Perceived government-expert discord and evaluation of COVID-19 policies in Hong Kong

ABSTRACT

This study examines citizens’ retrospective evaluation of government policies during the COVID-19 pandemic. It focuses specifically on the impact of perceived government – expert discord, i.e. public perceptions of disagreement between government officials and the medical expert community. Medical experts played important roles during the pandemic and enjoyed high levels of credibility, whereas government leaders were sometimes seen as driven by political motivations. Hence, perceived government – expert discord might negatively affect policy evaluation. Analysis of a post-pandemic survey in Hong Kong (N = 1,002) shows that people who perceive significant government – expert discord are likelier to blame the government for negative happenings during the pandemic, less likely to credit the government for positive outcomes, and less likely to see the government’s control measures as appropriate. These relationships are stronger among people who suffered more, economically or mentally, during the pandemic. Credit attribution, blame attribution, and perceived appropriateness of control measures, in turn, influence overall policy evaluation.

KEYWORDS:

• Policy evaluation
• public opinion
• experts
• credit-and-blame attribution
• COVID-19
• Hong Kong

1. Introduction

The COVID-19 pandemic has presented serious challenges to governments around the world. While the original task was to contain the virus, policy decisions became more complicated later, as governments needed to achieve the competing goals of saving lives and minimizing the negative consequences of strict pandemic control. Numerous studies have documented public opinion towards specific pandemic policy measures in various countries (Brouard et al., 2022; Collignon et al., 2021; Kerr & van der Linden, 2022). However, given that the COVID-19 pandemic had ended, how would people retrospectively evaluate the government’s pandemic policies? What factors influence such retrospective evaluations?

This study examines these questions. How people retrospectively evaluate government policies is worth examining because such evaluations could influence public opinion when a comparable crisis emerges in the future. For instance, Hong Kong’s experience with SARS shaped the society’s response to COVID-19 (S. Yuen et al., 2021). Similarly, what lessons people take from COVID-19 could shape their reactions to future health crises.

Analytically, this study focuses on the impact of perceived government – expert discord, i.e. perceptions of disagreement between government officials and the medical expert community on matters of public health. During the pandemic, medical experts played an important role in informing the government, the media, and the public regarding the character of the disease and the possible efficacy of various mitigation measures. Many governments proactively drew upon the authority of medical experts to legitimize their policies (Laebens & Ozturk, 2022; Weingart et al., 2022). However, there were also cases in which governments were criticized for politicizing the pandemic and ignoring expert advice (Gallo et al., 2022). We expect perceptions of government – expert discord to influence citizens’ attribution of credits and blames, evaluation of the appropriateness of the pandemic control measures, and overall evaluation of government policies.

This study examines Hong Kong, where the pandemic occurred in the midst of the Anti-Extradition Law Amendment Bill (Anti-ELAB) Movement. Therefore, the government had to combat the pandemic amidst very low levels of public trust (Chan, 2021), which could have influenced citizens’ willingness to cooperate with government policies and enlarged the credibility gap between officials and experts. Nevertheless, the protests also strengthened the mobilizing structures throughout society. Together with the experience of SARS, a civil-society-led pandemic response helped Hong Kong keep the number of infected cases and deaths low (S. Yuen et al., 2021) until the Omicron wave in early 2022. Normalization was signified by the end of the mask-wearing requirement in March 2023. Although the association of the pandemic with a huge protest movement is unique to Hong Kong, the key notion of government-expert discord is applicable to various contexts. The Hong Kong case should remain illustrative of how public opinions about pandemic policies were formed.

2. Public opinion during COVID and government–expert discord

Numerous studies have examined the factors shaping public opinion towards government COVID-19 policies. Many early publications highlighted the role of partisanship (Allcott et al., 2020; Chen et al., 2022; Gadarian et al., 2021). For example, Pignataro (2021) showed that, in Costa Rica, past voting records, evaluation of the economy, and approval of the government’s economic and health measures predicted support for the government during the pandemic (also see Ferraresi & Gucciardi, 2022; V. W. H. Yuen, 2022). However, degree of politicization of the pandemic depends on the availability of elite cues (Spalti et al., 2021). Flores et al. (2022) found that, in seven countries, cues from politicians elicited positive responses from the political in-group but negative responses from the out-group. Politician cues thus tended to polarize. In contrast, bipartisan cues and expert cues tend to depolarize and generate more policy support.

