Media use theory is applicable for analyzing the motivations, frequency, duration, and effects of media use on individuals’ behaviors and opinion formation (Schneiders et al., 2022). On the other hand, Media effects theory is suitable for studying the direct or indirect influences of media on individuals’ cognition, attitudes, and behaviors, encompassing various subfields such as social cognitive theory, selective perception, and motivation (Valkenburg et al., 2016). Media plays a significant role in shaping an individual’s risk perception by providing information about potential risks and strategies to mitigate them (Frh, 2017). Risk perception is individuals’ subjective judgments of the likelihood of negative events that may pose immediate or long-term threats to their health and well-being (Johnson et al., 2023). When the perceived risk exceeds their acceptable threshold, individuals may adapt their travel plans accordingly (Airak et al., 2023).

Media exposure and social influence are external factors that often influence risk perception (Liu M. et al., 2020). During the COVID-19 pandemic, the public relies on various types of mass media for information about COVID-19 and the associated travel risks (Ma et al., 2021). Consequently, different types of media contribute to the formation of individuals’ risk perception (Du et al., 2021). For example, exposure to COVID-19 information from mainstream media sources such as cable news channels, local news channels, and newspapers increases individuals’ perceived vulnerability (Olagoke et al., 2020).

The use of social media platforms is associated with higher levels of psychological and social anxiety regarding potential risks, consequently leading to increased risk perception. Moreover, research indicates that the use of new media (both official and unofficial sources) affects the public’s perception of risk (Zhou, 2022). Specifically, frequent exposure to epidemic information through new media platforms is linked to higher risk perception (Li et al., 2021). This risk perception may even surpass that of individuals relying solely on traditional media sources (Huang et al., 2018). Therefore, this study hypothesizes as follows:

Both media effects theory and social amplification theory of risk propose a relationship between media usage and behavior (Pidgeon, 2010). Under risk events, individuals perceive risks transmitted by the media and this affects their behavior (Wu and Li, 2017). As a result, the media plays a crucial role in promoting preventive behavior, amplifying the epidemic messages from the traditional media, and thus preventing large infection outbreaks. Social media provides the same effect of encouraging individuals to take preventive behaviors during a crisis (Seo, 2019). Furthermore, both threat and coping appraisals of public health events can enhance travelers’ protection motivations, which in turn affect their actual behaviors (Wang et al., 2019). Here tourism protection behavior intention is defined as the subjective tendency of individuals to take actions to protect, promote, or maintain their health and safety in the context of an epidemic. Amplification of official messages through media channels increases the likelihood that travelers will protect themselves against health risks. Previous research indicates that media use affects travel behavior intention. For example, the intensity of social media use affects tourists’ cognitive image of a destination and travel behavior intention (Huang et al., 2018). On the other hand, heavy Internet use is associated with reduced travel away from home (Lachapelle and Jean-Germain, 2019). Media usage can also influence adoption of environmentally friendly travel behaviors, with various types of media having different degrees of influence (Borowski et al., 2020). The effect of media exposure on preventive behavioral intentions has been proposed as due to it influencing subjective norms (Liu L. et al., 2020). The more epidemic information tourists are exposed to, the stronger effect on tourism behavioral intention (Cahyanto and Liu-Lastres, 2020). Therefore, this paper proposes the following hypothesis:

Protection motivation theory considers that people’s risk perception is a key factor affecting behavioral intentions (Floyd et al., 2000). Empirical studies have found that people decide whether to take protective behaviors depending on their perception of the risk situation (Kowalski and Black, 2021; Nazione et al., 2021). A study of the Ebola virus shows that the higher a person’s perceived susceptibility, severity and travel risk, the more likely they are to avoid travel (Cahyanto et al., 2016). Tourists with a higher perception of the risk of smog haze are likely to change their travel plans based on the weather conditions and their effect on haze levels (Wu and Li, 2017). Similarly for COVID 19, the higher the individual perceives their susceptibility to the coronavirus, the greater their intention to take protective travel behavior (Raina et al., 2022), by changing their travel patterns or reducing travel frequency (Abdelrahman, 2022). The stronger the risk consequence, the lower the travel intention of tourists (Cahyanto and Liu-Lastres, 2020).

