A Randomized Experiment on the Effects of Moral Appeals on A Randomized Experiment on the Effects of Moral Appeals on U.S. Parents’ Intentions to Vaccinate Their 5e11-Year-Old Children U.S. Parents’ Intentions to Vaccinate Their 5e11-Year-Old Children Against COVID-19 Against COVID-19

Background : Democrats and Republicans in the United States were divided on their COVID-related risk perceptions and their adoption of preventive measures (e.g., getting vaccinated). Based on moral foundations theory and the matching hypothesis, this study hypothesized that parents with a Democratic af ﬁ liation would be persuaded by messages featuring a harm/care or a fairness moral appeal, whereas parents with a Republican af ﬁ liation would be persuaded by messages featuring an authority or ingroup loyalty appeal. Method : An experiment was conducted among 567 parents with children aged 5 e 11, whereby each participant was randomly assigned to read one of the four moral appeals or a control message. Each participant then completed a questionnaire. Results : The results showed that, in general, the moral appeals did not interact with parents ' political af ﬁ liations, and the moral appeal messages did not signi ﬁ cantly increase the parents ' risk perceptions or vaccine uptake intent for their children. Additional analysis showed that trust in government and future orientation were strong predictors of parents ’ risk perceptions and vaccine uptake intent, whereas COVID fatigue was a weak predictor of their message evaluation. Conclusion : Moral framing in persuasive messages may have limited effects on a health problem widely known to the public. Instead, participants ' internalized value orientations and personal differences may be more predictive of their attitudes and adoption of preventive measures.


Introduction
A fter a highly politicized pandemic and the 2020 U.S. presidential election, Democrats and Republicans in the United States have shown different levels of risk perceptions toward COVID-19 in general and mistrust toward COVID-19 vaccines in particular [1,2].Republicans, compared with Democrats, perceive lower COVID-19 risks and have higher vaccine hesitancy [1].When facing a decision to vaccinate their children, parents of different political affiliations can exhibit similar risk perceptions and vaccine uptake intent for their children.Nine months after the U.S. Food and Drug Administration's Emergency Use Authorization for use among children 5e11 years of age, statistics show that the full vaccination rates among children aged 5e11 were lower in the more Republican-affiliated states than in the more Democrat-affiliated states [3].For example, the full vaccination rates among children aged 5e11 in New York, Connecticut, and Massachusetts were 39%, 45%, and 52%, respectively.In contrast, the full vaccination rates in Indiana, Tennessee, and Alabama were 21%, 17%, and 12% [3].
Recent theorizing and opinions often regard vaccinating against COVID-19 or getting children vaccinated as a moral act [4].The use of moral appeals can help facilitate attitude and behavioral change: Different moral appeals are more effective when matched with the target audience's political affiliation [5].This randomized experiment thus explores the efficacy of using moral appeals among parents with Democratic and Republican affiliations.In addition, it aims to understand the reasons that underlie parents' vaccine uptake intent for their children.Such an investigation can help understand the efficacy or failure of the health marketing strategies in the United States.This is particularly important because the United States has been a hotbed for COVID-19 infections and deaths.Higher rates of vaccine uptake can help contain COVID-19 and protect children's and adults' health [3].
Moral foundations theory states that human morality is pluralistic and that five major moral foundations underlie human thinking and behavior [6].These moral foundations include harm/care (i.e., care for and not inflict harm on others), fairness (i.e., reciprocity, equality, and justice), ingroup loyalty (e.g., loyal to one's group), authority (e.g., respect the authority and tradition), and purity (i.e., being pure from a religious perspective).Although different groups may subscribe to all five moral foundations, they may emphasize some but not others [6].More specifically, Democrats in the United States emphasize harm and fairness, whereas Republicans subscribe to ingroup loyalty, authority, and purity [6].
The matching hypothesis states that when a (moral) value appeal in a message is matched with the value that an audience holds, the audience will assign higher importance and relevance to the message than one that contains a mismatched value appeal [7].Furthermore, the literature on moral framing [8] states that using a moral value in persuasive messaging, if matched with the value orientation among the intended target audience, can guide the audience's moral thinking and provide an interpretive context for them.Hence, it can be more persuasive when information (e.g., risk) is communicated through a moral appeal matched with parents' moral values.
Our observations show that COVID-19 messaging in the United States and the world has used various moral appeals.COVID-19 can inflict severe illness and death and disrupt one's daily routines, work, and schooling.Vaccination can help prevent children from contracting COVID-19 (i.e., harm/care) or promote a healthy country or group (i.e., ingroup loyalty).It is also fair for every parent to vaccinate their children instead of hoping for other parents to vaccinate their children to prevent community spread (i.e., fairness).Lastly, vaccination messages can be conveyed via authoritative figures or health departments, for example, Pope Francis [4] or health commissioners.Considering the harm/care moral appeals, it is parents' responsibility to care for their children and prevent them from contracting COVID-19.Based on the theorizing in the preceding paragraph, when risks or the suffering associated with COVID-19 are conveyed along with a harm/care moral value (vs.without), parents will be more likely to have more favorable message evaluation, heightened risk perceptions, and vaccine uptake intent.
However, there is a caveat to the relationships discussed above.Political affiliation can moderate the effectiveness of moral appeals.According to Graham et al. [6], liberals (e.g., Democrats) are more likely to value harm/care and fairness, whereas conservatives (e.g., Republicans) are more likely to value ingroup loyalty and authority.As such, harm/ care and fairness moral appeals will be more effective for political liberals than conservatives because they value harm/care and fairness.At the same time, ingroup loyalty and authority appeals will be more effective for political conservatives than liberals.Feinberg and Willer [5] found that after seeing a message advocating the use of English as the official language of the United States, political liberals (vs.conservatives) showed stronger support when the argument was made from a fairness perspective.In contrast, political conservatives (vs.liberals) showed stronger support when the argument was made from the authority perspective.Democrats used and were more persuaded by moral appeals related to harm/care and fairness (vs. the other three moral values), whereas Republicans used and were more persuaded by moral appeals related to ingroup loyalty, authority, and religious purity/sanctity [e.g., 5].Based on the preceding, this project will examine the following: H1.(a) Harm/care and (b) fairness moral appeals will elicit stronger perceptions of COVID risks, COVID vaccine benefits and safety, value-expressive attitudes, and intention to vaccine their children among parents with a Democratic affiliation than with a Republican affiliation.
H2. (a) Ingroup loyalty and (b) authority moral appeal will elicit stronger risk perceptions, COVID vaccine benefits, value-expressive attitudes, and intentions to vaccinate their children among parents with a Republican affiliation than with a Democratic affiliation.
Previous research has provided some preliminary evidence that several other factors can influence individuals' reception of COVID-19 messages and prevention, including COVID-19 fatigue [e.g., 9, 10], future orientation [11], and trust in government [12].Therefore, the present research will statistically control for the influence of these variables using regression analysis and further explore the relationships between them and message evaluation, risk perceptions, and parents' vaccine uptake intent for their 5-11-year-old children.

