The efficacy of health experts’ communication in inducing support for COVID-19 measures and effect on trustworthiness: A survey in Hong Kong

Abstract

To control the spread of COVID-19, governments may implement freedoms-infringing health measures. Therefore, citizens' support for these measures is important. This study investigates: (1) whether health experts' communication induces support for COVID-19 measures, and (2) whether health experts' agreeing or disagreeing with government directives affects their trustworthiness. A cross-sectional online questionnaire was completed by 1072 adults in the Hong Kong general population between May 26 and June 3, 2021. Three COVID-19 measures were examined: contact-tracing mobile application, restriction-testing, and ban of public assembly. For each, participants were randomly assigned to three groups to view, respectively: vignettes with a neutral government announcement only; vignettes with a government announcement and a health expert's quote supporting the government's decision; and vignettes with a government announcement and a health expert's quote disagreeing with the government's decision. The result shows that positive health experts' communication increased the support for banning public assembly; no effects were found for the support for contact-tracing mobile applications and restriction testing. Participants who only viewed health experts disagreeing with the government had higher trust in health experts relative to participants who viewed health experts agreeing with the government at least once. The results render doubtful the strategy that health experts can be involved for garnering support for unpopular health measures without jeopardizing public trust in them.

1. Introduction

The COVID-19 pandemic, which started in 2020, continues to impact human life worldwide. Based on the recommendations of health experts, governments have imposed various public health measures such as social distancing, mandatory mask-wearing, and lockdown. Public support for government-imposed health measures is important to control the spread of contagious diseases. To increase compliance, health experts have assumed important roles—either invited to do so by their respective countries' governments, media, or out of personal civil responsibility—to communicate to the public the necessity of following COVID-19 restrictions. In general, scientists enjoy fairly high levels of public trust worldwide (Gallup, 2019; Pew Research Center, 2019). Research has revealed a “rallying effect” that boosted support for scientists and expertise during the pandemic (Battiston et al., 2020; Daniele et al., 2020). As an epistemological authority and a credible source of information, health experts are expected to command significant cooperation from society, but evidence in this regard is mixed at best and therefore warrant further research. Arceneaux et al. (2020) found evidence of greater support for COVID-19 restrictive measures endorsed by health experts as compared to measures endorsed by lawmakers only; Deslatte (2020) found less influence of health experts' endorsement than policymakers. Alsan and Eichmeyer (2021) found laypeople to be more effective than experts at promoting vaccination among low socio-economic classes. Does health experts' public communication have an effect on increasing support for governments’ COVID-19 measures? This is the first research question of this study.

In public decision-making, policymakers sometimes delegate strategies and decisions to experts. Alternatively, experts serve as consultants in committees to provide suggestions and advices, but ultimately the policymakers bear the responsibility of decision-making. In the latter case, directives of policymakers and opinions of health experts could diverge (Antoci et al., 2020). There are three reasons. First, while experts tend to make issue-specific recommendations based on their own expertise, policymakers have to consider diverse interests of the society (Moore and MacKenzie, 2020). For example, with the single goal of pandemic control, health experts can recommend stringent measures, but policymakers must consider economic, social, administrative, and fiscal aspects. Second, while experts presumably make recommendations in good faith, policymaking often involves political calculations. For example, Pulejo and Querubín (2020) found that incumbents who had to run for re-election implemented less stringent COVID-19 restrictions when elections were near. Third, opinions among health experts themselves could diverge. Making recommendations regarding COVID-19 is far from finding the best technical solution (Lavazza and Farina, 2020). Recommendations require application of scientific knowledge and making assumptions of the real world in case of uncertainty. Divergence also emanates from different values of individual experts and societies, such as the importance of freedom and priority of the right to live. Therefore, even though health experts were enlisted by governments to tackle the pandemic, diverse views among expert groups resulted in the co-existence of affirming and doubting opinions on government directives. This study focuses on whether health experts' agreements and disagreements with government directives affects experts’ trustworthiness—this is the second research question.

Experts’ trustworthiness was of interest because studies have reported lower trust in science and scientists after epidemics (Eichengreen et al., 2021; Feufel et al., 2010). In recent years, lower trust in experts has been observed in issues such as climate change (Anderegg et al., 2010) and Brexit (Bauer, 2017; Clarke and Newman, 2017) and become a subject of research interest (Gauchat, 2012) and public interest. The result also has implications on future risk management as the field literature suggests correlation between expert trust and compliance (Gilles et al., 2011; Siegrist and Zingg, 2014).

