Differential media exposure and perceptions of fear and behavior change in China and Israel during the COVID-19 pandemic

Abstract

Understanding the media effect on behavioral outcomes is critical during a health crisis. Mechanisms explaining the mediation effect of media induced risk perception on individual fears and preventive behavior adoption rarely attend to the assumptions and methods to make a causal inference, nor to explore how the effect differs by socioeconomic status. We applied a causal framework to estimate how differential media exposure motivates fear and behavioral reaction, and to what extent these effects can be explained by risk susceptibility and risk severity perceptions in Israel and China, and whether the effects are conditional on socioeconomic situation. Our results suggest that media consumptions are explanatory predictors for increased fear and behavior through provoked risk perceptions. Moreover, socioeconomic status is a pronounced moderator in differentiating the media effect. These findings emphasize the media effects in the context of pandemic and have potential implications for media campaign and policy making.

Keywords

Health behavior media effect moderation effect risk perception

Beginning in 2020, China saw an uncontrollable increase in reported COVID-19 cases. Individuals were seeking for reliable information and increasingly resorted on both conventional media and social media (Sun et al., 2021).

According to Chinese national statistics, the average time that households watch television per day was 5.85 hours, with news receiving the highest overall rating (Pengpai News, 2021). In contrast, the radio ratings (Li and Zhang, 2020) and the newspaper circulation (National Press and Publication Administration, 2021) both declined in 2020, besides pandemic-relevant radio programs that rose by 24.5% in February. Considering that there have been a rising number of Chinese netizens in recent years, with 904 million as of March 2020 (China Internet Network Information Center, 2020), social media become an important venue for information-seeking, notably in the search for health information (Di et al., 2015). Keyword searches about the outbreak (Wang et al., 2022) and debates in one of the most popular social media, Weibo, in China (Zhao et al., 2020) surged.

In Israel, media behavior has drastically changed in the digital era. Traditional media outlets, including television, radio, and print, are struggling for financial sustainability and business strategies (Dorot, 2020). The penetration rate of social media, on the other hand, has been rising over the last 3 years and reached 90% in 2022. Among them, 75% of users regularly use social media, such as the most popular platforms, YouTube, Facebook, and Instagram in Israel, for information consumption. Television news consumption went second for media use, and both of them far outpaced the use of radio and print (Altshuler, 2020). Following the discovery of the first case on 21 February 2020, the Israeli Ministry of Health issued the National Information Center (Israel Ministry of Health, 2020) and updated statistics, guidelines, and precautionary measures. That information has been compiled into texts, pictures, and videos and disseminated through all kinds of media outlets. Seized by overloaded information from media, the majority increased their news consumption compared to normal (Bodas et al., 2015) and reported media fatigue shortly at the first month of the outbreak (Solomon, 2022).

Media play an essential role in the social reaction to a pandemic. Amid crises with uncertainties and threats, people increase their reliance on media and put them as a primary source of information (Allan, 2002; Ball-Rokeach and DeFleur, 1976). Users search actively for information and are passively exposed to countless pandemic contents (e.g. statistics, policies, health practitioner’s advice and instructions) in diverse formats (e.g. text, video, graphs) through different media outlets (e.g. television, radio, newspapers, websites, and social media). Meanwhile, they also use media platforms to share personal experiences, thoughts, and behaviors. Such exposures trigger public’s cognitive and emotional changes and provoke particular responses (Zhang et al., 2015). These effects have been coined as “media effects” to define the deliberate and non-deliberate short- and long-term within-person changes in cognition, emotions, attitudes, beliefs, and behavior that results from media exposure (Potter, 2012).

Media exposure and its downstream effects on the mental and behavioral health have called wide attention from the academic research (Garfin et al., 2020). While prior fundamental studies provided us with an enlightening black-box view of the phenomenon, they are unable to explain how and why such associations arise. Thus, the main purpose of this study is to understand how risk evaluations informed by media exposure influence behavioral outcomes of the pandemic. A unique contribution of this study is that we expand previous studies by measuring many media characteristic and categorize them into four kinds of media exposure, daily exposure, media types, media length, and media contents, to understand the role of media on perceived risk. We assess the fear and health behavior adoption in terms of pandemic outcomes, which have not been fully estimated as the outcome of various media exposure simultaneously in the crisis of Covid-19. We also aim to find out whether the media causal effect differs by socioeconomic status (SES). If that is true, then we will assess how groups from different SES differ in pandemic-related behavioral outcomes.

Literature review

Media exposure in the pandemic

Daily exposure

Research links media exposure enhancement to more significant mental and behavioral health reactions (Bonanno et al., 2010), because after receiving repeated broadcasting of global news about the virus day and night, individuals tend to overestimate the threats to their local communities (Garfin et al., 2020). For example, a study of the Boston Marathon bombings found that the psychological suffering for people with the heaviest media exposure was higher than that for those directly involved in the scenario (Holman et al., 2014).

Media types

In addition to the amount of media exposure, media types and frequency of usage also matter to individuals’ attitudes and behavior responses (Bonanno et al., 2010; Gao et al., 2020). Legacy media, like newspapers, television, and radio, has long been providing high news coverage. The use of digital media (websites, social media) provides access to additional sources of information and even have gradually nibbling the preponderance of the legacy media. By virtue of its convenience and the prevalence of personal computers and cell phones, many people use digital-based media for pandemic information and communication with others. Thus, media behaviors shifted. For example, 12 years ago, people preferred newspapers and magazines over radio and television for crisis and routine information (Avery, 2010). While during the H1N1 crisis, social media and mobile phones became main sources for news (Hu and Zhang, 2014). During the Covid-19 pandemic, social media was identified to play an essential role in news access and communication, and research found that people using them as primary sources of information were more likely to report mental health problems (Bendau et al., 2021). Nevertheless, some studies (Casero-Ripollés, 2020) claimed that legacy media had regained their predominance during this outbreak as shown by high consumption levels and perceived credibility. This finding has been confirmed by another study (Allington et al., 2021) that legacy and social media impact the behavior following opposite trajectories: it reported a negative relationship between health-protective behaviors and the use of social media and a positive relationship between these behaviors and the use of legacy media (TV and radio).

Besides legacy media and digital media, interpersonal communication also has a vital role in crisis management. Studies showed that interpersonal communication with professionals was superior in efficiency on risk awareness over diseases (Brashers et al., 2004; Morton and Duck, 2001) and health behavior (Coe et al., 2012; Seo and Matsaganis, 2013). Compared with non-impersonal, interpersonal communication especially with trustful and skilled experts or medical professionals were seriously taken into consideration by the public when it comes to awareness of the health crisis, things to know about the disease and its prevention, as well as the adoption of new behaviors (Simba and Kakoko, 2012).

Media message length

Another change in media behavior in the digital era is that people tend to consume information with a short length, and these characterized as browsing and scanning, keyword spotting, and less time on in-depth and concentrated reading (Liu, 2005). Given information-overloaded, people may prefer shorter news. They tend to achieve higher efficiency from their media consumption by absorbing as much information in a given time. This preference might influence people’s focus and their physiological arousal (Lang et al., 2005), and lead to a lack of information and misinformation, consequently escalating fear (Taylor et al., 2004). To investigate this possibility, in this study, we measure the message length preference and investigate its role on perceived risk and the adoption of preventive behaviors.

Media content

People’s media exposure has largely been accounted by how media producers frame the contents. In terms of public health emergencies, framing mainly identified the detected cases, governmental policies (Shih et al., 2008), sensationalism with substantial coverage of horrible narratives (Dudo et al., 2007), and risk descriptions by credible sources (Berry et al., 2007). It is likely that differences in exposure will influence peoples’ perception of the information and lead to behavioral changes differently (Price et al., 1997). Overwhelmed by high volume of message, the disposition-content-congruency hypothesis posits that, individuals tend to seek out media contents that do not diverge too much from their pre-existing cognitions, emotions, attitudes, beliefs, and behavior (Klapper, 1960). These dispositional congruent contents can be processed fluently by individuals (Alter and Oppenheimer, 2009) and are more likely to lead to media effects (Klapper, 1960). In this study, we investigate what kinds of contents out of constellation of messages attract individuals’ attention, and how these contents influence them.