Indeed, the role of experts is another frequently addressed theme. The public tended to see experts as knowledgeable, politically neutral, and oriented towards public interests. Motta and Benegal (2023) found that the pandemic strengthened people’s belief in scientific expertise. Granados Samayoa et al. (2021) found that a belief in scientific expertise could enhance people’s compliance with COVID-related measures. Admittedly, certain factors may limit the power of experts. Collignon et al. (2021) showed that degree of populism affects attitudes towards pandemic science. More importantly, given the pandemic’s complexity, expert consensus cannot be taken for granted (Vicentini & Galanti, 2021). Lack of expert consensus can lead to worries among the public, which in turn lead to weaker support for mitigation policies (Kerr & van der Linden, 2022).

Despite these limitations, some studies directly compared people’s attitudes towards experts and politicians and found that people tended to see experts as more important and trustworthy. Morani et al. (2022) found that, in the UK, the news audience preferred to hear more from medical experts and health professionals than from politicians. Barcelo et al. (2023) showed that expert endorsement, but not politicians’ endorsement, consistently influenced people’s choice of vaccine. Cole et al. (2022) showed that COVID policies suggested by experts received stronger support than policies suggested by politicians.

As a result, to the extent that the expert community does have a consensual view, people are likely to expect the government to base its policy decisions on expert advice. Hence, perceived disagreement between government officials and the expert community could negatively affect people’s evaluation of the government. Various authors have noted that the more politicians see the pandemic as serious, the likelier they are to claim to be following expert advice (Laebens & Ozturk, 2022; Weingart et al., 2022). Government leaders tend to avoid being seen as putting forward measures without support from the scientific community. Indeed, when government – expert discord arises, people might see government leaders as politicizing the pandemic to serve their own purposes (Gallo et al., 2022; Wright et al., 2022).

Nevertheless, the extant literature has not directly examined the degree to which people perceive the presence of government-expert disagreement, as well as how perceptions of such disagreement influence public opinion. This study thus fills the gap by examining how perceived government – expert discord undermines people’s evaluation of COVID-19 policies. For this purpose, the next section will elaborate on the components of policy evaluation.

3. Components of COVID-19 policy evaluation

During the pandemic, governments around the world put forward a wide range of policies, such as lockdowns, social distancing, and vaccination requirements. Many studies have examined public opinion towards specific policies, such as lockdowns (e.g. Collignon et al., 2021; Peretti-Watel et al., 2021), lifting of restrictions (e.g. Stoeckel et al., 2022), and vaccination programmes (e.g. Medina et al., 2021; V. W. H. Yuen, 2022). In contrast, this study examines people’s overall evaluation of COVID-19 policies.

Lee (2022) provided a model that could inform the present study. He started with the idea that, based on the logic of retrospective evaluation, severity of the pandemic should relate negatively to public evaluation of the government, i.e. when number of cases and deaths increased, the pandemic would seem to be out of control, and public evaluation should become more negative (Chen et al., 2020). However, two factors mitigate this relationship. The first is credit-and-blame attribution. Just as the impact of the economy on public opinion can depend on whether people see the government as responsible for economic conditions (Fiorina et al., 2003; Rudolph, 2003), the pandemic situation may influence public opinion towards the government differently depending on whether people credit the government when the situation improves and blame the government when the situation worsens.

Second, an improved pandemic situation may not lead to a more positive policy evaluation because it might increase people’s grievances against control measures. Control measures, such as lockdowns and social distancing, have social costs. People’s livelihoods and mental health can suffer (Onyeaka et al., 2021; Panchal et al., 2023). People might see some control measures as unnecessarily strict, and they could become particularly disgruntled when the pandemic seemed to be under control. Indeed, Ferraresi and Gucciardi (2022) showed that, in Italy, disapproval against the government was strongest in places where the pandemic situation was positive.

Lee’s (2022) framework was aimed at understanding the dynamic evolution of aggregate-level public opinion, whereas this study is concerned with individual-level evaluation of government policies. However, his framework highlights the factors that need to be considered, namely, credit-and-blame attribution and perceived appropriateness of the strictness of the pandemic control measures. These two can be treated as mediators between perceived government-expert discord and overall policy evaluation. That is, when people perceive the government and the expert community as disagreeing with each other, they may become less likely to credit the government for positive results, likelier to blame the government for negative results, and less likely to see the control measures as appropriately strict. Both credit-and-blame attribution and perceived appropriateness of the control measures would then shape overall policy evaluation.