Many studies have investigated the factors associated with the acceptance of various COVID-19 measures that restrict freedom (Arceneaux et al., 2020; Bargain & Aminjonov, 2020; Guglielmi et al., 2020; Guillon and Kergall, 2020; Kuiper et al., 2020; Yu et al., 2020). It was believed that people living in democratic countries have lower tolerance for restrictive measures than those in authoritarian countries. In the context of Hong Kong, a major social movement had broken out in 2019 and made citizens sensitive about measures that restricted personal freedom. At the same time, trustworthiness of the government dropped to an all-time low in February 2020 when the pandemic began (Hong Kong Public Opinion Research Institute, 2022b). The existing tension made it difficult to support the government's freedom-infringing COVID-19 measures. Under such circumstance, health experts' recommendations—although based entirely on scientific grounds—might have been viewed with political skepticism.

The three COVID-19 measures tested in this study were the use of contact-tracing app, restriction-testing, and ban of public assembly. A contact-tracing app, “Leave Home Safe” was developed by the Hong Kong government for scanning and storing information of entry into premises. The use of this app was not mandatory but encouraged at the time of the survey conducted as part of this study. According to another survey, the use of digital contact-tracing was the most disagreed—approximately 60%—among the four measures surveyed in Hong Kong participants (Voo et al., 2021). Hong Kong people showed more resistance against digital contact-tracing than their counterparts in Malaysia, Singapore, the US, and South Korea (Huang et al., 2021; Voo et al., 2021).

Another controversial measure was the ban of public assembly. Under the Prevention and Control of Disease Regulation (Cap. 599 F) enacted because of the COVID-19 emergency, public gatherings of more than four people were prohibited during the survey period. Offenders could face a maximum fine of $25,000 and/or imprisonment for up to six months (Government of Hong Kong, 2021a). In Hong Kong, which has built its reputation as a “city of protest” (Dapiran, 2017) and “China's rebel city,” (South China Morning Post, 2020) public demonstrations and protests were banned because of the pandemic, which was in stark contrast to the fierce social movement of 2019. This measure was criticized by activists on grounds of “prevent (ing) the return of anti-government demonstrations, while largely allowing other groups to gather with impunity.” (Time, 2020).

“Restriction-testing” refers to a short-term lockdown of a small area consisting of a few buildings or a housing estate, and lasts from a half-day to about a week. During such lockdowns, residents within the “restricted areas” were mandated to undergo polymerase chain reaction testing of virus and not allowed to leave the areas. Between January 2021 and March 2022, restriction-testing operations were imposed approximately 250 times, affecting 350,000 people, or approximately 5% of the population (Ming Pao, 2022). Restriction-testing operations were akin a lockdown, albeit at a small scale. It was not a common COVID-19 measure implemented in other countries and to the best of the author's knowledge, no study has examined support for such a measure.

2. Methods

2.1. Study design

An online survey was conducted by a survey company (Dynata) from May 26 to June 3, 2021 using quota sampling to mimic the general Hong Kong adult population by age and sex. Points to exchange for coupons were given by the survey company as an incentive. A total of 1105 samples were collected anonymously. After eliminating incomplete responses and machine-detected ballot-stuffing using built-in functions of Qualtrics, a survey management platform, 1072 valid samples were analyzed. Electronic consent was obtained from participants before the survey began. Participants could opt out any time during the survey. The collected data were retrieved from the online survey platform protected by passwords. Ethics approval was obtained from the corresponding author's affiliated institution.

Daily new confirmed COVID-19 cases ranged from zero to seven during the survey period with no new death case (Government of Hong Kong, 2022). The epidemic was stable at a low level. However, the pandemic control measure had been held at a high level despite the drop in number of cases (Fig. 1 ). Following the strict pandemic control policy of China, the stringency index of Hong Kong stood at 71, whereas the world average was 58 (Hale et al., 2021). Social distancing and mask-wearing requirements continued; vaccination requirements and limits on capacity and operating hours were in place for catering business and various premises. The government explained that it was to avoid a rebound (Government of Hong Kong, 2021b). According to Hong Kong Public Opinion Research Institute (2021)'s representative sample collected during the survey period, respondents estimated the risk of infection to be 12%; 54% of the respondents were dissatisfied with the government; an average of 94% respondents thought that the social gathering ban was too strict.

Fig. 1.

Daily new confirmed COVID-19 cases and stringency index in Hong Kong (2020–2021).

Source: Government of Hong Kong (2022); Hale et al. (2021).