Risk perception mediation effect

Scholars have conjectured and assessed many mechanisms to explain how people process information. Models from prior studies like selective perception (Klapper, 1960), internal states (Anderson and Bushman, 2002), reception-activity orientation (Bryant and Oliver, 2009), exposure status (Potter, 2009), and message processing (Petty et al., 2009) introduced the mediation effect of the mental and physiological process as a causal link connecting exposure and outcome. Significantly, the differential susceptibility to media effects model (Valkenburg and Peter, 2013) points out that media response states such as subjective risk appraisal are significant mediators that originate from media use and affect the consequence.

Risk susceptibility and severity are key appraisal predictors of health consequences in the health belief model (Rosenstock, 1974a). They measure how likely individuals perceive they will be affected by infection and how severe they perceive the personal consequences of the infection (Rosenstock, 1974a). In some mechanisms, they have been conceptualized as explanatory mediators between independent variables and fear (Jackson, 2009). Due to the threat fueled by media exposure (Taylor et al., 2004), perception of vulnerability to disease and anxiety proneness could relate to fear from contagion (Bakshi et al., 2021).

Additionally, threat exposure could also have constructive meaning for motivating people to conduct health behaviors mediated by risk perception (Jones et al., 2015). A meta-analysis research has shown that triggered by both perceived susceptibility and severity, people have a significant effect on behavioral intention for flu vaccination (Brewer et al., 2007). And empirical evidence from Covid-19 studies also proved that information cues received by traditional media, social media, and interpersonal communication with health professionals resulted in different compliance of restriction behaviors because of risk perception (Ranjit et al., 2021; Zeballos Rivas et al., 2021).

Media effect on fear in the pandemic

Fear was posited as a primary emotion for people in response to a decrease in the sense of power (Kemper, 1987); it is universally experienced by individuals and provides the basis for various more complex mental health problems (Turner, 2014). Thus, better understanding of the antecedents and consequences of perceived fear of contagion of the disease is a critical foundation to elucidate other mental health outcomes during the crisis (Bader et al., 2020).

Psychopathologists summarized that public generally perceive four domains of fear: fear for the body, for significant others, fear of not knowing, and inaction (Schimmenti et al., 2020). Polls from the United States in late January 2020 (Aubrey, 2020) and Canada in early February 2020 (Kurl and Korzinski, 2020) revealed that threats from virus’ novelty and uncertainties raised people’s worry on likelihood of infection and its consequence. At the same time, government and health care system’s ability to tackle the outbreak, lack of treatment, financial and employment difficulties increased people’s stress and fuel an individual’s total fear of the pandemic. Empirical research had widely supported that excessive exposure to media was associated with greater fear and poor mental health from previous crisis events, such as the Ebola outbreak (Thompson et al., 2017), the Great East Japan earthquake (Nishi et al., 2012), H1N1 (Zhang et al., 2015), and Covid-19 (Gao et al., 2020; Sasaki et al., 2020). Research to date, however, has not tested how does the individuals’ fear for the pandemic surrounding and its consequences triggered by risk perception on the base of media exposure.

Media effect on behavior in pandemic

In addition to adverse mental health, health-related behavior can also be derived from the messages conveyed in the media (Wakefield et al., 2010). Some research has found that increased exposure to media about pandemic information was related to higher avoidance of preventive measures adoption, such as social distancing and washing hands with soap (Melki et al., 2022; Scopelliti et al., 2021), volunteer self-management of health behavior, such as physical activity (Goodyear et al., 2021). Other research, nevertheless, suggested that heightened psychological response resulting from repeated media exposure could bring disproportionate or overreacting behaviors in response to actual threats (Garfin et al., 2020), such as panic buying, and then overburden critical healthcare facilities and fuel the panic climate in society.

Theoretically, risk perception involves these decision-making processes (Rosenstock, 1974b), while the explanatory effectiveness of its constructs in empirical research regarding health behavior is inconsistent (Carpenter, 2010; Janz and Becker, 1984). For example, higher social media uses in the MERS Outbreak research in South Korea (Choi et al., 2017), Covid-19 Pandemic in Saudi Arabia (Angawi and Albugmi, 2022) and China (Liu et al., 2020) all reported increased risk perceptions. While they failed to find an association with behavioral response, they believed that media effect is essential in the preventive behavior research. Results from a Bolivian research on Covid-19 information (Zeballos Rivas et al., 2021) supported the positive association between social media exposure and risk perception, as well as the positive association between risk perception and preventive attitudes and behaviors combined, yet the path of how media influence behaviors through risk perception still need further elaboration.

Socioeconomic moderation effect

Evidence proved that social factors influence how people access, process, and respond to media information about health topics (Galarce et al., 2011), potentially leading to differential emotional arousals among groups with different SES. This communication inequality (Viswanath, 2006) may eventually drive disparities in health outcomes. Studies unveiled that people with higher levels of income and education can benefit from media information about health issues compared with their lower SES counterparts (Katapodi et al., 2004; Viswanath, 2005, 2006). However, research to date has not yet fully addressed the potential roles of SES inequality in the media effect, especially in the context of the pandemic.

Conceptual model

Scholars have elaborated on these mediating and moderating mechanisms in various situations, while few have considered the assumption and methodology for causal inference and sensitivity analysis. Moreover, it is valuable to learn how this mechanism functions in different subpopulations with various social capitals. Thus, we included participants from two countries: Israel and China, because of their differences in the outbreak status, cultural backgrounds, and population dispositions to provide empirical insights to these mechanisms. We also applied SES as an axis of variance to explore the difference within subpopulations.

As enlightened above, we proposed a conceptual model (Figure 1) to investigate how people in Israel and China perceive risks from media exposure, and in turn to influence their mental health and behavioral changes, and how these effects are conditional on SES. The findings should help us to better understand the mechanisms between media exposure and mental and behavioral responses during the pandemic contexts.

In this study, the central question is whether media exposure leads to higher levels of fear or higher behavior adoption by increasing participants’ risk perception. The reader should keep in mind that in the hypothesis when we refer to media exposure, we refer to the multiple measures that are included in the analysis (i.e. daily exposure, media types, media length, and media contents). Based on our proposed model and the literature reviewed, we would like to test following hypotheses:

H1. Media exposure is positively associated with (1) individual fear and (2) preventive behavioral adoption.

H2. Risk susceptibility mediates the association between (1) media exposure and fear; (2) media exposure and preventive behavioral adoption.

H3. Risk severity mediates the association between (1) media exposure and fear; (2) media exposure and preventive behavioral adoption.

Regarding our mediation expectation, our research questions will be

RQ1. How large the sensitivity parameter ρ would be for the mediation effect to be zero?

RQ2. How does the socioeconomics affect media mediation effect?

Methodology

Measurement

Under the research protocol, participants older than 14 years old with reading literacy were recruited between March and April 2020 from two professional online research firms in Israel and China. We interlocked the sex and geographical quotas in both nations and utilized a proportional quota sampling method to guarantee a wide representation of respondents. After data cleaning, answers from a total of 1449 respondents were valid for analysis.

The measurements were adapted by two publications (Leung et al., 2003; Liu et al., 2016). The questionnaire was originally in English and its validity was determined by a professor from social science discipline. After being translated into Hebrew and Mandarin, the readability and clarity of items were tested by two doctorate students from their native languages. Modifications on the questionnaire were made if respondents find unclear about the questions during the pilot study.

Media exposure

In this study, media exposure variables are independent variables that included (1) duration of daily exposure to pandemic information: an item that asked respondents to indicate how many hours per day they had been following the pandemic information; (2) frequency of old media use (newspaper, television, radio), digital media use (website and social media), and other (medical professional consulting) to access to pandemic information, 1 = infrequently use, 2 = sometimes use, 3 = frequently use, 4 = always use; (3) preference for information length (short, medium, long, extra-long) of text and video message; and (4) following or not certain contents about pandemic (statistics, narrative, policy, and behavior instructions).

Risk perception

The risk perception variables are mediation variables (mediators) as media response states that help to figure out how independent variables influence the outcome variables in an indirect approach. We measured two risk perceptions: risk susceptibility perception (Do you think you are susceptible to being infected with the coronavirus after receiving information from media?) and risk severity perception (Do you think getting coronavirus is a severe risk for your health after reading information from media?). We used a 5-point scale from “strongly disagree” to “strongly agree” to collect participants’ perception levels for each media usage.

