentions scale also correlated strongly with both the risks (−.66, p < .001) and with the lack of confidence (r = −.87, p < .001) subscales.

Table 1. Descriptive statistics for the dependent variables.

	Intentions	VHS: Risks	VHS: Lack of confidence
Strong norm
Reference group: Young people	3.96 (1.23)	2.80 (1.09)	2.10 (1.00)
Reference group: People in general	4.20 (1.13)	2.56 (1.11)	1.93 (0.85)
Weak norm
Reference group: Young people	3.84 (1.24)	2.85 (1.06)	2.21 (0.96)
Reference group: People in general	3.82 (1.25)	2.72 (1.05)	2.17 (0.94)
NHS information	4.03 (1.25)	2.59 (1.10)	2.08 (0.98)
Baseline condition	3.86 (1.29)	2.68 (1.00)	2.17 (0.94)
Total sample	3.95 (1.23)	2.70 (1.07)	2.11 (.95)

The table shows means and standard deviations for the dependent variables Intentions and Vaccine Hesitancy (VHS; subscales Risks and Lack of confidence). Higher Intention scores indicate higher intentions to take the coronavirus vaccine, and higher hesitancy scores indicate greater vaccine hesitancy. All were measured on five-point scales.

Hypothesis testing

We had pre-registered that we would set alpha to .01 instead of the conventional .05, to adjust for the fact that the dependent variables were expected to correlate with each other. To test H1 and H2, we conducted three 2 (norm strength; strong vs weak) x 2 (reference group; young people vs people in general) between groups ANOVAs, i.e., one for each dependent variable. In relation to H1 and intentions to take the vaccine, we found a weak effect of norm strength, F(1, 430) = 4.51, p = .03, η_p² = . 01. In line with our expectations, participants reported slightly stronger intentions to take the vaccine when they learn that 85% (vs. 45%) of others plan to take the vaccine. We also hypothesized that the norm effect would be stronger when the reference group consist of young people vs. people in general (H2). There was no support for this hypothesis, as the norm strength by reference group interaction effect was non-significant, F(1, 430) = 1.21, p = .27, η_p² = .003.

As with the main effect on vaccine intentions, norm strength had a weak effect on the lack of confidence subscale, F(1, 430) = 3.89, p = .049, η_p² = .009, yielding weak support for H1. Also, there was no norm strength by reference group interaction effect, F(1, 430) = .45, p = .50, η_p² = .001, lending no support for H2.

Finally, there was no significant effect of norm strength on the risks subscale, F(1, 430) = 1.09, p = .30, η_p² = .003, nor a norm strength by reference group interaction effect, F(1, 430) = .27, p = .61, η_p² = .001. Thus, neither H1 nor H2 were supported with respect to perceived risks.

We proceeded with simple contrasts to test our remaining hypotheses, which concern comparing strong (H3) and weak (H4) norms with standard appeals from the NHS and the baseline/no information condition, respectively. The difference between strong norms and NHS message was not significant with respect to vaccine intentions, contrast estimate = .05, SE = .14, p = .72, 99% CI [−.32, .42]. Nor was the difference between strong norms and baseline significant, even though it was slightly larger, contrast estimate = .22, SE = .15, p = .14, 99% CI [−.16, .59]. Similarly, for lack of confidence, the difference between strong norms and NHS message was not significant, contrast estimate = −.06, SE = .11, p = .58, 99% CI [−.35, .22], while the difference between strong norms and baseline was larger yet also non-significant, contrast estimate = −.16, SE = .11, p = .16, 99% CI [−.44, .13]. In other words, conveying a strong norm did not produce clear effects beyond that of a standard authority message, rendering no support for H3.

Regarding H4, the difference between weak norms and the NHS message was not significant with respect to vaccine intentions, contrast estimate = −.20, SE = .14, p = .16, 99% CI [−.57, .17]. Nor was there a difference between weak norms and baseline, contrast estimate = −.03, SE = .15, p = .82, 99% CI [−.41, .34]. For lack of confidence, the difference between weak norms and the NHS message was non-significant, contrast estimate = .12, SE = .11, p = .29, 99% CI [−.17, .40], as was the difference between weak norms and baseline, contrast estimate = .02, SE = .11, p = .83, 99% CI [−.26, .31]. H4 was therefore not supported.