Fig. 2 shows the flow of the survey. Part 1 asked participants about their COVID-19 experiences and political attitudes. Part 2 presented vignettes (excerpts from news reports) of three different COVID-19 measures in a random order: the use of contact-tracing mobile application, restriction-testing, and the ban of public assembly. For each response, a participant viewed one vignette, which was randomly drawn from consent, dissent, or control groups. This implies that the number of respondents in each group is roughly one-third of the total sample, barring any variations in randomization. The consent group referred to health experts' agreement with the government measure. The dissent group referred to disagreement. The control group did not contain health experts’ communication. Immediately after viewing the vignette, the participant was asked for their extent of support for the measure stated in the vignette. After viewing all three vignettes, participants were asked about their trust level in the government and health experts. Finally, demographic information was collected. The opinion, preference, and attitude questions were asked before the vignettes because they were used as control variables in this study. Viewing the few vignettes might affect what respondents think about the seriousness of the epidemic, and hence affects, for example, how they weight between pandemic control and freedom; and whether they think the Hong Kong government is competent. Therefore, these questions were ordered before the vignettes. The variables aimed to be treated only included trustworthiness and compliance. Demographics were fixed for the respondents and would not be affected by the treatment, and therefore could be asked in rear.

Fig. 2.

Flow of the survey.

Three broad categories of control variables were used: experience with COVID-19, political attitudes, and demographics. The controls for COVID-19 experience follow the crisis and trust literature and are relatively standard (Aassve et al., 2020; Aksoy et al., 2020; Albertson and Gadarian, 2015; Arceneaux et al., 2020; Citrin and Stoker, 2018; Gill, 2007; Goldfinch et al., 2021; Jensen and Naumann, 2016; Newton, 2020; Reinhardt, 2015; Sibley et al., 2020). Political attitudes have to be controlled because partisanship affects how people view issues (Bullock and Lenz, 2019, Holmberg, 2007), which is also true in the COVID-19 situation where compliance and trust were divided between party supporters (Goldstein and Wiedemann, 2021; Kerr et al., 2021; Ward et al., 2020). In Hong Kong, heated social movement and conflict between the government and the residents would require control of these variables. How much people had been exposed to the partisanship effect was gauged by the political attentiveness variable.

2.2. Health experts’ communication treatment groups

The vignettes viewed by the control group contained only a neutral government announcement. The consent group received the same government announcement along with a supportive communication from health experts, health expert groups, or public health organizations. The dissent group received the government announcement together with a disagreeing remark given by other similar health experts and groups. The principle for making the pairs of treatment vignettes was similarity. Variations could be due to the length of the vignettes, the persons or groups presented, organization affiliation of the persons cited, and their arguments. This study attempts to make the vignettes in the pairs as similar as possible in all these aspects. For each COVID-19 measure, the vignettes used in consent and dissent groups were made to similar length. All experts used in the vignettes were presented with titles related to healthcare. The adoption of a person was paired with another person with a similar title, whereas the use of a medical organization was matched with another organization. In making the vignettes, results that appeared higher in google search result were considered first because it implied that the news report was more popular among the public. The goal was to mimic real-world information dissemination as closely as possible. The vignettes of public assembly ban were presented as an example. Please refer to Section S1 of Supplementary materials for the full sets of vignette.

3. Control group

It has been reported that on a number of occasions since the No-gathering Order came into force, the police have issued notices of objection to public meetings and processions on grounds of epidemic prevention, and have issued to participants a fixed penalty for violating the Order.

3.1. Consent treatment group

It has been reported that on a number of occasions since the No-gathering Order came into force, the police have issued notices of objection to public meetings and processions on grounds of epidemic prevention, and have issued to participants fixed penalty for violating the Order. Professor Wong Tze Wai, from the School of Public Health and Primary Care in the Chinese University of Hong Kong, said that the virus is apparently spreading rapidly through high-efficiency person-to-person transmission. Therefore, the public is urged to avoid densely populated places, including public assembly.

3.2. Dissent treatment group

It has been reported that on a number of occasions since the No-gathering Order came into force, the police have issued notices of objection to public meetings and processions on grounds of epidemic prevention, and have issued to participants fixed penalty for violating the Order. Dr. Leung Chi Chiu, Chairman of Advisory Committee on Communicable Diseases of the Hong Kong Medical Association, said that if everyone wears a mask during an assembly, and stands 1-m apart, there is no big problem. The risk of virus spread is low in outdoor assembly and open spaces.