Fear and behavior

Fear and protective behavior are two outcome variables of how people score the consequences after exposure through a specific mechanism. We measured two kinds of fears: personal fear (How afraid you are to get the coronavirus?) and altruistic fear (How afraid you are that one of your close family members and friends will contract the coronavirus?). Each fear indicator was scored on a 4-item scale from “1 = not at all” to “4 = highly afraid,” then we summed them up as individual total score (Cronbach’s alpha = .82). The behavior adoption measures individuals’ compliance with preventive behavior that officials and media recommend. We included 19 behavioral items, and participants needed to choose the frequency of conducting each behavior by a 5-item scale from “1 = never” to “5 = always.” We added scores on the 19 items together as the total score of individual behavior adoption (Cronbach’s alpha = .88).

Socioeconomic status

The socioeconomic score is the moderation variable (moderator) to study the contingence of how the exposure variable affects the outcome variable between participants with different social status. We considered two indicators: income and education. Income measured the level of personal monthly income by a national standard. It ranges from “1 = significantly under average” to “5 = significantly higher than average.” A 7-option measured the education from “1 = No schooling completed” to “7 = graduate completed.” The total score of SES was calculated by summing up the scores of two indicators. Participants with higher scores are more socioeconomically advantaged.

Controlling variable

To control the confounding of the effect, we included six indicators as covariates: sex, age, family status, diagnosis with coronavirus, diagnosis with other diseases, and isolation.

Analytical strategy

This study estimated three kinds of effects: total effect, direct effect, and mediation effect (indirect effect). The total effect quantifies the overall effect of media exposure on the hypothesized outcome variable; the mediation effect is the portion of the total effect that transmitted through risk perceptions; and the direct effect represents all other possible causal mechanisms besides our hypothesized mediators that explain why media exposure lead to fear and behavior adoption.

In order to understand the whole picture of the causality, we applied the potential outcome framework (Holland, 1986) for total effect and added the assumption of sequential ignorability (Imai et al., 2010a) for direct effect and mediation effect. To bolster the confidence of the assumption, we included pre-exposure covariates which may confound the relationship between media use and outcome variables in the estimation. Thus, with observative data collected, we assumed that sequential ignorability was approximately true conditioning on all our measured covariates. We also conducted a sensitivity analysis to quantify the robustness of this assumption-based empirical finding.

We define $i$ as an object that we observe. $E_{i}$ denotes the exposure variable where $E_{i}$ = 1 implies object $i$ had been exposed to targeted media, and $E_{i}$ = 0 implies it did not. The $M_{i}$ denotes the mediator, and $M_{i}$ [e] denotes the potential value of the mediator for object $i$ under the exposure status of $E_{i}$ = e. We use $Y_{i}$ to denote the outcome variable, then $Y_{i} [e]$ denotes the potential outcome that $i$ receives the exposure of e. For example, $Y_{i}$ [1] represents the fear outcome for respondent $i$ with exposure of statistics following, and $Y_{i} [0]$ represents the fear outcome for the same respondent without following. With this, $Y_{i} [e, m]$ denotes the potential outcome of object $i$ with exposure status e and mediator m.

Given that a single individual cannot be both exposed and non-exposed to a specific exposure level simultaneously, we cannot quantify the individual-level causal effect of media exposure on the outcome variable. Therefore, we use the group level to calculate the average causal effect.

The total effect can be defined as: ${ToE}_{media} = Y_{i} [1] - Y_{i} [0]$ or ${ToE}_{media} = Y_{i} [$ 1, $M_{i} (1)$ ] $- Y_{i}$ [0, $M_{i} (0)$ ]. It identifies the observed difference in mean outcomes between the exposed and the non-exposed groups.

And the direct effect as: ${DiE}_{media} = Y_{i} [1$ , $M_{i} (e)$ ] $- Y_{i} [0$ , $M_{i} (e)$ ]. It identifies the causal effect of media exposure on fear and behavior while holding on to the mediators of risk perception.

The mediation effect as: $M e E_{media} = Y_{i}$ [e, $M_{i} (1)$ ] $- Y_{i}$ [e, $M_{i} (0)$ ]. It answers the counterfactual questions: what would the $M e E_{media}$ changes for a fitting E(e) if the $M$ (1) changes to $M$ (0). This means if the mediator change from what would be realized under the exposed condition $M$ (1) to what would be observed under the non-exposed condition $M$ (0) while holding the exposure status as e. In the current study, one of these scenarios is that if an object did not follow statistics content, when his severity perception changes from high to low, then what would happen to his fear?

The proportion of mediation effect in total effect is: %mediated = $M e E_{media} / {ToE}_{media}$ .

Estimation

Following the algorithm (Imai et al., 2010a), we first regressed the mediators on the independent variables and the covariates (equation (1)). We then regressed the outcome variables on the media variables, the mediators, and the covariates (equation (2))

M_{i} = β_{m 1} E_{i} + \sum_{i = 1}^{i} (β_{m 2} U_{i}) + K_{m} + ε_{i m}

(1)

Y_{i} = β_{y 1} E_{i} + β_{y 2} M_{i} + \sum_{i = 1}^{i} (β_{y 3} U_{i}) + K_{m} + ε_{i m}

(2)

In this study, the E indicates the media variables, M indicates the risk perception, Y indicates the fear scores and scores of behaviors adoption, U the covariates, K the constant, and ε the error.

Then, we simulated model parameters for equations (1) and (2) from sampling distribution. Based on equation (1), we also stimulated the potential values of the mediators for each object. We generated multiple sets of predicted mediator values for each object under different exposure levels and compared the effects of mediators at the observed level, for example, $M$ [ $E_{i} (1)$ ] and each other levels, for example, $M$ [ $E_{i} (0)$ ], on the outcome for each exposure level E(e = 0, 1, . . ., e-1).

After that, we used equation (2) to impute the potential outcome prediction. First, under a specific exposure condition, we obtained the predicted outcome value and predicted mediator value, for example, $Y_{i}$ [ $E_{i} (1)$ , $M_{i} (1)$ ]. We then calculated the predicted counterfactual outcome, that is, where the exposure indicator is still set to e = 1, but the mediator is set to one of the other predicted values under the other exposure level that obtained in last calculation, that is, $Y_{i}$ [ $E_{i} (1)$ , $M_{i} (0)$ ].

Therefore, we can compute the total effect, direct effect, and mediation effect by averaging the difference among the predicted outcome under the values of the mediators across objects.

Finally, we tested its statistical significance by employing a 5000-time sampling bias-corrected bootstrapping under the quasi-Bayesian Monte Carlo approximation (King et al., 2000) and reported the (average) point estimate and confidence interval.

Sensitivity analysis

The estimation reported via the mediation mechanism may be biased if the sequential ignorability assumption does not hold. Even though we adjusted several confounding variables by regressing them in models as covariates, more unobserved confounders can still affect both the mediators and the outcome variables. The proposed sensitivity analysis (Imai et al., 2010b) based on the correlation between error $ε_{m}$ from equation (1) and $ε_{y}$ from equation (2) to quantify the robustness of the mediation estimates to the remaining unobserved confounding. Its parameter is denoted as ρ applying the iterative feasible generalized least-squares algorithm of the seemingly unrelated regression (Zellner, 1962). A smaller value of ρ suggests that the modeling causal mediation is more sensitive to the potential violation of the sequential ignorability assumption.

Moderation

To test the moderation effect of SES functioning on the mediation effect, we estimated an ordinary least squares (OLS) regression model predicting risk perception ( $M_{m o}$ ) from media exposure (E), SES (W), and their product (EW), and covariates ( $U_{i}$ ) as equation (3)

M_{m o} = α_{1} E + α_{2} W + α_{3} E W + \sum_{i = 1}^{i} (α_{4} U_{i}) + K_{w} + ε_{w}

(3)

Combined with equation (2), the quantification of the moderation effect of W on mediation is calculated by equation (4)

ω = (α_{1} + α_{3} W) β_{y 2} = α_{1} β_{y 2} + α_{3} β_{y 2} W

It indicates a linear function of W with $α_{1} β_{y 2}$ as intercept and $α_{3} β_{y 2}$ as slope. The slope is the weight of how the mediation effect had been moderated by moderator W.