In sum, we found weak support for the hypothesis that people exposed to social norms conveying that most other people plan to take (vs. not take) the vaccine would report greater (lower) willingness to take the vaccine, and more (less) confidence in it. The effects were significant at the conventional alpha level, but not at the stricter (p < .01) level. More importantly, the effect sizes were very small. Moreover, we found no apparent norm effects on perceived risks. Finally, signaling norms does not appear to be more effective than simply providing standard NHS information.

Discussion

While older individuals have a more direct reason to take a vaccine against COVID-19 (i.e., protecting their own health), young people have less to lose from refraining from such a vaccine, and it is thus particularly important to find ways to influence their attitudes. There is a growing interest among global health scholars and practitioners in understanding how the socio-cultural context influences people’s health-related behavior, including vaccination (Cislaghi & Heise, 2018). The present study adds to this literature by examining effects of signaling descriptive norms on COVID-19 vaccination intentions and hesitation. Previous research on social influence in the context of vaccine attitudes has focused on the HPV vaccine (Xiao & Borah, 2020). In comparison, the decision of whether to get vaccinated against HPV may be perceived as less apparent and urgent, while the decision making concerning COVID-19 vaccines are situated in an ongoing pandemic that affects everyone’s lives, making it particularly interesting to explore the power of social norms on vaccine attitudes in this context.

The hypothesis that signaling strong as opposed to weak pro-vaccination norms leads to greater willingness to take the vaccine was only weakly supported. Our findings suggest that young people who are being led to believe that more as opposed to fewer people plan to take a new vaccine might form slightly more favorable attitudes toward the vaccine. Indeed, communicating accurate information about what others in one’s group or community do and approve of (“correcting misperceptions”) sometimes succeeds in changing perceived norms, which in turn may influence attitudes (Morris et al., 2015), especially for individuals with low expectations and those who were previously unaware of how common the behavior in question is (Legros & Cislaghi, 2020). Vaccination is a relatively private and less visible behavior, compared to other protective behaviors during the pandemic, such as physical distancing and wearing face masks (Rimal & Storey, 2020). It is thus possible that people’s awareness of whether vaccine hesitancy is common or not, is quite low. Although speculative at this point, it seems plausible that for those who (erroneously) believe that few people intend to take the vaccine, and for those who are unaware of how common pro-vaccine attitudes are, correcting misperceptions could be worthwhile.

However, the norm effects were weak, and did not differ significantly from the health authorities’ message or from the control (baseline) group. Therefore, we cannot conclude that signaling a strong norm – that most people want to take a certain vaccine – has an effect beyond that of standard appeals from the authorities. Based on these findings, it would be premature to recommend official appeals to invest resources in communicating norm information at the expense of other types of promising messages, such as reassuring messages that focus on vaccine safety. However, multiple messages sometimes produce stronger effects on health-related intentions (Ratcliff et al., 2019), and perhaps combining messages from health authorities with norm messages would produce stronger effects on intentions and hesitation. Similarly, it is possible that combining descriptive and injunctive norms are more effective (Schultz et al., 2007); future research could explore these possibilities.

Besides lack of confidence, another aspect of vaccine hesitation that we focused on were perceived risks, on which we did not observe any apparent norm effects. It is possible that perceived risks with new vaccines are more resistant to social influence, compared to lack of confidence and intentions, although further research is needed to determine whether this is the case. Interestingly, the norm study by Xiao and Borah (2020) instead found effects on risks but not on intentions. However, direct comparisons between their and our results can be problematic, as their study focused on a different disease (HPV), and because their study was more exploratory and only reported the results while controlling for several variables in the analyses.

According to two of the most influential theories in the field of social psychology – social identity theory and social categorization theory – individuals who define themselves as group members often incorporate the ingroup’s norms into their personal identity. These norms then become standards which they evaluate their own behavior against (e.g., Sassenberg et al., 2011). We therefore expected the effects of conveying norms to be particularly strong when the reference group was young people. In the field of health communication and attitudes to vaccination, this norm proximity hypothesis has not been put to the test before. However, it did not receive support. From a social identity perspective, the prescriptive force of the descriptive norm should be strongest when the reference group is important to one’s identity, or when one has a strong desire to be accepted as (a more central) group member (Christensen et al., 2004). Perhaps the reason for not observing the expected interaction between norm strength and reference group, is that our participants did not identify strongly with their age group. Furthermore, the reference group of young adults was compared to “people in general”, and not to other age groups, and it is possible that a comparison between young versus old people would have produced the expected effect. In sum, the reference group may not have been a strong ingroup to begin with, and moreover, there was no “outgroup” to constitute a sharp contrast. Based on these findings, we do not recommend that norm messages are tailored for specific age groups.