While Protection Motivation Theory (PMT) includes efficacy beliefs, this study’s primary focus is on risk perception for several reasons. In high-risk scenarios like pandemics, risk perception often supersedes efficacy beliefs in shaping behavior (Bish and Michie, 2010; Dryhurst et al., 2020). The urgency and immediacy of the COVID-19 threat made risk perception an apt focal point (Siegrist et al., 2021). Furthermore, focusing on risk perception streamlines our hypothesis testing, providing a targeted understanding of the situation. This focus is crucial in time-sensitive cases like pandemics, as demonstrated during the influenza A (H1N1) outbreak where risk perception significantly impacted protective measure adoption (Weerd et al., 2011). However, we recognize that efficacy beliefs, alongside trust and risk perception, form an important aspect of PMT, shaping behavioral outcomes (Siegrist et al., 2021). Our study has not incorporated this perspective, but it provides an enriching opportunity for future research in this domain. Therefore, in light of these findings, this paper posits the following hypothesis:

An individuals’ interpretation of information and events will directly affect whether they act (Albarracín and Wyer, 2000; Ma and Cao, 2019). Related media psychology research has found that individuals who use media can either be harmed or helped, depending on their personal cognition, which acts as a bridge between external information and behavioral consequences (Stever et al., 2021). Previous tourism research has found that risk perception plays a mediating role between media use and tourism behaviors. However, the intermediary role of risk perception on the relationship between media use and tourism behavior has not been examined (Chien et al., 2016). Instead, studies have focused on the relationship between destination image and the epidemic situation. For example, destination image perception plays a complete intermediary role between social media user generated content and tourism intention (Huang et al., 2018), and personal perception plays a mediating role between epidemic situation and protective behavior (Luo and Wang, 2021), or tourism behavior intention (Lebrun et al., 2021). Other studies have examined how media affects individual health attitudes and encourages individuals to adopt protective behaviors (Giustini et al., 2018). Therefore, this paper proposes the following hypotheses:

Section 2.3 of our study explores why risk perception, rather than efficacy beliefs, is more crucial during high-risk crises like pandemics. Hence, our research narrows down its focus to the impact of media use intensity on tourism behavior in the COVID-19 pandemic, emphasizing risk perception. Efficacy beliefs, although noteworthy, are not included in the current study. Future research might explore this aspect, but the current research primarily probes risk perception’s influence. This approach is based on our premise that high-risk crises make risk perception a more pressing and influential factor in shaping behaviors.

This study aims to understand how media use affects behavioral decision-making during the COVID-19 pandemic, taking into account the theoretical perspectives of the Media Use Theory (MUT), Media Effect Theory (MET), Social Amplification Theory (SAT), and Protection Motivation Theory (PMT). These integrated theories provide a rigorous theoretical foundation for the conceptual model and enhance the analysis of the relationships between media use, risk perception, and travel protection behavior intentions. The revised conceptual model hypothesizes that an individual’s exposure to epidemic information in the media, influenced by the Social Amplification Theory, Media Setting Theory, and Protection Motivation Theory, directly affects their risk perception and travel protection behavior intentions and can also indirectly affect travel protection behavior intentions through the mediation of risk perception.

In this revised model, the Media Effect Theory emphasizes how media exposure shapes an individual’s risk perception by providing information about potential risks and how to avoid them. The Social Amplification Theory suggests that media usage can amplify the transmission of risk messages and influence individuals’ behavior during risk events. The Protection Motivation Theory considers the role of risk perception in influencing individuals’ behavioral intentions and decision-making in response to threats. By incorporating these integrated theories into the conceptual model, the study establishes a rigorous theoretical foundation that justifies the continued relevance of the research (see Figure 1).