Current study hypothesized that a single moderator simultaneously and independently functioned on every media variable to each mediator and then each mediation effect was modeled as a linear function of every moderator. If the estimated index of the slope is different from zero, it indicates that the W moderates the mediation process. We used the pick-a-point approach (Preacher et al., 2006) and chose the mean and plus and minus one standard deviation from the mean of socioeconomic scores’ distribution as “middle,” “high,” and “low” groups by which to report point estimation. Since the index of a single calculation is subject to sampling variation, we conducted 5000 times bootstrap in the sampling distribution and reported the 95% bootstrap confidence interval for the slope of every group. If the confidence interval does not include zero, it statistically supports the significance of the moderated effect.

Results

Baseline characteristics and media consumption

We included 706 Israeli participants (male = 47.9%, female = 52.1%) and 733 Chinese participants (male = 51.4%, female = 48.6%). Among Israelis, most participants (93.2%) are less than 70 years old, and most Chinese participants (85.9%) are between 30 and 70 years old. Twenty participants had been diagnosed with coronavirus, in which 5 cases (0.7%) came from Israel and 15 cases (2%) came from China. At the same time, 127 Israeli participants (18%) and 72 Chinese participants (9.8%) reported being diagnosed with other diseases. In total, 77 Israeli participants (10.9%) and 160 Chinese participants (21.8%) were isolated when they filled out the questionnaire.

As shown in Figure 2, about 75% of Israeli participants and 50% of Chinese participants spent more than 1 hour a day on pandemic-related information consumption. In both countries, people widely use television, news websites, and social media for information. Chinese participants reported higher reliance on news websites and social media. Most participants preferred short and medium lengths of text and video information, and some Chinese preferred long videos when they learned about the pandemic. In terms of the contents they followed, more than 60% of Chinese participants were followers of all selected contents, while about 70% of Israeli participants did not follow statistics, policy, and behavior, and 96% did not follow narratives.

Direct effect and total effect of media exposure on fear and behavior

Online Appendices Table B presents the detailed results for the direct effects of media exposure on the mediating variables and on the outcome variables. When controlling the sex, age, family status, disease diagnosis, Covid-19 diagnosis, and isolation, we observed that Israeli participants’ susceptibility perceptions were positively related to higher daily exposure duration and more frequent television, radio, and website usage. Their severity perceptions were positively associated with a higher amount of daily exposure and more frequent legacy media utilizing. At the same time, Chinese participants who have longer daily exposure and frequent use of newspaper, radio, and medical consulting for pandemic information reported higher levels of susceptibility perception. On the contrary, those who followed preventive behaviors and policies reported decreased susceptibility perceptions. Their severity perception appeared to be positively associated with digital media, preference for video with a longer length, and following statistics, yet negatively associated with frequent use of newspapers. We also found statistically significant associations between media exposures and outcome variables from direct effect estimation.

We presented estimates for the total effects of media exposure on fear and behavior adoption in Table 1. It showed that Israelis’ fear was positively associated with daily exposure, more frequent use of television and news websites, and a preference for longer videos. Chinese respondent’s fear was observed to be positively associated with a broader spectrum of media exposure, including longer daily exposure, frequent use of almost all media types (except medical consulting), and preference for longer text and video messages. For participants from both countries, each higher unit in these media exposure was related to 0.13 to 0.34 units higher in fear. Meanwhile, Israelis’ behavior adoption was positively associated with daily exposure duration, frequent use of all legacy media and websites, preference for longer videos, and following media contents of statistics and behavior. Among which, one unit higher in frequent television use, preferring longer video and following behavioral information reported 2.51 to 3.33 units higher in preventive behavior adoptions. However, people who followed statistics reported about 2 units lower in fear. Chinese participants revealed higher behavior adoption by frequent use of digital media, preference for longer video messages, and following contents of statistics, policies, and behavioral information. Among these media consumptions, a higher frequent usage of digital media and following behavioral instruction revealed a stronger magnitudes association of 2–3 units higher in behavior adoption, whereas more frequent use of newspapers reported about 1 unit lower in preventive behavioral adoption.

Table 1.

Analysis exploring the media total effect ( $T o E_{m e d i a}$ ) of media exposure on outcome variables through mediators among participants.

Mediation	Israel fear		Israel behavior		China fear		China behavior
Mediation	Susceptibility	Severity	Susceptibility	Severity	Susceptibility	Severity	Susceptibility	Severity
Exposure time
Daily exposure	0.26^† [0.17, 0.36]	0.26^† [0.16, 0.36]	1.47^† [0.71, 2.25]	1.80^† [1.02, 2.61]	0.34^† [0.2, 0.48]	0.34^† [0.2, 0.49]	0.41 [−0.32, 1.16]	0.77 [−0.03, 1.62]
Frequencies of media type use
Newspaper	−0.06 [−0.16, 0.04]	−0.02 [−0.12, 0.08]	1.87^† [1.1, 2.66]	2.12^† [1.36, 2.92]	0.19^† [0.08, 0.3]	0.13^† [0.02, 0.25]	−0.87^† [−1.45, −0.27]	−1.17^† [−1.82, −0.5]
Television	0.22^† [0.12, 0.32]	0.23^† [0.14, 0.34]	2.51^† [1.77, 3.28]	2.87^† [2.12, 3.66]	0.27^† [0.14, 0.41]	0.29^† [0.15, 0.43]	0 [−0.71, 0.72]	0.35 [−0.43, 1.17]
Radio	0.08 [−0.01, 0.18]	0.08 [−0.02, 0.18]	1.60^† [0.86, 2.37]	1.85^† [1.1, 2.65]	0.16^† [0.05, 0.28]	0.15^† [0.03, 0.27]	−0.35 [−0.96, 0.26]	−0.22 [−0.88, 0.48]
Website	0.16^† [0.07, 0.26]	0.14^† [0.05, 0.25]	1.03^† [0.27, 1.81]	1.26^† [0.49, 2.08]	0.24^† [0.06, 0.42]	0.34^† [0.17, 0.52]	2.08^† [1.18, 2.99]	2.97^† [2.01, 3.98]
Social media	−0.02 [−0.12, 0.06]	−0.04 [−0.13, 0.05]	0.05 [−0.65, 0.76]	−0.07 [−0.77, 0.65]	0.16 [−0.01, 0.32]	0.27^† [0.11, 0.44]	1.72^† [0.89, 2.56]	2.71^† [1.83, 3.64]
Medical consulting	0 [−0.09, 0.09]	0.00 [−0.09, 0.09]	0.20 [−0.52, 0.95]	0.35 [−0.38, 1.12]	0.11 [−0.01, 0.24]	0.07 [−0.06, 0.19]	−0.02 [−0.66, 0.62]	−0.10 [−0.81, 0.63]
Preferences for media lengths
Text	0.26 [−0.05, 0.58]	0.21 [−0.1, 0.54]	1.50 [−0.95, 4.01]	1.55 [−0.92, 4.13]	0.15^† [0.04, 0.26]	0.14^†[0.03, 0.26]	0.42[−0.15, 1]	0.40 [−0.22, 1.06]
Video	0.2^† [0.04, 0.37]	0.22^† [0.05, 0.39]	2.56^† [1.27, 3.9]	2.87^† [1.57, 4.23]	0.28^† [0.15, 0.41]	0.31^† [0.19, 0.45]	1.31^† [0.64, 1.99]	1.82^† [1.09, 2.59]
Following media contents
Statistics	−0.05 [−0.27, 0.17]	−0.1 [−0.32, 0.12]	−1.88^† [−3.63, −0.12]	−2.25^† [−4, −0.44]	−0.16 [−0.42, 0.11]	0.02 [−0.24, 0.3]	1.4^† [0.02, 2.78]	2.86^† [1.39, 4.4]
Narratives	−0.29 [−0.78, 0.23]	−0.34 [−0.85, 0.19]	−0.54 [−4.56, 3.54]	−0.62 [−4.64, 3.57]	−0.03 [−0.27, 0.22]	0.03 [−0.21, 0.28]	0.29 [−0.98, 1.55]	0.79 [−0.59, 2.22]
Policies	0.04 [−0.17, 0.27]	0.05 [−0.16, 0.28]	−0.54 [−2.26, 1.2]	−0.37 [−2.08, 1.43]	−0.12 [−0.37, 0.14]	−0.03 [−0.28, 0.24]	1.9^† [0.58, 3.22]	2.16^† [0.72, 3.67]
Behaviors	0.15 [−0.06, 0.37]	0.18 [−0.03, 0.41]	2.93^† [1.22, 4.67]	3.33^† [1.64, 5.12]	0.02 [−0.23, 0.28]	0.11 [−0.14, 0.37]	2.91^† [1.64, 4.18]	2.85^† [1.45, 4.32]

“Israel Fear”: dependent variable is Israeli individual fear scores; “Israel Behavior”: dependent variable is Israeli individual behavior adoption scores; “China Fear”: dependent variable is Chinese individual fear scores; “China Behavior”: dependent variable is Chinese individual behavior adoption scores; “Mediation”: the mechanism by which the part of mediation effect functioning. Numbers in the brackets are 95% confidence interval calculated by 5000 times bias-corrected bootstrap.