In addition, we expect an interaction effect between negative impact of the pandemic on oneself and perceived government-expert discord on credit-and-blame attribution and perceived appropriateness of pandemic control measures. As noted, the pandemic had negatively impacted people’s livelihoods and mental health. Individuals who were adversely affected to a larger degree were likelier to find pandemic-related information personally relevant. Personal relevance could motivate more systematic information processing (Noe & Hammitt, 1992; Petty & Cacioppo, 1996). This should strengthen the impact of perceived government – expert discord. In other words, while perceived government – expert discord should relate negatively to credit-and-blame attribution and perceived appropriateness of the control measures, the negative relationships should be stronger among people who suffered economically or mentally to larger degrees during the pandemic.

4. Context and hypotheses

The COVID-19 outbreak in Hong Kong started in late January 2020. The pandemic came after months of protests of the Anti-ELAB Movement (Cheng et al., 2022; Lee et al., 2019). According to the Hong Kong Public Opinion Research Institute, only 19.1% of Hong Kong citizens trusted the government in January 2020, whereas 69.2% expressed distrust.¹

Nevertheless, the number of cases and deaths remained low for a long period of time. By the end of January 2022, there were only around 30,000 confirmed cases and around 200 deaths. Scholars have attributed the success of pandemic control to a strong civil-society-led response (Hartley & Jarvis, 2020; Wan et al., 2020; Wong, 2022; S. Yuen et al., 2021).

Distrust would continue to negatively impinge on government efforts, e.g. through undermining citizens’ willingness to vaccinate (Chung et al., 2022; Ye et al., 2023). Meanwhile, the public was concerned with whether the Hong Kong government was forced to follow China’s pandemic policy despite the vast differences in social conditions between Hong Kong and the mainland. When Hong Kong suffered from the Omicron wave in early 2022, with the number of deaths jumping to 9,400 in three months, the city was seen as being caught between China and the world (Wang & Ramzy, 2022).

From the beginning of the pandemic, the Hong Kong Government solicited help from medical experts. Some prominent medical experts were well-known to be government advisors, and some citizens might see the views of these experts as ‘quasi-official’. However, the government and the medical expert community remain distinctive entities. The media regularly solicited the views of medical experts who had few ties with the government. More importantly, occasional expert – government disagreements did arise (V. W. H. Yuen, 2023, p. 6) and were reported by the media (Heung, 2022). There is no strong reason to assume that citizens would treat the government and the medical expert community as so intertwined as to be non-distinguishable. Perceptions of government-expert discord could arise to different degrees across individuals. When people perceive the presence of such disagreement, they should be particularly likely to see the government as acting unreasonably.

Considering the context and the earlier conceptual discussions, several hypotheses and research questions can be stated. First, the analysis will explore the correlates of perceived government – expert discord. Political attitudes can shape the evaluation of government performance and affect perceptions of government – expert discord. People opposed to the government may be more prone to see the government as acting inappropriately and, hence, more prone to see the government as deviating from expert advice. Besides, in the pandemic, public confusion may come from other sources, such as information overload, disagreement among experts (Vicentini & Galanti, 2021), or the presence of misinformation (Brennen et al., 2020). Our analysis thus differentiates among: a) perceived government – expert discord; b) exposure to COVID misinformation, i.e. the extent to which people perceive the presence of misinformation surrounding the pandemic; and c) COVID information confusion, i.e. a general perception that information about the COVID pandemic is confusing and difficult to handle. These factors are likely to be interrelated. A general research question was posed:

Q1:

How does perceived government – expert discord relate to political inclination, exposure to COVID misinformation, and COVID information confusion?.

As explicated, this study expects perceived government – expert discord to relate to credit-and-blame attribution and perceived appropriateness of the pandemic control measures. For the latter, in addition to perceived appropriateness of the measures’ strictness, the analysis also examines the perceived appropriateness of the timing of lifting the control measures (Stoeckel et al., 2022). It is a variant of perceived appropriateness of the control measures’ strictness, but with an emphasis on temporality. Besides, following earlier discussions, the various relationships are expected to be stronger among people who had experienced stronger negative economic or mental impact during the pandemic. The three hypotheses are stated as follows:

H1:

Perceived government – expert discord is associated with a lower likelihood of crediting the government for positive results and a higher likelihood of blaming the government for negative outcomes.

H2:

Perceived government – expert discord is associated with a lower likelihood of seeing the government’s control measures as appropriate in terms of strictness and time of lifting.

H3:

The relationships stipulated in H1 and H2 are stronger among people whose economic interests and mental wellbeing were negatively affected to larger extents during the pandemic.