The questionnaire consists of two main parts. The first collects the demographic characteristics of the respondents, including gender, age, education level, occupation, and monthly disposable income. The second contains the measurement items of the variables in the research model. By design, the measurement instrument drew upon scales used in past research to ensure the instrument’s consistency, reliability, and validity. The measures were adapted to the context of the analysis, which was the pandemic situation. The model includes four latent variables. The intensity of media use means the extent to which a person is actively engaged in different media supported activities. The article examines the previous measurement standards and analysis indicators of media intensity (Rosen et al., 2013; Du et al., 2021; Oh et al., 2021). To understand the degree of dependence of the respondents on different media under the condition of multiple media exposure, this paper has designed indicators of new media use intensity and traditional media use intensity. This article explores the research on how people perceive the risks of an epidemic (Bae and Chang, 2020; Sánchez-Cañizares et al., 2021), and understand the views of the respondents in terms of the severity of the epidemic itself, and the susceptibility and severity of infection sequelae. Seven dimensions for measuring protective tourism behavior are proposed (Wang et al., 2019; Neuburger and Egger, 2020; Lebrun et al., 2021). The specific items are all measured using a 5-point Likert-type scale. The intensity of media use is measured by asking participants the extent that they have been exposed to media information about epidemic risks (1 = not at all, to 5 = to a great extent) (see Table 1).

Data was collected from the residents of Shanghai Songjiang University Town. Songjiang University Town residents have the highest trips per year in Shanghai. The residents therefore will be expected to have been strongly affected by the COVID-19 lockdowns. Before the formal survey, the researcher conducted a trial survey (N = 76) in early June 2021. After removing invalid questionnaires, the researcher analyzed the reliability and validity of 71 valid questionnaires and deleted and revised the problematic items. The distribution and collection of formal questionnaires were carried out in late July and August of 2021. In July 2021, Shanghai reported a total of 108 confirmed cases of COVID-19, all of which were imported cases. In August 2021, Shanghai reported a total of 171 confirmed cases of COVID-19, including 10 local cases and 161 imported cases. In late July 2021, the Shanghai Culture and Tourism Administration issued a notice on strict implementation of epidemic prevention and control measures. On August 6, Shanghai suspended trans-provincial group tours (SMACT, 2021). The questionnaires were distributed online and collected by snowballing. A total of 650 online questionnaires were distributed and 594 were returned. A consistency test of the samples was carried out. A total of 103 invalid questionnaires were excluded due to blanks and omissions or to most questions on a response being the same number. A total of 491 valid questionnaires were retained (82.66%).

The sample consisted of 43.6% male and 56.4% female respondents, which is in line with the local area gender characteristics. In the sample, those 18–25 years old accounted for 41.1% of the total and 25–50 years old respondents accounted for 42.4%. Most of the participants have a higher education level; 59.1% of the participants are college/undergraduate students, 22.4% have a master’s degree or above. Monthly disposable income was found in the following ranges: less than 2,480 Rmb accounting for 40.5%, 2,481–5,000 Rmb accounting for 18.1%, and 5,001–10,000 Rmb accounting for 33.2%. The occupations of the participants were mainly students accounting for 39.5%. The sample population’s educational attainment, income level, and other status variables are associated with the higher proportion of students in the sample area, as well as the demographic characteristics of the university town (Table 2).

Given the youthful demographic of the University Town, we surmise that the influence of new media was potentially stronger in this sample. Nevertheless, it is crucial to note the increasing universality of smart device use. As of 2021, there were 950 million smartphone users in China, an adoption rate of 68%. Almost all eligible age groups in Shanghai had widespread smartphone access. Particularly, Shanghai statistics indicate that 88.1% of residents aged 60 and above own a smartphone. Consequently, the likelihood of information acquisition via new media is growing across all age groups. Research by Deloitte in 2022 further corroborates this when it revealed increasing smartphone dependency across all age groups during the COVID-19 pandemic. Truthfully, younger populations might have more exposure to new media, but older populations are also trending in this direction.