3.3. Support for COVID-19 measures

After viewing each vignette, participants answered the following question: “To what extent do you support [the COVID-19 measure]?” They were asked to rate the answer using an 11-point scale, from “0” (least support) to “10” (strongly support). A total of three responses were recorded.

3.4. Health–expert–government interaction

As discussed above, for each COVID-19 policy measure, a participant was assigned randomly to the control, consent treatment, or dissent treatment groups. As there are three COVID-19 measures, each participant was assigned to a total of three random vignettes. For each participant, dummy variables concerning interaction between health experts and the government were constructed as follows:

Discord: participants viewed only dissent and control vignettes.

Contradiction: participants viewed a combination of consent, dissent, and control vignettes.

Consensus: participants viewed only consent and control vignettes.

3.5. Health expert and government trustworthiness

Participants were posed two questions in random order: “To what extent do you find health experts trustworthy?” and “To what extent do you find the Hong Kong government trustworthy?” They provided answers on an 11-point scale, with 0 being very untrustworthy and 10 being very trustworthy. Two responses were recorded.

3.6. Experience with COVID-19

Participants’ COVID-19 experience, COVID-19 knowledge, perception of responsibility, and economic preferences were collected as control variables. COVID-19 experience was measured via yes/no answers to the two statements: “I have lost my job or main source of income since the (COVID-19) outbreak,” and “I have been in quarantine because of COVID-19.”

Participants’ knowledge of COVID-19 was tested using four questions. Sample statements included: “a person recovered from COVID-19 will not be infected the second time.” Participants were asked to indicate whether the item was correct or not by choosing from “true,” “false,” and “not sure.” One mark was awarded for each correct answer. Total score ranged from 0 to 4, with higher scores reflecting better knowledge.

Perception of responsibility was assessed using the question: “The spread of COVID-19 is natural, not human-induced”; it was rated on an 11-point scale, ranging from “0” (strongly disagree) to “10” (strongly agree).

Economic preference was measured by the extent to which they agreed that “economic recovery is more important than pandemic control.” They responded on an 11-point scale, ranging from 0, which indicated that pandemic control was more important, to 10, which favored economic recovery.

3.7. Political attitudes

Prior to treatment, political attitudes were measured in four dimensions. The first item was about expectations with the government; participants were asked to what extent they agreed that “The Hong Kong government is competent in tackling COVID-19.” The second item was about freedom preference: “Freedom is more important than pandemic control.” The third items measured political attentiveness: “In general, do you pay attention to politics?” The answers to these questions were rated on a scale of 0–10, with 0 (10) indicating the strongly disagree (agree) for the first two items and least (most) attention for the third item, respectively.

The participants were also asked to self-report their political stance. In Hong Kong, the major political cleavage is not between the left and right but between being pro-government and pro-democracy. The choices provided in the survey were “pro-establishment,” “moderate/center,” “pan-democratic,” “pan-localist,” “others,” and “don't know.” For statistical analysis, these responses were grouped into three categories: pro-government (pro-establishment), opposition (democrat and localist), and others (not included in the above categories).

4. Demographics

The demographic background of participants including sex, education, age, and income.

4.1. Statistical analysis

The Chi-squared test and one-way ANOVA were used to confirm that the independent variables across the treatment and control groups were not statistically different. Two sets of ordinary least square linear regression analyses were conducted. The first set tested the associations between health experts’ communication and support for COVID-19 measures. The second set tested the association between health-expert–government interaction and trust level. All statistical analyses were performed using Stata version 16.1 (StataCorp LLC, College Station, Texas, USA). Statistical significance was set at p < 0.05.

5. Results

5.1. Descriptive statistics

5.1.1. Demographics of participants

Among the 1072 participants, more than half were females (53.6%) and had university education (56.5%); the average age was 39.8 years. The majority (54.4%) had income between 20 and 55 percentiles of the general population (Table 1 ).

Table 1.

Descriptive statistics of responses.

	N	% of full sample
Male	497	46.40%
University Education	606	56.50%
Age (years)	39.8 (12.2)
Income group: Bottom 20%	127	11.90%
Income group: Middle 20%–55%	583	54.40%
Income group: Top 45%	362	33.80%
Experience with COVID-19
Loss of income	204	19.00%
Quarantine experience	72	6.70%
Knowledge of COVID-19 (0–4)	2.7 (1.2)
COVID-19 is a natural occurrence (0–10)	4.0 (2.9)
Economic recovery over freedom (0–10)	2.9 (2.1)
Political attitude
Political stance: Pro-government	99	9.20%
Political stance: Opposition	301	28.10%
Political stance: Others	672	62.70%
Government's competence (0–10)	4.7 (2.8)
Freedom over pandemic control (0–10)	3.1 (2.4)
Political attentiveness (0–10)	6.1 (2.3)

Note: Mean and standard deviation in brackets are shown for discrete variables. Income groups are approximations of percentiles using 2021 Q2 data of the General Household Survey, Census and Statistics Department.