Procedure and sample
A randomized online experiment was conducted to examine the effectiveness of moral appeals (harm/ care, fairness, ingroup loyalty, authority, and control).G*Power was used to determine the sample size: We set alpha to .05 and power to .80.The number of experimental groups was five, and a small-medium effect size f ¼ 0.17 was used.The required sample size was 418.Alternatively, a sample of 567 could detect an effect size f ¼ 0.15.Participants were randomly assigned by Qualtrics' online survey algorithm to view one of the five moral appeal messages and then responded to several batteries of questions.There were approximately equal numbers of participants in each condition.
The final sample (N ¼ 567) included a Qualtrics sample of 496 participants and an MTurk sample of 71 participants in November and December 2021.For the Qualtrics sample, Qualtrics coordinated data collection, including setting a quota of racial composition to mirror that of the United States, sending the survey link to randomly sampled panel members, and screening those to meet the criteria for participation (i.e., parents with children ages 5e11).A total of 1318 Qualtrics panel members first answered screening questions related to their parental status and the age of their children.Those who had children aged 5e11 continued the experiment and read an informed consent form.Due to Qualtrics' difficulty recruiting male parents, additional data were collected via Amazon Mechanical Turk, resulting in a sample of 71 male parents.Independent-samples t tests showed no differences in the dependent variables between the male participants from the Qualtrics and MTurk samples.The final sample included 68.3% female and 31.7%male parents.Four non-binary parents were not included in the final analyses because the number of nonbinary participants was small and did not allow statistical comparison.The racial makeup was as follows: 5.5% Asians, 13.4% Blacks, 10.4% Latinos, 3.2% Native Americans, 64.6% White, and 3.0% other racial backgrounds.The average household income was US$64,321 (SD ¼ 51,432), the average number of years of education was 12.8 (SD ¼ 4.45), and the mean of political philosophy was 4.27

Ethical issue
The procedure and the questionnaire received ethical approval (HSRO #05022321) for the procedure and the questionnaire from Rochester Institute of Technology University in the United States.