Finally, both credit-and-blame attribution and perceived appropriateness of the control measures should influence overall policy evaluation. Therefore, they should mediate the relationships between perceived government – expert discord and overall policy evaluation, i.e. they serve as the mechanisms through which perceived government – expert discord influences overall policy evaluation. A hypothesis is stated as follows:

H4:

Credit-and-blame attribution as well as perceived appropriateness of the pandemic control measures mediate the relationship between perceived government – expert discord and overall policy evaluation.

5. Methods and data

5.1. Sampling

The data analysed below came from a telephone survey conducted between May 10 and 9 June 2023, by the Center for Communication and Public Opinion Survey at the Chinese University of Hong Kong. The survey was conducted two months after the government lifted the mask-wearing requirement, the last pandemic control measure. The respondents were Cantonese-speaking Hong Kong residents aged 18 or above. A telephone number database was created by combining all four-digit prefixes in use for landlines and mobile phones with the full set of possible 10,000 four-digit suffixes (i.e. 0000 to 9999). Specific numbers were randomly drawn by the computer during the survey. For landline numbers, the most recent birthday method was used to select the target respondent. A total of 1,002 interviews were completed. The response rate is 34.5% following American Association of Public Opinion Research’s response rate formula 3.²

The sample included 55.7% females; 16.6% aged between 18 and 29, and 32.4% aged 60 or above; 41.9% had a tertiary degree. The profile is close to that of the population in age and gender (52.9% females, 13.4% aged between 18 and 29, and 35.3% aged 60 or above). The sample had substantially higher educational achievement, as only 29.9% of Hong Kong citizens had a tertiary degree. The sample was weighted according to the sex × age × education distribution of the population when conducting the analysis.

5.2. Operationalization

Evaluation of COVID policies was measured by a question asking the respondents to provide an overall evaluation of the Hong Kong Government’s pandemic control policies in the previous three years. Answers were registered on a five-point Likert scale ranging from 1 = very bad to 5 = very good (M = 3.01, SD = 1.02).

5.2.1. Credit-and-blame attribution

Two questions asked the respondents to judge, in the previous three years, who was responsible when the pandemic situation 1) was light and 2) was severe. The answering options were ‘citizens’, ‘handling by the government’, and ‘both were equally important’. Given this study’s focus on policy evaluation, a scale was created so that 1 = crediting/blaming the government, 0.5 = ‘both were equally important’, and 0 = not crediting/blaming the government (crediting the government: M = 0.33, SD = 0.44; blaming the government: M = 0.55, SD = 0.47).

5.2.2. Appropriateness of pandemic control measures

The respondents were asked to indicate if, overall speaking, they saw the pandemic control measures as being too tight, too loose, or appropriate. The concern of the study is whether the key independent variable would lead people to evaluate the measures positively or negatively. Since both ‘too tight’ and ‘too loose’ are negative judgements, the item was turned into a dichotomous variable with 1 = appropriate (55.4%) and 0 = too tight or too loose. Another question asked if the respondents saw the lifting of the measures as coming too early, too late, or at an appropriate time. It was similarly turned into a dichotomous variable (% appropriate = 47.4%).

5.2.3. Perceived government – expert discord

A set of four questions asked the respondents to indicate, using a four-point scale (1 = never and 4 = frequently), whether they had: 1) been exposed to COVID-related misinformation or rumours, 2) felt confused due to disagreement between officials and medical experts, 3) felt confused due to disagreement among experts, and 4) felt confused or tired due to having too much information. The items were positively correlated (r ranged from 0.25 to 0.77). However, they represent distinctive concepts for this article. Item (1) represents exposure to COVID misinformation (M = 2.58, SD = 1.07). Item (2) represents perceived government – expert discord (M = 2.64, SD = 1.13). Both the third and fourth items are pertinent to the notion of COVID information confusion. They were highly correlated (r = 0.63). For analytical parsimony, they were combined through principal component analysis into a single variable representing COVID information confusion (M = 0.00, SD = 1.00).

5.2.4. Partisanship

Respondents were asked about their political leanings. Answering options included ‘localists’, ‘democrats’, ‘centrists’, ‘pro-establishment’, and ‘no political leaning’.³ Two dummy variables were produced: 1) pro-opposition (democrats or localists, 18.0% in total) vs. others, and 2) pro-government (i.e. pro-establishment, 9.6%) vs. others.