5.2. Experience with COVID-19

Approximately one-fifth of the participants (19%) declared loss of job or a major source of income since the outbreak of COVID-19. A small percentage (6.7%) were in quarantine because of COVID-19. The mean knowledge score was 2.7 out of 4 (Standard Deviation [SD]: 1.2). Their assessment of whether COVID-19 was a natural occurrence was overall 4/10 (SD: 2.9). Their economic preference over pandemic control was on average 2.9/10 (SD: 2.1) (Table 1).

5.3. Political attitude

The percentage of participants who declared themselves to be either pro-government or in opposition were 9.2% and 28.1%, respectively. Their expectation regarding government's competence to cope with COVID-19 was on average 4.7/10 (SD: 2.8). The average preference for freedom over pandemic control was 3.1/10 (SD: 2.4). The average political attentiveness was 6.1/10 (SD: 2.3) (Table 1).

5.4. Test for statistical difference in independent variables across treatment groups

Their distribution across the three treatment groups was 32% (consent), 32% (dissent), and 36% (control), respectively; it was about the same in all three COVID-19 measures. Using the Chi-squared test for categorical variables and ANOVA for other variables, we could not reject the null hypothesis that any of the characteristics listed are not systematically different across groups (all p-values >0.05) (Table 2 ). Thus, we were confident that the independent variables used in this sample were not statistically different across groups.

Table 2.

Statistical test result for difference in independent variables across treatment groups.

	Health experts' communication (consent/dissent/control)			Expert–gov interaction
	Contact-tracing	Restrictive-testing	Public assembly ban	(discord/contradiction/consensus)
Sex	0.654	0.289	0.908	0.911
Education	0.418	0.609	0.552	0.751
Age	0.963	0.789	0.409	0.186
Income group	0.328	0.422	0.475	0.735
Experience with COVID-19
Loss of income	0.558	0.178	0.34	0.218
Quarantine experience	0.841	0.258	0.443	0.354
Knowledge of COVID-19	0.649	0.873	0.843	0.354
COVID-19 is a natural occurrence	0.57	0.908	0.555	0.648
Economic recovery over freedom	0.944	0.485	0.7	0.834
Political attitude
Political stance: Pro-government	0.725	0.892	0.745	0.401
Political stance: Opposition	0.251	0.239	0.8	0.957
Government's competence	0.474	0.604	0.546	0.764
Freedom over pandemic control	0.241	0.422	0.817	0.931
Political attentiveness	0.69	0.385	0.924	0.581

5.5. Unconditioned support for COVID-19 measures and trustworthiness

Table 3 shows the unconditional support for different COVID-19 measures across health experts' communication treatment groups. In all three measures, the consent group had the highest support rate. In contrast, only the dissent group of the use of contact-tracing app had the lowest support rate. In the other two measures, the control group had the lowest support. Health experts' trustworthiness is higher than government's trustworthiness by 2.42 points on a 0–10 point scale. The difference across treatment groups for restriction-testing is significant with p value equals to 0.044. No other result is statistically significant across groups.

Table 3.

Support for COVID-19 measures.

COVID-19 measures Support rating (0–10)	Dissent	Control	Consent	Comparison across treatment groups (p-value)	All
Contact-tracing app	4.8	5.32	5.35	0.237	5.16
Restriction-testing	5.84	5.69	6.00	0.044	5.84
Public assembly ban	5.57	5.45	5.83	0.264	5.61
Health-expert–government interaction	Discord	Contradiction	Consensus
Health experts' trustworthiness (0–10)	6.65	6.26	6.37	0.338	6.46
Government's trustworthiness (0–10)	3.97	3.97	4.14	0.286	4.04

Regression for Support of COVID-19 Measures.

The regression results are shown in Table 4 (full table in Section S2 of Supplementary materials). In these regressions, the control group is used as the base to study whether being assigned dissent and consent groups affects the support for COVID-19 measures. No treatment effect of statistical significance was detected for health experts’ communication except for public assembly ban (Model 3), in which the consent group showed higher support compared to the control group (coefficient = 0.333, p = 0.045). Belief in the natural occurrence of COVID-19, being pro-government, having higher expectation from the government, and older age were found to be associated with higher support for all COVID-19 measures tested. Having an opposition political stance and higher freedom preference showed lower support. The models used here explained 48.6%–56.5% of the support for COVID-19 measures.