Experimental manipulation
We created and pretested five health messages (Table 1) after reviewing the COVID-19 vaccination messages in the United States.The main structure and the message in each condition were kept consistent across the conditions; the only difference was that the four experimental conditions included an additional sentence that emphasized one of the four moral foundations (i.e., harm/care, fairness, ingroup loyalty, and authority).These messages were pretested and rated by a group of 44 undergraduate students in the United States.The ratings showed that each moral message was rated higher on their respective moral foundation than the control message ( ps < .05;Table 1).The messages with the intended moral appeals received higher ratings on the intended moral dimension and lower ratings on the other four dimensions.
Items were adapted from or constructed based on the definitions in the literature [13e18].Scale items, associated information, and references were listed in Table 2.All the scale items loaded on their respective factors based on confirmatory factor analysis and demonstrated good construct validity and reliabilities.

Main analysis
The means and standard deviations of the dependent variables are shown in Table 3. ANOVA tests showed that none of the comparisons were significant.
To test H1 and H2, this present research examined the interaction effects between political philosophy and the experimental conditions using moderated multiple regression because political affiliation was measured as an interval-level variable.Each of the four experimental conditions was coded as 1 for that specific condition, which created four variables.The control condition was coded as 0 for all four variables and was used as "contrast."Four interaction terms between mean-centered political affiliation and each of the specific experimental conditions were created.Then, the mean-centered political philosophy was entered in the first step, the experimental conditions in the second, and the interaction terms in the third.
The results (Table 4) showed that for message evaluation and perceived susceptibility, the interactions between the experimental conditions and political philosophy were nonsignificant.The experimental conditions did not influence participants' message evaluation or perceived susceptibility.
For value-expressive attitudes, the interaction effects were not significant.The main effect of the fairness condition showed that the fairness condition increased the participants' value-expressive attitudes (B ¼ 0.32, p ¼ .040).Other main effects of the experimental conditions were not significant.
For parents' vaccine uptake decisions for their children, the interaction effect between political affiliation and the care manipulation was significant (B ¼ 0.20, p ¼ .044).Further analysis showed that for those in the care condition, the relationship between political affiliation and vaccine uptake intent was B ¼ 0.35, p < .001.For those in the control condition, the relationship between the two variables was 0.15, p ¼ .022.
Taken together, the results provided limited support for the theorizing on the matching effects of moral appeals to one's political affiliations.Overall, the block of experimental conditions and the interaction effects added 0.0%e0.8% of the variance to the four dependent variables and were not significant.

Additional analysis
The additional analysis examined the role of demographic, personality, and value-based variables without including experimental conditions.The inclusion of experimental conditions did not change the magnitudes of the reported relationships.Table 5 showed that females were less likely to evaluate the messages favorably (B ¼ -0.29, p < .001),held less strong value-expressive attitudes (B ¼ -0.19, p ¼ .042),and were less likely to a All experimental manipulations were followed by " …, it is more important now to get children 5e11 years of age (or "them" in the control condition) vaccinated against COVID-19.With the emergence of the Delta variant in recent months, more children have tested positive relative to older adults.This is largely because more older Americans are vaccinated and children have returned to in-person schooling.Pfizer's COVID-19 vaccine has been tested and proven safe and effective among young children 5e11 years old.It has been approved by the FDA.Get your child vaccinated against COVID-19 when one becomes available in your area." b Pair-samples t tests showed that the differences between the intended moral appeal manipulation and the control condition (means in the same row sharing superscript b) were significant ( p < .05).
COVID-19 fatigue was only negatively related to message evaluation (B ¼ À0.08, p ¼ .019)and was not related to perceived susceptibility, valueexpressive attitudes, and vaccine uptake intent.
Future orientation showed a positive relationship with message evaluation (B ¼ 0.18, p < .001)and perceived susceptibility (B ¼ 0.29, p < .001).Trust in government showed a positive relationship with message evaluation (B ¼ 0.51, p < .001),perceived