Emphasizing the sample’s affiliation to Songjiang University Town - an up-and-coming, technologically advanced area in Shanghai - underscores the potential futuristic relevance of the findings. Despite a presently limited representation, particularly from seniors and individuals below 18, the research offers valuable insights into travel decisions that may indeed shape those of future Chinese populations. This younger demographic, with their strong reliance on new media and rational decision making, could potentially form a model for examining and predicting travel behaviors and risk perceptions in the rapidly evolving technology and information landscape.

Exploratory factor analysis was used to test the reliability and validity of the questionnaire variables. It is important to note that the EFA was carried out using data derived specifically from the trial survey, with a sample size of 76 participants. Cronbach coefficient is the most used measure of reliability. Cronbach’s alpha values of the four variables are 0.798, 0.845, 0.814, 0.846, above the acceptable level of 0.7. The Kaiser-Meyer Olkin (KMO) test and Bartlett test were used to find if the data were suitable for factor analysis. The KMO of the four variables are 0.712, 0.787, 0.707, 0.910 respectively, and the p value of the four variables is 0.000, KMO > 0.7, p < 0.05, indicated the adequacy of the sample size as well as the existence of the latent variables. The KMO value for the total scale items is 0.876, and Bartlett’s test of sphericity (χ² = 3,165.854, p = 0.000) is significant, indicating that the scale is appropriate for factor analysis. Exploratory factor analysis of the scale items was carried out using principal component analysis. Four factors with a factor load greater than 0.5 were extracted. The result after rotation shows that the loading of TI2 was less than 0.5 and hence it was deleted. After deleting this item, factor analysis was carried out again, and five factors were extracted with factor loadings greater than 0.5, and the cumulative explained variance was 62.985%. This indicates that the internal consistency of scale structure was valid, and the screened factors are representative.

To further investigate the validity of the scale, this paper uses AMOS24.0 software to conduct confirmatory factor analysis to test the relationship between observed variables and latent variables. Notably, the CFA was undertaken utilizing a comprehensive dataset derived from the total survey, which included a sample size of 491 participants. First the convergent validity of the scale the discriminant validity among variables was tested. Table 3 shows the results of Convergent validity test analysis. Convergent validity is usually measured with CR and AVE. If CR > 0.7 and AVE > 0.5, it indicates that the aggregation validity is good. If AVE < 0.4, the variables need to be deleted to improve the convergence effect. The result shows that the AVE of tourism protection behavior intention is only 0.395, and the loadings of TI7 and TI6 are low, so they should be deleted. After deleting the items, the AVE value of tourism behavior intention is 0.402. This AVE value is less than 0.5 but greater than 0.4 due to the sample size, so it is within the acceptable range. Overall, the convergence validity of the scale is high. Table 3 shows the results of the reliability and conversion validity test analysis.

Discriminant validity is used to test the difference between two factor constructs (Iacobucci, 2010). Discriminant validity was tested by comparing the square root of AVE for individual constructs with the correlations among the latent variables. As shown in Table 4, the square root of AVE exceeded those of the off-diagonal elements, which means the measurement scale of each latent variable has good discriminant validity.

In a confirmatory factor analysis, model fit refers to how closely observed data match the relationships specified in a hypothesized model (Ribeiro et al., 2017). The commonly used fit indexes include two judgment criteria: the similarity index and dissimilarity index between the two models. The similarity index includes the goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI), comparative fit index (CFI), and Tucker-Lewis index (TLI). Dissimilarity indicators include the root mean square error of approximation (RMSEA) and the standardized root mean square residual (SRMR). As shown in Table 5, the model has a good degree of fit since it meets critical value requirements of each index.

After the reliability and validity of the scales meet the requirements, the path analysis is used to measure the relationships between the independent and dependent variables to verify whether the hypotheses are established. AMOS24.0 software is used to analyze the path of the model and the following results are obtained.