5.2.5. Negative impact of COVID on oneself

The respondents were asked if, during the pandemic, they or their family had 1) suffered from economic losses and 2) suffered mentally. Answers were registered with a three-point scale (1 = no loss/impact, 2 = some loss/short-term impact, and 3 = much loss/enduring impact). The two items were positively correlated at r = 0.36. For analytical parsimony, they were combined through principal component analysis to form an index (M = 0.00, SD = 1.00).

5.2.6. Control variables

Other variables used in the multivariate analysis included four demographics (age, sex, educational level, and family income), attention to health news (average of two items), active information seeking (average of five items), and general confusion about health information (average of two items). Details of operationalizations are omitted due to space limitations and are available upon request.

6. Analysis and findings

6.1. Predicting perceived government – expert discord

The first column of Table 1 summarizes the bivariate correlations between perceived government – expert discord and the controls. Pro-opposition and pro-govenrment citizens are more and less likely, respectively, to perceive significant government – expert discord. Hence perceptions of government – expert discord are indeed related to people’s political predilection. However, perceived government – expert discord is most strongly correlated with COVID information confusion. As column 2 shows, when all variables are put into a single regression model, only perceived COVID information confusion relates strongly to perceived government – expert discord.

Table 1. Predicting perceived government–expert discord.

	Bivariate correlations	Regression model 1	Regression model 2
Sex	−.05	−.07**	−.06
Age	−.16***	.03	.00
Education	.15***	.01	.04
Income	.13***	.02	.04
Attention to health news	.07*	−.02	−.02
Active information seeking	.17***	.05	.13***
Pro-opposition	.20***	.06**	.14***
Pro-government	−.10**	−.03	−.07*
Negative impact of COVID	.27***	.04	.25***
General health info. confusion	.10**	.04	.11***
COVID misinformation	.25***	.03	.17***
Perceived COVID info. confusion	.73***	.69***
Adjusted R^2		0.539***	0.179***

Entries are Pearson correlation coefficients and standardized regression coefficients. Missing values in income were replaced by the mean. Missing values for other variables were deleted pairwise. N = 977 for the regression analysis; ***p < 0.001; **p < 0.01; *p < 0.05.

If perceived COVID information confusion is removed, active health information seeking, the two political stance variables, negative impact of COVID on self, confusion about health information in general, and exposure to COVID misinformation all relate significantly to perceived government – expert discord. On the whole, perceived government – expert discord can be considered primarily a part of people’s perceptions of an ‘infodemic’. Political attitudes have some, but not very strong, impact.

6.2. Predicting credit-and-blame attribution and perceived appropriateness of control measures

We now turn to H1 to H3. Multiple regression analysis was conducted to examine the predictors of credit-and-blame attribution. The full model contains all control variables, perceived government – expert discord, and an interaction between perceived government – expert discord and negative impact of COVID on self.⁴ The first four columns of Table 2 summarize the findings. Supporting H1, perceived government – expert discord relates negatively to the dependent variable both before and after the interaction term was added. Notably, neither exposure to COVID misinformation nor perceived COVID information confusion relates significantly to the dependent variable. Perceived government – expert discord stands out as the factor that influences credit attribution. Moreover, there is a significant negative interaction effect between perceived government – expert discord and negative impact of COVID on self (β = −0.07). That is, the negative impact of perceived government – expert discord on crediting the government is stronger (i.e. more negative) among those who suffered more economically and/or mentally during the pandemic.

Table 2. Impact of perceived government–expert discord on credit-and-blame attribution and perceived appropriateness of control measures.

	Crediting the government		Blaming the government		Degree of strictness appropriate		Timing to lift measures appropriate
Sex	−.04	−.04	−.00	.00	1.46**	1.45**	1.35*	1.34*
Age	.03	.03	−.12**	−.12**	1.02	1.02	1.10***	1.11***
Education	−.09*	−.09	−.17***	−.17***	0.99	0.99	0.94	0.94
Income	−.03	−.04	.02	.02	0.97	0.97	1.04	1.04
Attention to health news	.05	.05	−.07	−.07	1.18	1.21	1.19	1.20
Active information seeking	.03	.03	.02	.02	1.13	1.11	0.89	0.88
Pro-opposition	−.02	−.03	.20***	.20***	0.28***	0.27***	0.43***	0.42***
Pro-government	.13***	.13***	−.02	−.02	1.77*	1.76*	1.53	1.52
Negative impact of COVID	.07*	.08*	.05	.04	0.94	0.98	0.79**	0.81**
General health info. confusion	−.03	−.02	−.06*	−.07*	1.19	1.22*	1.12	1.14
COVID misinformation	−.01	−.01	.02	.02	0.99	1.00	1.01	1.01
COVID info. confusion	.03	.02	.01	.02	0.96	0.95	1.02	1.01
Perceived gov.–expert discord	−.09*	−.09*	.17***	.17***	0.75**	0.75**	0.86	0.86
Perceived gov.–expert discord X negative impact of COVID		−.07*		.03		0.83**		0.86*
Adjusted R^2	.042***	.045***	0.112***	0.112***
Nagel-karke R^2					0.176	0.187	0.148	0.155
Chi-square					136.00***	144.75***	113.43***	118.64***
N	975	975	975	975	962	962	963	961