†

The 95% confidence interval does not include zero.

Mediation effect of media exposure on fear

Table 2 reports estimates of the mediating effects of risk perception on the association between media exposure and fear.

Table 2.

Analysis exploring the media mediation effect ( $M e E_{m e d i a}$ ) of media exposure variables on fear through mediators.

	Israel				China
	${MeE}_{media}$ by susceptibility	% mediated	${MeE}_{media}$ by severity	% mediated	${MeE}_{media}$ by susceptibility	% mediated	${MeE}_{media}$ by severity	% mediated
Exposure time
Daily exposure	0.05^† [0.02, 0.08]	18.63	0.04^† [0.01, 0.08]	16.62	0.02^† [0, 0.05]	5.96	0.02 [−0.01, 0.07]	7.24
Frequencies of media type use
Newspaper	0.01 [−0.02, 0.04]	−10.31	0.04^† [0.01, 0.08]	−62.30	0.02^† [0, 0.03]	7.98	−0.04^† [−0.08, −0.01]	−29.87
Television	0.04^† [0.01, 0.07]	17.66	0.05^† [0.02, 0.09]	23.40	0.01 [0, 0.04]	5.02	0.03 [−0.01, 0.07]	9.65
Radio	0.04^† [0.01, 0.07]	41.17	0.03^† [0, 0.07]	39.77	0.02^† [0, 0.04]	9.56	0.01 [−0.03, 0.04]	3.93
Website	0.05^† [0.02, 0.08]	28.79	0.03 [0, 0.07]	20.04	−0.01 [−0.04, 0.01]	−5.49	0.09^† [0.04, 0.15]	26.91
Social media	0.00 [−0.03, 0.02]	6.11	−0.02 [−0.05, 0.01]	31.24	0.00 [−0.03, 0.02]	−2.11	0.10^† [0.05, 0.15]	36.73
Medical consulting	0.02 [−0.01, 0.05]	−22.69	0.02 [−0.01, 0.06]	−22.11	0.02^† [0.01, 0.05]	20.93	−0.02 [−0.06, 0.01]	−22.61
Preferences for media lengths
Text	0.04 [−0.04, 0.14]	15.83	0.00 [−0.1, 0.11]	−0.03	0.00 [−0.02, 0.02]	0.17	0.00 [−0.03, 0.03]	−1.33
Video	0.03 [−0.01, 0.09]	16.51	0.05 [−0.01, 0.11]	21.21	0.01 [−0.01, 0.03]	2.97	0.05^† [0.01, 0.09]	14.46
Following media contents
Statistics	−0.01 [−0.07, 0.06]	5.31	−0.06 [−0.14, 0.01]	39.04	−0.02 [−0.06, 0.02]	9.97	0.16^† [0.08, 0.25]	69.47
Narratives	0.04 [−0.11, 0.2]	−9.75	−0.02 [−0.19, 0.16]	4.49	0.00 [−0.04, 0.03]	1.61	0.05 [−0.02, 0.13]	25.02
Policies	0.02 [−0.04, 0.09]	11.48	0.03 [−0.05, 0.1]	16.29	−0.04^† [−0.09, −0.01]	24.66	0.05 [−0.02, 0.12]	−21.52
Behaviors	0.03 [−0.03, 0.1]	19.13	0.06 [−0.01, 0.15]	32.92	−0.07^† [−0.12, −0.02]	−31.33	0.02 [−0.05, 0.09]	11.10

Each cell shows an estimated coefficient of media mediation effect obtained by calculating the average causal mediation effect, the outcome variable is fear score. Numbers in the brackets are 95% confidence interval calculated by 5000 times bias-corrected bootstrap. Negative percentage in the “% mediated” indicates an opposite direction of mediation effect and total effect.

†

The 95% confidence interval does not include zero.

Among Israeli respondents, daily exposure and media types explained substantial portions of the mediation effects of risk perceptions on fear. Specifically, 0.05 units of fear (accounted for 18.63% of total fear) was related to heightened risk susceptibility caused by 1 unit higher in daily exposure. It also revealed that similar additional fears were associated with risk susceptibility caused by more frequent use of television, video, and website for pandemic information, and these fears accounted for 17.66%, 41.17%, and 28.79% of the total effects. At the same time, 16.62–62.30% of fear came from the increased severity perception because of a higher amount of daily media exposure and a higher frequency of legacy media exposure, and 1 unit higher in these exposures reported 0.03 or 0.05 units higher in fear through severity perception.

For Chinese respondents, daily exposure and media types are less critical in inducing fear through risk perception than that for Israeli participants. To be more specific, 1 unit higher in daily duration, frequent use of newspapers, radio, and consulting each heightened 0.02 units in fear through susceptibility and accounted for small portions of their total fear (5.96%, 9.56%, 20.93%). The followers of policies and behavioral instruction reported decreased fear than those who did not show any interest in these kinds of content. These 0.04 units (24.50% of total fears) and 0.07 units (29.64% of total fears) of weakened fears were due to lower susceptibility perception to the virus, even though their total effects on fear were not significant. Via severity perception, 0.04 units less in fear was observed in every unit higher in frequent newspaper reading, and this fear ease can offset 29.05% of total fear. Moreover, the more extended frequency of digital media use had fueled people’s severity perception, leading to about 0.1 units higher in fear (27.32–37.53% of total fear). And a longer video length preference and following in statistics also contributed to a 0.05 unit and 0.16 unit higher in fear because people gained a higher severity perception.

Mediation effect of media exposure on preventive behavior

Table 3 presents the estimation of the mediation effect of risk perception on individuals’ preventive behavior adoption.

Table 3.

Analysis exploring the media mediation effect ( $M e E_{m e d i a}$ ) of media exposure variables on preventive behavior adoption through mediators.

	Israel				China
	${MeE}_{media}$ by susceptibility	% mediated	${MeE}_{media}$ by severity	% mediated	${MeE}_{media}$ by susceptibility	% mediated	${MeE}_{media}$ by severity	% mediated
Exposure time
Daily exposure	−0.07 [−0.22, 0.05]	−5.00	0.26^† [0.06, 0.5]	14.29	−0.11^† [−0.24, −0.01]	−20.46	0.25 [−0.14, 0.67]	31.61
Frequencies of media type use
Newspaper	0.00 [−0.06, 0.04]	−0.24	0.24^† [0.06, 0.47]	11.24	−0.07^† [−0.17, 0]	8.35	−0.38^† [−0.69, −0.06]	32.28
Television	−0.05 [−0.16, 0.05]	−1.83	0.31^† [0.12, 0.55]	10.76	−0.07 [−0.18, 0.02]	6.59	0.28 [−0.1, 0.69]	50.62
Radio	−0.04 [−0.16, 0.04]	−2.79	0.20^† [0.01, 0.43]	10.99	−0.08^† [−0.18, 0]	16.97	0.06 [−0.26, 0.4]	−12.47
Website	−0.06 [−0.2, 0.05]	−6.07	0.17 [−0.02, 0.41]	13.78	0.05 [−0.05, 0.18]	2.54	0.94^† [0.48, 1.45]	31.78
Social media	0.00 [−0.04, 0.05]	0.64	−0.12 [−0.31, 0.06]	20.15	0.02 [−0.09, 0.12]	0.89	1.00^† [0.57, 1.48]	37.06
Medical consulting	−0.02 [−0.1, 0.03]	−3.74	0.13 [−0.05, 0.34]	24.98	−0.13^† [−0.25, −0.03]	19.96	−0.21 [−0.55, 0.15]	37.99
Preferences for media lengths
Text	−0.04 [−0.27, 0.1]	−2.25	0.00 [−0.64, 0.66]	0.00	0.00 [−0.08, 0.07]	−0.12	−0.02 [−0.32, 0.3]	−4.06
Video	−0.03 [−0.17, 0.06]	−1.30	0.27 [−0.05, 0.65]	9.43	−0.04 [−0.14, 0.05]	−3.01	0.47^† [0.11, 0.85]	25.89
Following media contents
Statistics	0.01 [−0.09, 0.13]	−0.43	−0.36 [−0.84, 0.08]	15.94	0.09 [−0.07, 0.3]	6.54	1.55^† [0.83, 2.35]	54.52
Narratives	−0.03 [−0.33, 0.19]	0.98	−0.11 [−1.15, 0.96]	3.48	0.02 [−0.15, 0.19]	1.85	0.51 [−0.16, 1.23]	49.71
Policies	−0.02 [−0.15, 0.08]	1.28	0.16 [−0.27, 0.63]	−12.35	0.20^† [0.03, 0.43]	10.29	0.46 [−0.24, 1.2]	21.24
Behaviors	−0.03 [−0.19, 0.07]	−1.02	0.38 [−0.04, 0.88]	11.29	0.25^† [0.05, 0.52]	8.73	0.20 [−0.48, 0.91]	6.92