As shown in Figure 2, the intensity of new media use is significantly associated with both epidemic risk perception and tourism protection behavior intention. However, the intensity of traditional media use has a significant negative impact on the perception of epidemic risk but has little correlation with the behavioral intention of tourism protection. The results of this research suggest that individuals who perceive a higher risk of an epidemic are more likely to have a stronger intention to engage in protective behaviors concerning tourism.

Table 6 provides the results of the path analysis and hypotheses verification. The path analysis showed that, except for the insignificant effect of traditional media use intensity on tourism protection behavior intention, the other variables were all significantly related. The data does not support H4 but does support the other hypotheses.

This paper examines whether epidemic risk perception mediates the relationship between media use intensity and tourism protection behavior intention, by estimating the total, direct, and indirect effects. The results of the effect analysis conducted using AMOS24.0, are presented in Table 7. The total effect of an independent variable on another dependent variable is the sum of its direct effect and indirect effects. The total effect of traditional media use intensity on tourism protection behavior intention is −0.139, of which the direct effect is −0.005, while the indirect effect through epidemic risk perception is −0.134. This means the total effect of traditional media use intensity on tourism protection behavior intention is significant, and the negative relationship is mediated by epidemic risk perception. H7 is supported.

The total effect of new media use intensity on tourism protection behavior intention is 0.514, of which the direct effect is 0.333, while the indirect effect through epidemic risk perception is 0.181. Epidemic risk perception partially explains the relationship between new media use intensity and tourism protection behavior intention. H6 is supported.

Multi-Group Analysis (MGA) is a statistical method used in structural equation modeling (SEM) when research involves comparison between two or more groups (Li et al., 2023). This approach facilitates the examination on whether construct relationships, as theorized in our model, hold uniformly across various groups. Notable variations emerge based on different demographic factors. Gender variations, for instance, have distinct behavioral intentions (Yang et al., 2022c). Similarly, age or generational groups exhibit unique intentions (Wu et al., 2023), while geographic aspects also contribute to travel behavior (Yang et al., 2022a). These analyses offer insights into demographic influences on the relationship among media use intensity, travel risk, and behavioral intentions.

Table 8 represents the model fit test for the multiple group analysis (MGA) conducted on different demographic variables. The significant Chi-square values suggest that the demographic variables in the model have an influence on the relationship between media use intensity, risk perception, and travel protective behavior intention. The good fit of the model further supports the validity of the MGA in analyzing these relationships.

Our investigation revealed a significant positive association between new media use intensity and both epidemic risk perception, as well as an intention to adopt safety-conscious tourism behaviors. This is consistent with other scholars’ research on the influence of social media on perception (Huang et al., 2018; Liu M. et al., 2020), and also in line with research about the impact of social media on tourism intentions (Koo et al., 2016; Yoo et al., 2016). This suggests that as individuals’ new media usage escalates, their perceived sense of epidemic risk intensifies, which in turn has the potential to spark an increased intent to enact more vigilant tourism behaviors (Pakpour et al., 2021; Xie, 2021; Jiang et al., 2022).

Yet, our exploration into the role of traditional media use yielded contrasting results. We found a direct inverse association between traditional media usage and risk perception, but surprisingly, there wasn’t any marked influence on protective behavior. While this finding resonates with some previous studies (Liu L. et al., 2020), it contradicts others (Luo and Wang, 2021), exhibiting the contentious nature of traditional media’s influence on public risk perception.

These disparities might be attributed to several factors. For instance, the level of pandemic control, the promptness of epidemic information release, and the effectiveness of prevention measures (Liu M. et al., 2020). All these elements influence the impact traditional media use has on risk perception. The contradictory results underscore the need for further research to create a more nuanced understanding of traditional media’s role in shaping perceptions and behaviors during epidemics (Popiołek et al., 2021).

Upon examination of different demographics, we found that the influence of media usage and risk perception varies across age, education level, and occupation, which corroborates with what Salman et al. (2011) found. However, a novel finding in our study was that government and public institution employees displayed lower perception of COVID-19 severity and susceptibility, which is not commonly reported in existing literature. We also identified the influence of income level on media usage patterns, which is an underexplored area.