Entries in the first four columns are standardized regression coefficients. Entries in the last four columns are odds ratios. Missing values in income were replaced by the mean. Missing values for other variables were deleted pairwise in multiple regression and listwise in logistic regression. ***p < 0.001; **p < 0.01; *p < 0.05.

Columns 3 and 4 of Table 2 show that, also consistent with H1, people who perceive significant government – expert discord are likelier to blame the government both before and after the interaction term was added. However, the interaction term is insignificant. Therefore, H3 is only partially supported.

Since perceived appropriateness of the control measures is represented by two dichotomous variables, logistic regression analysis was conducted to examine H2. The fifth and sixth columns show that people who perceived significant government – expert discord are less likely to see the strictness of the control measures as appropriate (odds ratio < 1.00). Meanwhile, there is a significant interaction effect between perceived government – expert discord and negative impact of COVID when it was added into the model. The odds ratio of the interaction effect term is smaller than 1.00 (odds ratio = 0.70), meaning that the negative impact of perceived government – expert discord is even more strongly negative among people who suffered more negative consequences during the pandemic.

The last two columns show that perceived government – expert discord has an odds ratio smaller than 1. Those who perceived government – expert discord were less likely to see the timing of lifting the control measures as appropriate. However, the corresponding logistic regression coefficient was not significant. Nonetheless, the interaction between perceived government – expert discord and negative impact of COVID on self is statistically significant and in the expected direction (odds ratio = 0.86).

Analysis using PROCESS MACRO was done to interpret and check the robustness of the interaction effects. The significant interaction effect on perceived appropriateness of the control measures’ strictness was retained. The impact of perceived government – expert discord is significant for people who scored above −0.507 on negative influence of COVID (66.5% of the respondents fulfilled this condition). Meanwhile, the interaction effect on perceived appropriateness of the timing of lifting the control measures and the interaction effect on credit attribution become non-significant (p = .12 and .11 respectively). Overall, the findings provide good support for H1 to H2, but only partial support for H3.

6.3. Predicting overall government performance evaluation

Multiple regression analyses were conducted to examine the last hypothesis. We first tried to predict people’s evaluation of COVID policies by using the regression models in Table 2. We then added credit-and-blame attribution and perceived appropriateness of the control measures into the full model. As the first column of Table 3 shows, perceived government – expert discord relates significantly negatively to the dependent variable, whereas perceived COVID information confusion, general confusion about health information, and exposure to COVID misinformation do not. Perceived government – expert discord stands out as the variable with a significant relationship with policy evaluation. Meanwhile, the interaction term does not have a significant coefficient when added.

Table 3. Predicting overall policy evaluation.

	Model 1	Model 2	Model 3
Sex	.09**	.09**	.07*
Age	.16***	.16***	.10**
Education	−.11**	−.11**	−.12**
Income	.03	.03	.03
Attention to health news	.06	.06	.02
Active information seeking	−.02	−.02	−.02
Pro-opposition	−.21***	−.21***	−.10***
Pro-government	.13***	.13***	.09**
Negative impact of COVID	−.04	−.03	−.01
General health info. confusion	.00	.00	−.03
COVID misinformation	−.05	−.05	−.05
Perceived COVID information confusion	−.03	−.03	−.02
Perceived gov-expert discord	−.12**	−.11**	−.04
Perceived gov-expert discord X negative impact of COVID		−.02	.01
Credit the government			.08**
Blame the government			−.17***
Strictness appropriate			.18***
Timing appropriate			.22***
Adjusted R^2	0.228***	0.228***	0.389***

Entries are standardized regression coefficients. Missing values in income were replaced by the mean. Missing values for other variables were deleted listwise. N = 975; ***p < 0.001; **p < 0.01; *p < 0.05.