Table 4.

Linear regression analysis of COVID-19 measures.

Model	1	2	3
	Contact-tracing	Restrictive-testing operations	Public assembly ban
Health experts' communication (base = control)
-Dissent	−0.186	−0.183	0.191
	(0.172)	(0.158)	(0.166)
-Consent	0.0729	−0.0913	0.333*
	(0.171)	(0.156)	(0.166)
Control for COVID-19 experience	Yes	Yes	Yes
Control for political attitudes	Yes	Yes	Yes
Control for demographics	Yes	Yes	Yes

Standard errors in parentheses; ***p < 0.001, **p < 0.01, *p < 0.05.

5.6. Regression for trustworthiness

In Table 5 (full table in Section S3 of Supplementary materials), health-expert–government interaction dummies were found to have effect on health experts' trustworthiness (Model 4). Here, the discord group is used as the base to study whether being assigned the contradiction and consensus groups affects trustworthiness. Participants in the contradiction (coefficient = −0.374, p = 0.019) and consensus groups (coefficient = −0.339, p = 0.010) found experts less trustworthy compared to those in the discord group. The same effect was not found in government's trustworthiness in Model 5. In Model 6, the interaction term between government's trustworthiness and the contradiction and consensus dummies were included as independent variables on health experts' trustworthiness. The higher the trust in government of the participants in the contradiction (coefficient = 0.180, p < 0.001) and consensus groups (coefficient = 0.252, p < 0.001), the higher trust they had in health experts. In Model 7, a new variable, “average support for COVID-19 measures” was introduced, averaging the support rating for the three measures asked. This variable and its interaction terms with the contradiction and consensus dummies were added as independent variables in the model. Higher support for COVID-19 measures correlated with health experts' trustworthiness with statistical significance (coefficient = 0.161, p < 0.001). In the consensus group, participants with higher support for COVID-19 measures showed higher trust in experts (coefficient = 0.145, p = 0.001) as opposed to the other two groups. Across all three models of health experts' trustworthiness, participants who had higher expectations with the government, lower freedom preference, and higher political attentiveness, had higher trust in health experts. Middle and upper income participants had lower trust in health experts compared to those whose income is in the bottom 20 percentile.

Table 5.

Linear regression analysis of health experts' and government's trustworthiness.

Model	4	5	6	7
	Experts' trustworthiness	Government's trustworthiness	Experts' trustworthiness	Experts' trustworthiness
Health-expert–government interaction (base = discord)
-Contradiction	−0.374**	−0.0639	−1.081***	−0.674**
	(0.159)	(0.159)	(0.233)	(0.342)
-Consensus	−0.339***	0.0136	−1.365***	−1.165***
	(0.131)	(0.130)	(0.194)	(0.281)
Gov. Trustworthiness X Contradiction			0.180***
			(0.0442)
Gov. Trustworthiness X Consensus			0.252***
			(0.0357)
Average support for COVID-19 measures				0.161***
				(0.0404)
Average support for COVID-19 measures X Contradiction				0.0507
				(0.0552)
Average support for COVID-19 measures X Consensus				0.145***
				(0.0453)
Control for COVID-19 experience	Yes	Yes	Yes	Yes
Control for political attitudes	Yes	Yes	Yes	Yes
Control for demographics	Yes	Yes	Yes	Yes

Standard errors in parentheses; ***p < 0.001, **p < 0.01, *p < 0.05.

6. Discussion

During the COVID-19 pandemic, governments worldwide struggled with citizens' compliance with disease containment measures. Moreover, health experts served as the source of authority for governments to take appropriate actions and persuade the public to support COVID-19 measures. This study finds some evidence that health experts' communication helps increase support for ban of public assembly. However, if health experts’ opinions agree with government directives, their trustworthiness may be reduced.

The three measures examined impacted freedom and rights in different dimensions. In the case of contact-tracing, although at the time of survey, the mobility data were only stored in users’ phone as proclaimed by the government, the prolonged use of contact-tracing app—and the institution once set up—could be an intrusion of privacy (Chan and Saqib, 2021; Dowthwaite et al., 2021) — a reason widely cited for reluctance in adoption. Restriction-testing is essentially a limited-scale lockdown plus mandatory testing, which was an intrusion of personal freedom of movement. Ban of public assembly was in violation of civil rights.