These mixed results highlight the complexity of media influences, suggesting that the impact of media on different population segments is not uniform. This is an important detail that is often overlooked in current research. Our study contributes to a more comprehensive understanding of the multifaceted impacts of media during health crises, adding depth to the current knowledge landscape.

Our study provides meaningful conclusions, sketching an assertive picture of the critical role media, both new and traditional, play in molding public perception and subsequent actions during an epidemic. These platforms carry a substantial responsibility to transmit reliable and accurate information to the public, a role that has been amplified in our findings.

The differing influences of traditional and new media indicate that an effective epidemic communication strategy should not be a one-size-fits-all approach. Rather, it necessitates a finely-tuned balance, fully leveraging the strengths and acknowledging the weaknesses of each media type. For instance, taking into account our findings, public health officials could utilize new media as a tool to enhance risk perception and safety-oriented behaviors, owing to its positive correlation with these aspects. Simultaneously, the role of traditional media in managing lower risk perceptions should not be overlooked, hinting at its use as a tool for assuaging public panic during health crises.

Next, a key insight stemming from our study is the highlight of socio-demographic disparities in media usage and their corresponding influences. This indicates a pressing need for segmented and personalized messaging. For instance, strategies should recognize that younger individuals, more likely to use new media, may perceive higher epidemic risks due to this medium’s usage. On the other hand, older individuals, who lean towards traditional media, might require more reassurances, given the negative correlation between traditional media usage and risk perception. Public health officials could consider these media usage patterns and perceptions while devising segmented communication strategies.

Lastly, our analysis underscores the significant mediating role of perceived epidemic risk in shaping protective behaviors. This finding is particularly invaluable for public health strategists, as it indicates a firm path towards achieving enhanced adherence to safety protocols (Golets et al., 2021). This insight can guide efforts to heighten risk perception through appropriately tailored messaging, thus promoting protective behaviors and helping manage public sentiment both during the current and future health crises.

In response to our findings, we make the following recommendations for enhancing public communication strategies during any health crisis, like an epidemic. These underscore the need for addressing misinformation, enable the leveraging of trusted sources to shape perceptions, and promote suitable behaviors.

Firstly, given the distinct impact of different media types, it is paramount to take into account the specific media platform used for risk communication. Owing to the heightened influence of new media, communication strategies must be directed towards curtailing panic and misinformation that often proliferate on such platforms (Xie and Liu, 2022). Simultaneously, traditional media should be harnessed to disseminate timely, accurate, and soothing information that can reassure and inform the public, hence balancing risk perception.

Secondly, recognizing that media usage and interactions markedly differ among socio-demographic groups, mass communication approaches should be tailored to address these disparities effectively. For instance, targeted messaging and programming could be developed to cater to different age groups, education levels, and occupations. Interventions could also take the form of promoting media literacy initiatives to help the public better decipher and judge the credibility, reliability and salience of the media content they consume (Akritidis et al., 2022).

In addition, collaborations with influencers on new media could be pursued to reach and engage with younger cohorts more effectively. This can establish a more robust public rapport and a more receptive audience to health advisories and recommendations (Xue et al., 2021). This approach further emphasizes the need to keep changing communication strategies, as the media landscape is swiftly transforming.

Lastly, our study underlines the crucial role of perceived epidemic risk - this could anchor strategic communication. It would be beneficial to ensure frequent, clear, and open communication about epidemic statistical trends, procedural changes, and progress updates (Wang and Zheng, 2019). By accurately managing this perception, public understanding can be fostered, promoting appropriate response behaviors. It also enables effective preparation for future public health crises by learning from the current experiences.

In conclusion, our study’s findings suggest a vital need for delicate balance and adaptability in public communications during health crises. It urges towards an audience-centric, data-driven, and dynamic approach to manage risk perception effectively and promote protective behaviors.