When credit-and-blame attribution and perceived appropriateness of the control measures are added, all four variables significantly relate to the dependent variables: people who credited the government for positive outcomes and who regarded the control measures as appropriate evaluated COVID policies more positively, whereas people who blamed the government for negative outcomes evaluated the policies more negatively. After controlling for the four variables, perceived government – expert discord no longer relates to the dependent variable (i.e. the relationship is completely mediated by the four variables).

A SEM analysis was conducted to check if each of the four factors was a significant mediator.⁵ The results show that perceived government-expert discord has significant indirect effects on policy evaluation via blame attribution (indirect effect = .024, p < .01), perceived appropriateness of strictness (indirect effect = .023, p < .01), and perceived appropriateness of timing (indirect effect = .012, p < .05), but not credit attribution. In SEM, the total effect of perceived government-expert discord on overall policy evaluation is 0.109 (p < .001).

7. Discussion

This article attempts to understand how the public retrospectively evaluates government COVID policies. The conceptual focus is on the impact of government – expert discord. Experts generally gained trust as a result of the pandemic (Motta & Benegal, 2023), and various governments around the world tried to draw upon the legitimacy of medical experts when promoting policies (Weingart et al., 2022). However, there were occasions when officials and experts disagreed. As expected, citizens who perceived significant government – expert discord are likelier to blame the government for negative happenings, less likely to credit the government for positive outcomes, and less likely to see the pandemic control measures as appropriate. Credit-and-blame attribution and perceived appropriateness of the control measures in turn shape overall policy evaluation.

Perceived government – expert discord is highly correlated with COVID information confusion and exposure to COVID misinformation. During the pandemic, people often struggled to handle the deluge of COVID-related information. This has led to problems of information overload, fatigue, and frustration (Hwang et al., 2022; Mao et al., 2022). However, when the variables are controlled against each other, perceived government – expert discord stands out as the factor that leads to negative policy evaluation. In contrast, people who find health information generally confusing are less likely to blame the government for negative happenings and likelier to see the control measures as appropriately strict. A plausible interpretation is that people who find all kinds of health-related information confusing may tend to rely more on the authorities. In any case, not all kinds of confusion lead people to evaluate the government negatively, but the perception of government – expert discord consistently does.

This study fills a gap in the literature, as previous studies have not directly examined the extent to which people perceive the presence of disagreement between government officials and medical experts and the influence of such perceptions. Previous studies have mainly focused on the differential degrees of trust and influence enjoyed by politicians and experts (Barcelo et al., 2023; Cole et al., 2022). While government leaders might be perceived as having political motivations, experts are likelier to be seen as neutral. This differential perception is the basis for this study to expect that people would tend to evaluate the government negatively when its views and deeds deviate from those of the expert community. Support for the basic arguments thus helps extend the current literature.

In the Hong Kong case, the differential levels of credibility enjoyed by government officials and medical experts were exacerbated by the political context. The pandemic occurred at a moment when people’s trust in the government was low. The extent of autonomy of the Hong Kong Government has long been a matter of concern. Throughout the pandemic, there were questions regarding whether the government was allowed to adopt policies that deviated from the mainland’s. These contextual factors mean that the negative influence of perceived government – expert discord may be particularly strong in Hong Kong.

However, respect for experts during the pandemic was a worldwide phenomenon, and government-expert discord is also not unique to Hong Kong. The hypotheses and arguments were constructed based on research around the world. Perceived government – expert discord can be expected to have similar consequences in other contexts. Certainly, instead of merely replicating the finding in other contexts, it would be even more meaningful to develop and test arguments about the contextual factors that might shape the influence of perceived government – expert discord. For instance, one might examine if the degree to which medical experts are incorporated into governmental decision-making bodies would affect perceived government-expert discord and its impact.

Another contribution of this article resides in the specification of credit-and-blame attribution and perceived appropriateness of the pandemic control measures as the immediate predictors of overall policy evaluation. During the pandemic, governments struggled to strike a balance between preventing an uncontrolled outbreak and reducing the negative impact of stringent control measures. Stringent measures may inadvertently lead to more negative evaluations of the government if people’s risk assessments are not aligned with those of the government’s (Lee, 2022). Meanwhile, credit-and-blame attribution determines whether government popularity would benefit (or suffer) from positive (or negative) social and economic conditions (Fiorina et al., 2003; Rudolph, 2003). Identifying these as mediators helps us better specify how and why certain factors can shape policy evaluation.