This study examined the efficacy of health experts’ communication in garnering public support for COVID-19 measures, including the use of contact-tracing app, restriction-testing, and public assembly ban; overall, the evidence was weak. In the sample, the support for the use of contact-tracing app was the lowest at 5.16/10 among the three measures (Table 3). The low support for digital contact-tracing in Hong Kong was related to the lack of trust in data privacy and efficacy and low trust in the government (Huang et al., 2021; Voo et al., 2021). The support for restriction-testing was the highest at 5.84/10 out of the three measures. No known survey has opinion rating on this measure for comparison. At the time of survey, most restriction-testing were limited to a few buildings and the operations usually began at dawn and ended the next morning. The restriction on personal freedom was little compared to a large-scale lockdown. Additionally, not many people were affected, as it only targeted a small population of Hong Kong. Regarding the support of public assembly ban, the support rate recorded in this study was 5.61/10. A comparable measure compiled by the Hong Kong Public Opinion Research Institute (2022a) on May 31, 2021 showed that 98.1% participants found the gathering ban too strict.

We found that viewing a vignette of a health expert, with a professional title, citing reasons, and urging the public not to attend public assembly can increase support for the public assembly ban when compared with only a government announcement of such a rule and the penalty involved. This is coherent with the finding in the US and the UK that a COVID-19 policy supported by a health expert received greater public support as compared to a policy supported only by lawmakers (Arceneaux et al., 2020). In the study, health experts increased support for measures that help reduce infection, such as mandatory stay-home orders and use of a contact tracing app; however, their opinions on indefinitely postponing elections were ignored. The expert effect on increasing compliance is still doubtful, given mixed evidence in the field. Deslatte (2020) tested on the messenger effect of experts and found no significant difference from the control. In fact, expert actors weaken the effect of a pro-public-health message relative to a federal official. Another study found that experts’ communication can induce participants of certain political stances, but not all, to have higher vaccine intent (Vlasceanu and Coman, 2022).

Overall, this study found weak evidence that experts’ communication can induce support for measures that restrict freedom; a few reasons may explain it. First, as the participants only briefly read the vignettes, it might not induce an effect strong enough to be detected. Second, there might be pre-exposure effect. In the world of information overload, participants may have already read similar communications. Third, support for different measures could be context-specific, which differs across countries and time of survey.

The second part of this study examines whether experts agreeing and disagreeing with government measures affects experts' trustworthiness. Participants who read only experts agreeing with government's directives (consensus group), and who read experts sometimes agreeing and sometimes disagreeing (contradiction group), had lower trust in experts compared to those reading only disagreements (discord group); the effect was statistically significant. A similar effect of lowering trust was not found toward government's trustworthiness.

Despite recent distrust in experts and the refrain: “had enough of experts,” (Clarke and Newman, 2017) the uncertainty and high stakes around COVID-19 helped experts gain a central role in pandemic management (Radaelli, 1999). Experts were called upon to offer scientific advice, formulate strategies, and communicate with the public, and earned prominent roles based on two main reasons: their epistemic authority in a field that is based on scientific reasoning and empirical support; and their independence. They are trusted as virtuous and personally disinterested individuals, implying no personal stakes or desire for immediate rewards. Governments rely on experts because: (1) they are likely to provide efficacious solutions; and (2) help curb controversy (Lavazza and Farina, 2020). By involving them in finding solutions, or simply by consulting them, governments can delegate—or appear to delegate—power to experts, thereby reducing or negating social conflicts (Flinders and Dimova, 2020).

Possible explanations are provided for decrease in health experts' trustworthiness when they agreed with government's COVID-19 policy. First, health experts had agreed with a low-trustworthiness entity. Several studies in Hong Kong indicated that the Hong Kong government had low trust (Ho et al., 2021; Huang et al., 2021; Yuen et al., 2021). In this sample, it had an overall trustworthiness of 4.04/10, which was much lower than that of experts (6.46). Agreeing with a low-trust entity may have a spillover effect on experts themselves. The result of Model 6 is consistent with this argument. It shows that for participants with higher trust in the government, when they were presented with experts' views that sometimes agreed with the government's, they had higher trust in the experts; when presented with experts' views that unilaterally agreed with the government, the trust in experts was even higher. It is noteworthy that the signs of these effects are opposite to the unconditional (negative) effects of reading expert views that are sometimes or always in agreement with the government's—the main result of Model 4. A similar study in Hong Kong found that medical experts bearing a government title could not induce more trust in medical experts themselves; a source critical of the government could enhance the credibility of official government messages (Sheen et al., 2021).