Each cell shows an estimated coefficient of media mediation effect obtained by calculating the average causal mediation effect, the outcome variable is behavior adoption score. Numbers in the brackets are 95% confidence interval calculated by 5000 times bias-corrected bootstrap. Negative percentage in the “% mediated” indicates an opposite direction of mediation effect and total effect.

†

The 95% confidence interval does not include zero.

Media exposure was found to have a strong relationship with Israelis’ behavioral adoption, while most of their mechanisms were undetected. Through the mediation of severity perception, only a higher amount of daily exposure and higher frequency in legacy media exposure gave rise to behavioral adoption by 0.20–0.31 units and accounted for about tenth of total effects. Specifically, each higher unit’s daily exposure can lead to 0.26 units higher in behavioral adoption, and more frequent use of legacy media contributed to 0.24, 0.31, and 0.20 units higher in adoption.

While among Chinese participants, daily exposure and media types seemed to decrease behavior adoption. Via the mediation effect of susceptibility perception, more prolonged daily exposure, frequent use of newspaper, radio, and medical consulting were reported to lower the adoption by 0.08–0.13 units and accounted for 8.35–20.46% of the total effect. By contrast, following policies and behavioral messages reported 0.20 and 0.25 units higher adoption (accounted for about 10% of the total effect). Only more frequent use of newspapers reported decreased behavior adoption by 0.37 units via mediation effect of severity perception. Furthermore, the severity perception they gained from digital media, preference for longer video lengths, and following statistics had consistently translated to higher behavior adoption by about 1, 0.47, and 1.55 units, respectively, which account for 25.89–54.52% of the total effects.

Figure 3 illustrates the direct, total, and indirect effects of risk perception for media exposure on fear and behavior adoption.

Sensitivity analysis

In Figure 4, we displayed the results of the sensitivity analysis; it showed the sensitivity parameters for every mediation effect of media exposure variable.

We predicted that the mediation effect of severity perception for media exposure on behavior adoption had the highest robustness in the study with a sensitivity parameter ρ near 0.44, followed by its effect on fear (the mean of the ρ is 0.25) among Chinese participants. Among Israeli participants, the ρ-parameters of the severity perception mechanism for media exposure on fear (the mean of the ρ is 0.22) are slightly higher than that for the susceptibility mechanism (the mean of the ρ is 0.19). Parameters for the susceptibility perception mediation effects on behavior are low among Chinese and Israeli participants.

Moderation

Figure 5 depicts that SES as a moderation effect functioned on the mediation effects of how media variables influenced outcome variables through risk perception.

Discrepancies observed among the estimations of different SES groups indicated that the proposed mediation effects were moderated by SES, even though some estimates were marginally different from others. We found that among legacy media exposure, Israeli participants with lower SES reported a higher mediation effect than their higher SES counterparts, while Chinese participants with lower SES groups reported a lower mediation effect than those who are socioeconomically advantaged. With few exceptions, the moderation effects are similar for both countries among digital media exposure, preference for longer text and video information, and policies following. Participants with higher SES reported a higher mediation effect. On the contrary, among statistics and behavior instruction following, participants with lower SES reported a higher mediation effect.

Discussion

The current study proposed a conceptual model to explore the possible mechanisms that explained the link between media exposure and mental and behavioral outcomes. The proposed model contributes to the literature by applying causal mediation and moderation analyses to investigate up-to-date public health crises. Furthermore, it compares the process in China and Israel to gather more knowledge on the role of cultural differences in the model. The empirical endeavor treated media exposure as a cue to risk perception, which directly and indirectly triggers fear and motivates behavioral changes, and this mechanism was conditional on SES. Results show that the provoked susceptibility perception explains −0.07 to 0.06 units increase in fear and −0.13 to 0.25 units increase in behavior adoption, and severity perception explains a −0.04 to 0.16 units increase in fear and −0.38 to 1.55 units higher in behavior adoption. The mechanisms of mediation effect of risk perception account for 6–40% of the media effect through susceptibility and about 10–70% of the media effect through severity. Moreover, we observe SES as a moderator in differentiating the media effect of legacy media, digital media, message length preference, and following contents.

Here are some insights that we would like to mention.

First, as posited by media dependency theory (Ball-Rokeach, 1985), people utilize various media resources to gain information about the pandemic and remove the uncertainty of their situations. Classical media exposures, like daily exposure duration, legacy media, digital media, and interpersonal communication, are demonstrated to contribute substantially to fear and behavior through the mediation path of risk perception. These results corroborate the findings of previous studies that excessive exposure to the outbreak information via media can lead to overestimating people’s risk, thus causing more fear (Bendau et al., 2021) and motivating preventive behavior (Ranjit et al., 2021). In addition, the current study further probes the roles of other media exposures that have rarely been assessed, such as the length preference for message and contents following, and provided evidence for their explanatory functions for media total effect and mediation effect. Those findings highlight the importance of efficient health communication strategies that should include a broader spectrum media exposure and consider the way of news framing.

Second, the study provides empirical evidence about the mechanisms of mediation effect through risk perception. The finding testifies the roles of susceptibility and severity perceptions in explaining how and why the media effect occurs, and it demonstrates the value of psychological arousal status in the fear and behavior study. We also provide a sizable suggestion that the mediation mechanism of severity perception elicited by media exposure is more accountable in fear arousal and behavior motivation than the mechanism of susceptibility perception because it is explainable for higher portions of the total effect. Meanwhile, we also find that severity perception is superior in facilitating the media effect on mental and behavioral health to susceptibility perception since the mediation effects conveyed from the former mechanism reported relatively greater in magnitudes. Thus, it is beneficial to consider their assessment, especially people’s severity perception, in designing the prevention and intervention programs.

Third, the study conducts sensitivity analyses to quantify the robustness of the mediation effects. We include the assumption of sequential ignorability in the identification and estimation to obtain causal effects, and the sensitivity analysis enables us to make valuable conclusions even with violation of this assumption. Given this, our results suggested that the causal mediation effect of severity perception linking the relationship between media use and behavior adoption is robust, and the magnitude of any unobserved variables needs to be large enough to alter this inference. Again, this emphasizes that severity perception can be a trusted causal mechanism implemented in a real-world campaign.

Fourth, the study enhances our understating about health inequalities derived from social disparities. Findings suggest that socioeconomic situation shapes individuals’ mental and behavioral responses outcomes disproportionately as a result of media consumption. It enlightens the relationship between health inequalities and social inequalities. Lack of financial support and essential knowledge undoubtedly reduce resilience to disadvantaged surroundings. Thus, reducing social difficulties is still a priority for a better health goal. At the same time, part of our results suggests that, under some circumstances, people from more advantaged social groups may suffer from greater adverse mental difficulties, which should also be considered in the prevention and intervention strategies.