A few limitations and possibilities for future analyses should be noted. First, the study was conducted after the end of the pandemic. It thus measures retrospective policy evaluation and judgement of government-expert discord. One may wonder how reliable people’s recalls are. However, the pandemic lasted for three years in Hong Kong, and government-expert discord did not occur only once. Even in September 2022, the media were writing ‘explanatory articles’ focusing on why the government and medical experts did not always agree with each other (Heung, 2022). In other words, the study did not ask respondents to recall an individual incident that occurred a long time ago. The demand on the respondents should not be seen as extraordinarily heavy. Besides, regardless of their bases, if citizens do form specific perceptions and retrospective evaluations, these are important as they can have real consequences. Hence the results of the study should be meaningful.

Second, some of the key variables in this study, including perceived government – expert discord, were measured only by a single item. On the one hand, support for the hypotheses is itself an indicator of the measure’s utility for the present purpose. Nevertheless, future research could try to develop better measures.

Third, while part of the arguments in this study concerns differential levels of credibility people placed on medical experts and the government, the study does not measure trust in the government and the expert community directly. Similar to partisanship, basic trust in the government might shape the degree to which people perceive government-expert discord. Nevertheless, the present study shows that perceived government-expert discord during the COVID pandemic was primarily a part of a general information confusion about COVID. Basic political trust is unlikely to completely explain away the relationship between perceived government-expert discord and the other variables. Meanwhile, a more specific notion of trust in the government’s information might be a consequence of perceived government-expert discord, i.e. when people perceive the government and the experts as disagreeing, they might lose confidence in the information provided by the government. In this sense, adding trust in government information may further elucidate why and how perceived government-expert discord matters. Future analysis may examine how basic political trust and trust specifically in governmental information matter.

Fourth, while this study focuses on people’s retrospective evaluations of government policies, the significance of studying such retrospective evaluations is partly premised on how the COVID experience might influence other policy attitudes after the pandemic. Hong Kong residents’ experience of SARS in 2003 conditioned their response to COVID-19 (S. Yuen et al., 2021), and research has started to show how attitudes towards experts developed during the pandemic may spill over into other arenas (Motta & Benegal, 2023). It would be meaningful to see if the perceptions and attitudes developed during COVID would have a lasting influence on people’s attitudes towards other government policies.

8. Conclusion

To conclude, this article has shown the importance of alignment between the government and medical experts for the government to gain public approval. An obvious implication is for governments to seriously incorporate expert advice into decision making. This may involve considerations regarding the roles to be played by experts in governing and/or consultation bodies. Besides, when public opinion is concerned, when expert advice is incorporated into policy decisions, it is advisable for governments to make the incorporation of expert advice conspicuous when communicating with the public, such as by inviting them to help explain and speak on behalf of government policies.

Certainly, governments needed to deal with not only issues of health but also issues of citizen rights and the economy during the pandemic. That is, governments need to consider matters that medical experts do not consider. Discrepancies between government policy decisions and expert advice might be inevitable in some cases. However, even in cases where a government cannot simply adopt expert advice, it remains important for officials to maintain proper communication with the expert community so that the latter is adequately informed of the rationale behind the policies. This may, again, depend on the degree to which the government can institutionalize the participation of experts in the policy-making process. In any case, obtaining the expert community’s understanding could prevent the overt and strong disagreement expressed by the expert community. This could help alleviate the (perception of) government – expert discord, which in turn can boost people’s confidence and improve the effectiveness of policies.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/17516234.2024.2355778

Additional information

Notes on contributors

Francis L. F. Lee

Francis L. F. Lee is Professor at the School of Journalism and Communication, Chinese University of Hong Kong, whereas Annisa Lee is Associate Professor.

Annisa Lee

Annisa Lee is Associate Professor at the School of Journalism and Communication, Chinese University of Hong Kong.

Notes

1. https://www.pori.hk/pop-poll/government-en/k001.html?lang=en

2. The RR3 formula took part of the calls with unknown eligibility into the denominator.

3. There were concerns about the quality of survey responses in contemporary Hong Kong. However, research about self-censorship in authoritarian contexts have argued that the degree of preference falsification is often exaggerated (Shen & Truex, 2020). In the present case, a degree of preference falsification may weaken the predictive power of the partisanship measures, but it should not introduce systematic bias to the findings.

4. The interaction term was centred around means to reduce multicollinearity. The largest variance inflation factor is only 2.35 for the regression models.

5. Since the interaction effect term has a close-to-zero coefficient when predicting policy evaluation, it was removed when running the SEM to simplify the model.