The second explanation is that by agreeing with government directives, health experts shoulder the responsibility for these public decisions, and if support is low for these COVID-19 measures, health experts also take the blame. Normally, policymakers—that is, politicians and bureaucrats—bear the responsibility of consequences in public decision. Technocrats prevail at situations of high uncertainty and when information and ideas are of high cost, such as the COVID-19. High salience of COVID-19 also leads to politicization of the issue (Haas, 1992). Health experts who performed prominent roles in COVID-19 management have to bear responsibility for the decisions taken. The British case of preferring herd immunity and the US policy of excluding disabled people from medical care during COVID-19 are examples (Lavazza and Farina, 2020). In these cases, health experts were no longer providing technical solutions limited to their area of expertise. The public, therefore, held them at least partially responsible for the consequences of these measures. Even though in a few countries the final decision largely rested with government officials, the act of concurring with the government resulted in sharing the responsibility. This is coherent with the blame defection argument for bringing in health experts (Flinders and Dimova, 2020), —regardless of whether blame defection is intentional or not.

The third explanation is that the disagreement signals independence. Health experts can signal independence by publicly disagreeing with the government public health measures, but not publicly agreeing with public health measures, even if they could still be independent in the latter case. However, this independence explanation is not distinguishable from the first (trust spillover) explanation in this single-case study of Hong Kong. But they can be separated in a state in which the government has high trust, then agreeing with the government would induce higher expert trustworthiness according to the ‘trust spillover’ explanation, but lower expert trustworthiness according to the ‘independence’ explanation.

In this survey, participants exposed to vignettes of health experts agreeing with government directives might have prompted them to think that experts were endorsing or supporting the decision, and the experts, therefore, shared the responsibility. The average support rate for the three COVID-19 measures, and its interaction terms with the contradiction and consensus dummies, were added to Model 7. Support for COVID-19 policies was statistically significant and associated with experts’ trustworthiness. Participants in the consensus group showing stronger support had higher trust in experts as opposed to those in the same group showing weaker support. The same interaction effect with support for COVID-19 measures was not found in the discord and contradicting groups; in these groups, discord between health experts and government directives might have relieved the experts from sharing the responsibility.

This study has several limitations. First, the result could be affected by the use of real-world vignettes, which have the advantage of social–ecological validity to help inform potential solutions to real-world problems (McDonald, 2020; Reis et al., 2000). However, their use also prevents the isolation of the exact factor in a health expert that induces (or fails to induce) trust and the associated causal mechanism—we leave this question for future research. A related limitation is that using a real-world vignette might have pre-treatment exposure effect, which may affect the results. However, given the salience and time-span of the pandemic, participants would have likely read ample COVID-19 information and recommendations anyway. This study partially captured the pre-treatment exposure by including variables such as knowledge of COVID-19 and COVID-19 experiences.

Additionally, participants were randomly allocated to experiment groups. Thus, there is no compelling reason to believe that participants assigned to different groups had different pre-treatment effects. Further studies could examine whether the effect of lowering expert trustworthiness could be generalized to other countries and the causal link or reasons behind lower trust in experts for concurring with the government.

7. Conclusion

This study finds limited evidence of the efficacy of using health experts to induce support for COVID-19 measures. However, health experts agreeing with government's COVID-19 directives could reduce health experts' trustworthiness, compared to health experts disagreeing with the government.

Governments involved health experts in policymaking for pandemic management. As it became a moral mission for experts to motivate public compliance for public good, they publicly supported government directives. However, evidence regarding the efficacy of such communication strategy in inducing compliance is mixed. The worse is providing recommendations concurring with government's directives to encourage compliance as that could reduce health experts' trustworthiness.

The implication of the findings from this study is profound. While the expected effect on compliance is not substantiated, what the health experts did could hurt their trustworthiness. Many studies have reported lower trust in science and scientists after epidemics (Eichengreen et al., 2021; Feufel et al., 2010). The repeated use of health experts for dissemination for information could be part of the reason. Trust in health experts, medical organizations, and health authorities has been widely cited as an important factor that correlates with citizens’ compliance during epidemics (Gilles et al., 2011; Siegrist and Zingg, 2014). Low trust in health experts and scientists could undermine the effectiveness of risk communication in the future.

Funding

The work was supported by the Hong Kong Institute of Economics and Business Strategy, University of Hong Kong.

Ethical issues

The study was approved by the Human Research Ethics Committee of the University of Hong Kong.

Author contribution

Vera Wing Han Yuen is the sole author of the manuscript.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

The data that has been used is confidential.