Study 3 sought to extend our analysis in two ways. First, it utilized an experimental design to provide the strongest test of the causal role of ingroup trust on COVID-19 risk. Second, it utilized validated multi-item measures of trust and risk behavior. Third, it included a control condition, with a view to investigating the inferences people make about trust and risk in the absence of explicit information about group membership. Fourth and finally, Study 3 aimed to provide an initial test of how the link between ingroup trust and COVID risk might be fruitfully used to inform the public health response. We reasoned that one of the greatest challenges of behavior change still ahead in combating the pandemic is persuading the global population to accept a COVID-19 vaccine. All vaccines have some (usually extremely small) risk associated with them, and these potential risks are often amplified and given undue emphasis by vaccine sceptics [27, 28]. In line with our theoretical model, we reasoned that people will perceive a vaccine as less risky, and indicate a greater willingness to take the vaccine, if it has been developed and endorsed by ingroup (rather than outgroup) members because of the greater trust we have for them.
Method
Participants were 216 Australian residents recruited for an online experiment via Prolific, an academic research web platform, between 13 and 17 August 2020 – when hundreds of vaccines were in development, but results of phase 3 trials were not yet available. Participants were 108 men and 104 women, with four people indicating that they were non-binary or choosing not to disclose their gender. Participants ranged in age from 18 to 72 (M = 30.67; SD = 10.55) and were diverse in terms of their educational attainment. Participants were paid the standard Prolific rate for their time (approximately AUD$8.30/h).
Study 3 was an experiment with three conditions: ingroup source, outgroup source, and control. Participants were asked to consider a hypothetical scenario in which a COVID-19 vaccine was announced in a press conference by Australia’s medical officials. Participants first completed a social identification scale to make their social identity as an Australian salient.
Participants were then asked to imagine that the following scenario in a speech attributed to Australia’s chief medical officer. The control condition excluded the text in square brackets and did not attribute the vaccine or the risk evaluation to any particular group.
“I am pleased to announce that a first vaccine for COVID-19 will, from tomorrow, be available to the Australian public.
The vaccine was developed by [Australian/French] scientists and has passed the final stage of testing. It showed an 85% success rate in preventing the virus and carries only a 2% risk of serious side effects, which [our/France’s] leading health experts consider an acceptable level of risk.
Of course, this vaccine is being made available around the world, including in Australia. A website has just gone online where everyone can sign up to receive the vaccination, and indicate the doctor’s surgery where they would like to receive it.
This will be the first stage in rolling out the vaccine to all Australians.”
French people were selected as the outgroup for several reasons. We sought a real-world outgroup which was perceived by Australians to be of similar status (e.g., unlike the United States which is typically perceived as higher status) and to be a nation about which Australians tend to have relatively neutral stereotypes (e.g., unlike Canada, which is typically perceived very positively).
The design, measures, hypotheses and analyses were pre-registered (https://aspredicted.org/h8qj9.pdf). The pre-registration also included a power analysis, which indicated a recruitment target of 250 to achieve a minimum sample size of 207 for analyses.
Measures
Trust
In Study 3, trust was measured using a comprehensive validated scale adapted from previous research [29]. As well as some items related to general trust, this measure included subscales relating to perceived integrity (e.g., “Sound principles seem to guide [Australian/French] scientists’ behavior”), benevolence (e.g., “[Australian/French] scientists would not knowingly do anything to hurt me”) and competence (e.g., “I feel very confident about [Australian/French] scientists’ skills”), each rated on a scale from disagree strongly [1] to agree strongly [5]. The underlined content was removed in the control condition. Two of the items from the original scale could not be readily adapted to the experimental context of Study 3, yielding a 19-item scale [7]. The scale has primarily been used as a single unitary indicator of trust in prior research [30] and the overall reliability (α = .91) suggested this was also appropriate here.