Some limitations of the study should be noted. First, the assumption’s validity for causal inference cannot be guaranteed. In the absence of randomization, the assumption of sequential ignorability in an observational study may not hold. For example, we are unable to eliminate all the selection bias. Elaborated by the media selectivity paradigm that people only attend to limited messages and only these attractive messages have the potential to influence them (Klapper, 1960; Knobloch-Westerwick, 2014), these consumptions will introduce the possibility of residual confounding. Thus, we collect multiple pre-exposure data and controlled them as exogenous covariates. This ensures the assumption to be at least approximately true and supplements the validity to identify the causal effect of interest. In the meantime, given that the identification of causal mechanism relies upon this untestable assumption, we conduct sensitivity analysis and report robust evaluations of the causal inference results to potential violations of this assumption.

In addition, the measurement and coding of the outcome variables could be problematic. The official recommendation and personal judgment of health behaviors are diverse and dynamic. For example, many countries defined the mask-wearing as recommended, while some mandated the wearing when we collected the data. Similarly, doing physical exercise is a regular lifestyle for some people while not for others. It is not easy to disentangle the degree of individual behavior caused by media effects from other sources. Thus, we use multiple indicators to measure people’s behavioral inclinations, include almost all virus-related preventive behavior, avoidance behavior, and lifestyle behaviors mentioned by the media and officials, and to provide a general and stable disposition of individual behavioral adoption.

Last but not least, the data are relatively small and lack representation for causal inference. Ideally, future studies can use randomized experiment design with more sophisticated measurements for data collection under this conceptual model.

Footnotes

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Gustavo S Mesch

Author biographies

Gustavo S Mesch is a Professor of Digital Sociology (Emeritus). He holds a PhD degree in Sociology from the Ohio State University (1993). His research interests are technology and society, social effects of new media, Media effects on social behavior.

Xue-Jing Liu received her Master of Public Health at the University of Haifa. Her research interest focuses on media, health behavior, and social epidemiology.

References

Altshuler

(2020) The truth about fake news. Israel Democracy Institute. Available at: https://www.idi.org.il/articles/28828 (accessed 8 January 2023).

Allan

(2002) Media, Risk and Science: Issues in Cultural and Media Studies. London: Open University Press.

Allington

Duffy

Wessely

, et al. (2021) Health-protective behaviour, social media usage and conspiracy belief during the COVID-19 public health emergency. Psychological Medicine 51(10): 1763–1769.

Alter

Oppenheimer

(2009) Uniting the tribes of fluency to form a metacognitive nation. Personality and Social Psychology Review 13(3): 219–235.

Anderson

Bushman

(2002) Human aggression. Annual Review of Psychology 53(1): 27–51.

Angawi

Albugmi

(2022) The impact of social media on risk perceptions during COVID-19 in Saudi Arabia. Frontiers in Public Health 10: 898041.

Aubrey

(2020) Poll: most Americans say U.S. “doing enough” to prevent coronavirus spread. National Public Radio, 4 February. Available at: https://www.npr.org/sections/health-shots/2020/02/04/802387025/poll-most-americans-say-u-s-doing-enough-to-prevent-coronavirus-spread

Avery

(2010) Contextual and audience moderators of channel selection and message reception of public health information in routine and crisis situations. Journal of Public Relations Research 22(4): 378–403.

Bader

Baker

Day

, et al. (2020) Fear itself: the causes and consequences of fear in America. Social Forces 99: e1–e3.

10.

Bakshi

Deshmukh

Kumar

(2021) The Psychology of Pandemics: Preparing for the Next Global Outbreak of Infectious Disease (ed. Taylor

). Newcastle upon Tyne: Cambridge Scholars Publishing.

11.

Ball-Rokeach

(1985) The origins of individual media-system dependency: a sociological framework. Communication Research 12(4): 485–510.

12.

Ball-Rokeach

DeFleur

(1976) A dependency model of mass-media effects. Communication Research 3(1): 3–21.

13.

Bendau

Petzold

Pyrkosch

, et al. (2021) Associations between COVID-19 related media consumption and symptoms of anxiety, depression and COVID-19 related fear in the general population in Germany. European Archives of Psychiatry and Clinical Neuroscience 271(2): 283–291.

14.

Berry

Wharf-Higgins

Naylor

(2007) SARS wars: an examination of the quantity and construction of health information in the news media. Health Communication 21(1): 35–44.

15.

Bodas

Siman-Tov

Peleg

, et al. (2015) Anxiety-inducing media: the effect of constant news broadcasting on the well-being of Israeli television viewers. Psychiatry 78(3): 265–276.

16.

Bonanno

Brewin

Kaniasty

, et al. (2010) Weighing the costs of disaster: consequences, risks, and resilience in individuals, families, and communities. Psychological Science in the Public Interest 11(1): 1–49.

17.

Brashers

Neidig

Goldsmith

(2004) Social support and the management of uncertainty for people living with HIV or AIDS. Health Communication 16(3): 305–331.

18.

Brewer

Chapman

Gibbons

, et al. (2007) Meta-analysis of the relationship between risk perception and health behavior: the example of vaccination. Health Psychology 26(2): 136–145.

19.

Bryant

Oliver

(2009) Media Effects: Advances in Theory and Research. London: Routledge.

20.

Carpenter

(2010) A meta-analysis of the effectiveness of health belief model variables in predicting behavior. Health Communication 25(8): 661–669.

21.

Casero-Ripollés

(2020) Impact of Covid-19 on the media system. Communicative and democratic consequences of news consumption during the outbreak. El Profesional de la Información 29(2): e290223.

22.

China Internet Network Information Center (2020) The 45th statistical report on internet development in China. Available at: https://www.cnnic.com.cn/IDR/ReportDownloads/202008/P020200827549953874912.pdf (accessed 8 February 2023).

23.

Choi

Yoo

Noh

G-Y

, et al. (2017) The impact of social media on risk perceptions during the MERS outbreak in South Korea. Computers in Human Behavior 72: 422–431.

24.

Coe

Gatewood

Moczygemba

(2012) The use of the health belief model to assess predictors of intent to receive the novel (2009) H1N1 influenza vaccine. Innovations in Pharmacy 3(2): 1–11.

25.

Junsheng

Ruosi

(2015) Clustering analysis of health information acquisition channels: active acquisition and passive receiving. Journal of International Communication 5: 81–93.

26.

Dorot

(2020) Media influence matrix: Israel. CEU Center for Media, Data and Society. Available at: https://cmds.ceu.edu/sites/cmcs.ceu.hu/files/attachment/basicpage/1860/mimisraelfunding.pdf (accessed 17 March 2023).

27.

Dudo

Dahlstrom

Brossard

(2007) Reporting a potential pandemic: a risk-related assessment of avian influenza coverage in US newspapers. Science Communication 28(4): 429–454.

28.

Galarce

Ramanadhan

Weeks

, et al. (2011) Class, race, ethnicity and information needs in post-treatment cancer patients. Patient Education and Counseling 85(3): 432–439.

29.

Gao

Zheng

Jia

, et al. (2020) Mental health problems and social media exposure during COVID-19 outbreak. PLoS ONE 15(4): e0231924.

30.

Garfin

Silver

Holman

(2020) The novel coronavirus (COVID-2019) outbreak: amplification of public health consequences by media exposure. Health Psychology 39(5): 355–357.

31.

Goodyear

Boardley

Chiou

S-Y

, et al. (2021) Social media use informing behaviours related to physical activity, diet and quality of life during COVID-19: a mixed methods study. BMC Public Health 21(1): 1–14.

32.

Holland

(1986) Statistics and causal inference. Journal of the American Statistical Association 81(396): 945–960.

33.

Holman

Garfin

Silver

(2014) Media’s role in broadcasting acute stress following the Boston Marathon bombings. Proceedings of the National Academy of Sciences 111(1): 93–98.

34.

Zhang

(2014) Channel selection and knowledge acquisition during the 2009 Beijing H1N1 flu crisis: a media system dependency theory perspective. Chinese Journal of Communication 7(3): 299–318.

35.

Imai

Keele

Tingley

(2010a) A general approach to causal mediation analysis. Psychological Methods 15(4): 309–334.

36.

Imai

Keele

Yamamoto

(2010b) Identification, inference and sensitivity analysis for causal mediation effects. Statistical Science 25(1): 51–71.

37.

Israel Ministery of Health (2020) COVID Information Center Updated info and official guidelines for daily life during COVID-19. Available at: https://corona.health.gov.il/en/ (accessed 30 March 2023).

38.

Jackson

(2009) A psychological perspective on vulnerability in the fear of crime. Psychology, Crime & Law 15(4): 365–390.

39.