Discussion
Table 4 showed that the effects of message appeals were limited.By the procedure of randomization, we removed the initial differences among the participants; that is, the participants were assumed to be equal across the conditions.As such, the limited effects on the dependent variables should be attributed to the ineffectiveness of the experimental manipulations.Table 5 showed that among the demographic and other individual variables, future orientation and trust in the government were stronger predictors of the dependent variables than were demographic variables and COVID-19 fatigue.The results have theoretical implications for using moral appeals and practical implications for COVID-19 vaccination programs and future pandemics.
First, our results, as well as previous research [19], indicate the need for reconceptualizing the use of moral foundations theory and the matching hypothesis in persuasive messaging.First, according to the current theorizing, care and fairness moral appeals should be more persuasive for those with a Democrat affiliation than for those with a Republican affiliation in the United States, whereas authority and ingroup moral appeals should be more persuasive for Republicans than for Democrats.Table 4 showed nonsignificant interaction effects between these moral appeals and political affiliation among the four dependent variables for all the moral appeals, except for one significant interaction effect between care and political affiliation on vaccine uptake intent.Similarly, Arpan et al. did not find a matching effect in using realistic environmental communication in facilitating favorable message evaluation or willingness to pay more for renewable energy [19].On the other hand, Feinberg and Willer [5] found a matching effect among hypothetical scenarios (using English as the official language).Arpan et al. stated that it could be difficult to reframe renewable energy messaging [19].
Second, our manipulation checks showed that the message manipulations were successful.However, all the messages were rated high on the care dimension because the ultimate moral in the vaccination messages is about saving lives, whether Note.N ¼ 567.Each of the four experimental conditions was coded as 1 for that specific condition, which created four variables.The control condition was coded as 0 for all four variables and was used as "contrast."þ p < .10,*p < .05,***p < .001.done via an ingroup appeal, an authority appeal, or a fairness appeal.We concur with Arpan et al.'s overall thesis [19] and believe that the matching effect may be unsuccessful if the ultimate moral of an issue or the previous, consistent messaging trumps the reframing efforts.That is, the moral framing/matching effects may be more pronounced when an issue is novel or does not have a dominant moral foundation (e.g., the care foundation for medicine-or life-and-death-related issues).Thirdly, additional analysis (Table 5) revealed that future orientation and trust in government, compared to the demographic variables and other individual difference variables (e.g., COVID-19 fatigue), were much more important.These variables explained a large proportion of the variance in message evaluation, risk perceptions, and vaccine uptake intent.The results indicate that vaccination programs' success can largely depend on the populace's pre-existing individual or cultural differences.That would have explained why COVID-19 promotions were more successful and encountered less resistance in some states (vs.others) in the United States and some countries than in others.
A few limitations should be acknowledged.The first limitation is associated with the use of an online sample.Internet panels are known for their opt-in characteristic and may not represent the general population or parents.Although the racial makeup in the sample was generally consistent with the national statistics, the sample included more female participants than male participants, probably because the topic and vaccination decisions are more relevant to mothers than fathers.Second, participants read only one message in one sitting, which might have heightened their sensitivity to the issue.On the other hand, participants recruited from the Internet may skip a message before completing the ensuing questionnaire.Thirdly, we point out one caveat that our experiment was based on one-time forced exposure.Repeated exposure to the messages in the media or elsewhere might produce a different effect.

Conclusion
The present research found that the matching effects of moral appeals and political affiliation were limited, probably due to the underlying consideration that all the messages were related to care and harm.As such, reframing the message using other moral appeals may not elicit more or less favorable responses.That is, these moral appeal messages were equally effective in facilitating message evaluation, risk perception, value attitudes, and vaccine uptake intent.Our study is of theoretical importance because it helps cumulate knowledge on the conditions when moral appeals may not fare better than a generic health message.We also found that internalized values were strong predictors of message evaluations, risk perceptions, value-expressive attitudes, and vaccine uptake intent for children.

Recommendations
With the absence of evidence to refute the use of different moral appeals, we posit that it is safe to continue to use these moral appeals and that they most likely will not result in lower persuasiveness than a health message without an explicit moral appeal.We recommend that future research further explores the role of moral appeals and political philosophy in other public health settings and be cognizant of the role of internalized values when designing health campaigns.

Table 1 .
Message manipulation and manipulation checks.

Table 2 .
Confirmatory factor analysis and standardized factor loadings of scale items of the variables.Note.N ¼ 567.Confirmatory factor analysis showed a good fit to the data: c 2 (329, N ¼ 567) ¼ 797.7, p < .001,root mean square error of approximation (RMSEA) ¼ 0.050, 90% CI of RMSEA [0.046 ~0.054], comparative fit index ¼ 0.96, and standardized root mean residual ¼ 0.059.Items were adapted from the literature and edited for language and context.

Table 3 .
Descriptive statistics of means and standard deviations of message conditions on the dependent variables.

Table 4 .
Moderated multiple regression analysis of moral appeals and political affiliation on parents' message evaluation, COVID-19-related beliefs, and vaccine uptake intent.

Table 5 .
Additional analysis of the demographic and value-related predictors of COVID-19 vaccination message evaluation, beliefs, and vaccine uptake intent for children.