Perceived COVID-19 vaccine risk
Perceived risk of the proposed vaccine and willingness to take the vaccine was measured using four items adapted from prior research [7], each measured on a continuous sliding scale from not at all [0] to extremely [100]. The items were “How risky do you think receiving the vaccine is?”, “How safe do you think having the vaccine is?”, “How risky do you think not having the vaccine is?”, and “How likely is it that you will sign up for the vaccine?”. Three items were reversed such that higher scores represented a greater perceived risk (and avoidance) of the hypothetical vaccine (α = .86). In addition to the Cronbach’s alpha, the unitary structure of these items was further supported by an exploratory factor analysis, which revealed only one factor with an eigenvalue > 1 that accounted for 70% the variance, with each item loading onto this factor at >.72.
Manipulation check
At the end of the study, participants were asked “In the scenario you were asked to imagine, which scientists had developed the vaccine?” with response options of “Australian”, “French”, “American”, “I can’t remember” and “It didn’t say which scientists developed the vaccine”. Participants in the ingroup and outgroup conditions were required to accurately identify the nationality of the scientists in their vignette to pass the manipulation check, while either of the final two options were acceptable to pass the manipulation check in the control condition.
Results
Initially, 250 people completed the study, however, in accordance with our pre-registered data management plan, 34 who failed the manipulation check were excluded. This left 216 participants in the final sample.
The hypothesis was tested using PROCESS [24] with 5000 bootstrapped samples. Experimental condition was entered as the (categorical) independent variable, trust in scientists was the mediator, and perceived vaccine risk was the dependent variable. All combinations of contrast coding were used in the mediation analysis such that comparisons were examined between each pair of conditions. Of these, only those comparing the ingroup source to the outgroup source were significant. This model is presented in Fig. 3.
Participants in the ingroup condition expressed greater trust in the vaccine scientists, β = .53, p < .001. Trust in vaccine scientists, in turn, predicted a reduction in perceived risk and avoidance of the COVID-19 vaccine, β = −.23, p = .005. Supporting the hypothesis, the indirect effect was significant, β = −12. (95% CI: −.27, −.02).
Comparisons between the control condition and each of the group membership conditions revealed no significant effects except the effect of trust on risk, which was robust in all conditions. However, inspection of the estimated marginal means (presented in Fig. 4) indicated that, rather than falling between the ingroup and outgroup conditions, the control condition most closely resembled the ingroup condition, albeit with a slightly wider variance that rendered its comparison with the outgroup condition nonsignificant.
Discussion
Study 3 provided several new insights. It replicated the previous studies in finding that ingroup trust is a key predictor of COVID-19 risk. However, Study 3 focused on a different domain of risk: perceived riskiness of and willingness to receive a COVID-19 vaccine. This is important because it suggests ways in which the general tendency towards ingroup trust might be utilized for effective public health messaging – by emphasizing that vaccine development is an ingroup initiative being spearheaded by ingroup members. Of course, for many people around the world, the scientists developing COVID-19 vaccines will be outgroup members (based on nationality). However, shared identity can nevertheless be emphasized on dimensions other than nationality, potentially encompassing all of humanity [31, 32].
Study 3 had several strengths over the prior studies, including its pre-registered experimental design and its use of more comprehensive validated measures of the constructs of interest. However, a limitation of Study 3 was that the findings may have been influenced by the specific ingroup and outgroup selected (Australia and France), and the real-world intergroup dynamics of these groups driven by normative content and status. The effects may have been quite different if, for example, a high status country (e.g., the United States) had been selected as the outgroup. Findings should thus be generalized with caution, and with consideration of these dynamics. This limitation is best addressed in research by employing a minimal group paradigm, where the groups in question have no real-world “baggage” associated with them. Although this was not possible in the context of the present study, previous research has shown that the effects proposed in SIMORT also emerge in the context of such paradigms [7]. The comparison to the control condition provided some tentative evidence that in the absence of group membership information, participants may infer a shared group membership – at least in the context of a local vaccine roll-out or similar public health initiative. This is promising for intervention efforts, because it suggests that in countries where vaccine developers are an outgroup, de-emphasizing the nation of origin may reduce vaccine hesitancy.