Limitations

There are several limitations to the present study. First, intentions may not translate directly into behavior (taking the vaccine), as contextual factors may intervene to strengthen or weaken the influence of the norm (Chung & Rimal, 2016). However, health-related intentions are often strongly linked to behavior (Fishbein & Ajzen, 2010), and that the actual offer to take the vaccine was eminent at the time of data collection is promising for this study’s ecological validity. Moreover, studying intentions is probably the closest we can get to actual behavior in this context, as behavior during vaccination is difficult to study in practice.

Regarding generalizability, we consider it a strength that our sample was not restricted to students. However, all participants were British residents age 18–30, and the findings may not generalize to all age groups or cultures. In the UK, people have generally reported high willingness to get vaccinated against COVID-19, whereas this willingness varies considerably across different European countries. It is possible that norm messages work differently depending on the level of vaccine skepticism among the population in question, and perhaps susceptibility to descriptive norms is greater among more vaccine hesitant groups. It is also possible that effects of conveying social norms depend on the specific vaccine or disease (Shapiro et al., 2018). Possibly, the more skeptical or worried people are of a certain vaccine, the more room there is for norms to assert their influence.

Further, we cannot rule out that the weak effect sizes of the norm manipulation could in part be explained by an insufficient distinction between the strong and weak norm experimental conditions. However, although claiming in the weak norm condition that e.g., 10% (instead of 45%) of people plan to take the vaccine would remedy this concern, it would also introduce credibility problems, and the credibility of the experimental manipulation was a priority concern when we designed the study. Importantly, this does not pose a threat to validity but is rather a question of generalizability, as our results may not generalize to effects of very weak norms.

Regarding the experimental manipulation, we also tried to keep the deception to a minimum, which is why we did not specify where the information about what other people plan to do came from. It is possible that adding a source behind the message would make the manipulation stronger. However, similar manipulations have proven to be effective in several other norm studies examining a variety of outcomes (e.g., Agerström et al., 2016; Goldstein et al., 2008), suggesting that our manipulation should be just as powerful. That most participants could correctly identify the correct alternative in the manipulation check, coupled with the fact that very few expressed suspicion about the purpose of the study, attests to the effectiveness of the manipulation, as it suggests that the norm information was noticed, remembered, and perceived as credible.

This research focused on general norm effects, and future studies might want to examine interactions between effects of norm messages and theoretically relevant personality traits. For starters, some individuals tend to be more vulnerable to social influence than others (Briñol & Petty, 2005). Second, the decision to get vaccinated or not may be easy for some people but complicated for others. Individual difference variables that may be of importance include health literacy; the set of abilities and skills that are required to navigate the health care system and health-related messages (Bernhardt et al., 2005), and more broadly, educational level. Finally, we did not measure prior vaccination attitudes, and research suggests that those who are vaccine skeptical to begin with can react in unpredictable ways to influence attempts (Nyhan & Reifler, 2015).

Conclusive remarks

Even though vaccines prevent several diseases that used to be widespread from returning, vaccine hesitancy and resistance persists, and efforts to minimize it is an international priority (Lane et al., 2018). The present study evaluated the possibility of signaling descriptive social norms as a means of increasing young people’s willingness to take a new vaccine during a pandemic. Although the results suggest that the practical usefulness of signaling social norms is limited, we hope that they will spawn more research efforts on effective ways to signal social cues such as norms, with the ultimate goal of reducing vaccine hesitancy and increasing vaccine acceptance.

Author contributions

SS planned the experiment with help from JA. SS supervised the data collection, conducted the literature search and the statistical analysis, and drafted the manuscript. Both authors revised the manuscript and approved the final version for submission.

Acknowledgments

The authors are grateful to Towe Nilsson for assistance with data collection.

Data availability statement

The data that support the findings of this study are available from the corresponding author upon request.

Disclosure statement

The authors declare that there are no conflicts of interest.