Janz

Becker

(1984) The health belief model: a decade later. Health Education Quarterly 11(1): 1–47.

40.

Jones

Jensen

Scherr

, et al. (2015) The health belief model as an explanatory framework in communication research: exploring parallel, serial, and moderated mediation. Health Communication 30(6): 566–576.

41.

Katapodi

Lee

Facione

, et al. (2004) Predictors of perceived breast cancer risk and the relation between perceived risk and breast cancer screening: a meta-analytic review. Preventive Medicine 38(4): 388–402.

42.

Kemper

(1987) How many emotions are there? Wedding the social and the autonomic components. American Journal of Sociology 93(2): 263–289.

43.

King

Tomz

Wittenberg

(2000) Making the most of statistical analyses: improving interpretation and presentation. American Journal of Political Science 44: 347–361.

44.

Klapper

(1960) The Effects of Mass Communication. New York: Free Press.

45.

Knobloch-Westerwick

(2014) Choice and Preference in Media Use: Advances in Selective Exposure Theory and Research. London: Routledge.

46.

Kurl

Korzinski

(2020) Half of Canadians Taking Extra Precautions as Coronavirus Continues to Spread Around the Globe. Vancouver, BC: Angus Reid Institute.

47.

Lang

Shin

Bradley

, et al. (2005) Wait! Don’t turn that dial! More excitement to come! The effects of story length and production pacing in local television news on channel changing behavior and information processing in a free choice environment. Journal of Broadcasting & Electronic Media 49(1): 3–22.

48.

Leung

Lam

, et al. (2003) The impact of community psychological responses on outbreak control for severe acute respiratory syndrome in Hong Kong. Journal of Epidemiology & Community Health 57(11): 857–863.

49.

Zhang

(2020) Investigation report for national broadcasting and frequency in 2020. China Radio Film & TV.

50.

Liu

Fraustino

Jin

(2016) Social media use during disasters: how information form and source influence intended behavioral responses. Communication Research 43(5): 626–646.

51.

Liu

Zhang

Huang

(2020) Media exposure to COVID-19 information, risk perception, social and geographical proximity, and self-rated anxiety in China. BMC Public Health 20(1): 1–8.

52.

Liu

(2005) Reading behavior in the digital environment: changes in reading behavior over the past ten years. Journal of Documentation 61: 700–712.

53.

Melki

Tamim

Hadid

, et al. (2022) Media exposure and health behavior during pandemics: the mediating effect of perceived knowledge and fear on compliance with COVID-19 prevention measures. Health Communication 37(5): 586–596.

54.

Morton

Duck

(2001) Communication and health beliefs: mass and interpersonal influences on perceptions of risk to self and others. Communication Research 28(5): 602–626.

55.

Nishi

Koido

Nakaya

, et al. (2012) Peritraumatic distress, watching television, and posttraumatic stress symptoms among rescue workers after the Great East Japan earthquake. PLoS ONE 7(4): e35248.

56.

Pengpai News (2021) Television statistics on 2020. Available at: https://ent.sina.cn/tv/tv/2021-01-12/detail-iiznezxt2368027.d.html (accessed 8 February 2023).

57.

Petty

Briñol

Priester

(2009) Mass media attitude change: Implications of the Elaboration Likelihood Model of persuasion. In Bryant

Oliver

(eds) Media effects: Advances in theory and research (3rd ed., pp. 125-164). London: Routledge.

58.

Potter

(2009) Conceptualizing the audience. In: Nabi

Oliver

(eds) Handbook of Media Effects. Thousand Oaks, CA: SAGE, pp. 19–34.

59.

Potter

(2012) Media Effects. Thousand Oaks, CA: SAGE.

60.

Preacher

Curran

Bauer

(2006) Computational tools for probing interactions in multiple linear regression, multilevel modeling, and latent curve analysis. Journal of Educational and Behavioral Statistics 31(4): 437–448.

61.

Price

Tewksbury

Powers

(1997) Switching trains of thought: the impact of news frames on readers’ cognitive responses. Communication Research 24(5): 481–506.

62.

Ranjit

Shin

First

, et al. (2021) COVID-19 protective model: the role of threat perceptions and informational cues in influencing behavior. Journal of Risk Research 24(3–4): 449–465.

63.

Rosenstock

(1974a) Historical origins of the health belief model. Health Education Monographs 2(4): 328–335.

64.

Rosenstock

(1974b) The health belief model and preventive health behavior. Health Education Monographs 2(4): 354–386.

65.

Sasaki

Kuroda

Tsuno

, et al. (2020) Exposure to media and fear and worry about COVID-19. Psychiatry and Clinical Neurosciences 74: 501–502.

66.

Schimmenti

Billieux

Starcevic

(2020) The four horsemen of fear: an integrated model of understanding fear experiences during the COVID-19 pandemic. Clinical Neuropsychiatry 17(2): 41–45.

67.

Scopelliti

Pacilli

Aquino

(2021) TV news and COVID-19: media influence on healthy behavior in public spaces. International Journal of Environmental Research and Public Health 18(4): 1879.

68.

Seo

Matsaganis

(2013) How interpersonal communication mediates the relationship of multichannel communication connections to health-enhancing and health-threatening behaviors. Journal of Health Communication 18(8): 1002–1020.

69.

Shih

T-J

Wijaya

Brossard

(2008) Media coverage of public health epidemics: linking framing and issue attention cycle toward an integrated theory of print news coverage of epidemics. Mass Communication & Society 11(2): 141–160.

70.

Simba

Kakoko

(2012) Primacy of effective communication and its influence on adherence to artemether-lumefantrine treatment for children under five years of age: a qualitative study. BMC Health Services Research 12(1): 1–8.

71.

Solomon

(2022) Information disruptions and disruptive information sources in the practice of law: obstacles in gathering information, through an Israeli lens. Journal of Librarianship and Information Science. Epub ahead of print 3 June. DOI: 10.1177/09610006221097975.

72.

Sun

Grossman

, et al. (2021) Comparison of COVID-19 information seeking, trust of information sources, and protective behaviors in China and the US. Journal of Health Communication 26(9): 657–666.

73.

Taylor

Asmundson

Hyprochondria (2004) Treating Health Anxiety: A Cognitive-Behavioral Approach. New York: Guilford Press.

74.

Thompson

Garfin

Holman

, et al. (2017) Distress, worry, and functioning following a global health crisis: a national study of Americans’ responses to Ebola. Clinical Psychological Science 5(3): 513–521.

75.

Turner

(2014) The evolution of human emotions. In: Stets

Turner

(eds) Handbook of the Sociology of Emotions: Volume II. Dordrecht: Springer, pp. 11–31.

76.

Valkenburg

Peter

(2013) The differential susceptibility to media effects model. Journal of Communication 63(2): 221–243.

77.

Viswanath

(2005) The communications revolution and cancer control. Nature Reviews Cancer 5(10): 828–835.

78.

Viswanath

(2006) Public communications and its role in reducing and eliminating health disparities. In: Thomson

Mitchell

Williams

(eds) Examining the Health Disparities Research Plan of the National Institutes of Health: Unfinished Business. Washington, DC: Institute of Medicine, pp. 215–253.

79.

Wakefield

Loken

Hornik

(2010) Use of mass media campaigns to change health behaviour. The Lancet 376(9748): 1261–1271.

80.

Wang

Xiao

Lin

, et al. (2022) Geographic social inequalities in information-seeking response to the COVID-19 pandemic in China: longitudinal analysis of Baidu Index. Scientific Reports 12(1): 1–10.

81.

Zeballos Rivas

Lopez Jaldin

Nina Canaviri

, et al. (2021) Social media exposure, risk perception, preventive behaviors and attitudes during the COVID-19 epidemic in La Paz, Bolivia: a cross sectional study. PLoS ONE 16(1): e0245859.

82.

Zellner

(1962) An efficient method of estimating seemingly unrelated regressions and tests for aggregation bias. Journal of the American Statistical Association 57(298): 348–368.

83.

Zhang

Kong

Chang

(2015) Media use and health behavior in H1N1 flu crisis: the mediating role of perceived knowledge and fear. Atlantic Journal of Communication 23(2): 67–80.

84.

Zhao

Fan

Basnyat

, et al. (2020) Online health information seeking using “# COVID-19 patient seeking help” on Weibo in Wuhan, China: descriptive study. Journal of Medical Internet Research 22(10): e22910.