ABSTRACT
Haze has become one of the most life-threatening problems in China and affects over one billion Chinese people’s health. Chinese people have become more dependent on receiving health information from social media, especially Weibo and WeChat, which shapes their health perceptions and behaviors. To investigate how Chinese people’s exposure to health information on social media influenced their health protective behaviors in response to haze, particularly wearing a PM2.5 anti-haze mask, we conducted a longitudinal web-based survey of mainland Chinese. The results from the structural equation modeling showed that (a) attitude and descriptive norm positively mediated the relationships between using Weibo for health information and behavioral intention while descriptive norm negatively mediated the relationship between using WeChat for health information and intention, and that (b) attitude, descriptive norm, and injunctive norm significantly predicted behavioral intention and wearing mask, but perceived behavior control did not. Theoretical and practical implications are discussed.
Notes
1. A full list of provinces where the participants resided is available upon request.
2. The one-month lag was chosen as a tradeoff between two needs: giving respondents enough time to change their intention and behavior (Schar, Gutierrez, Murphy-Hoefer, & Nelson, Citation2006) and not delaying so long as to lose more of the sample than we already had with the one-month follow-up.
3. We performed CFA for each variable separately by following the two-step modeling rule (Kline, Citation2011). Before identifying the structural model, we specified the measurement model first, where the measurement errors will not be corrected in the structural model. Following this logic, we first specified the CFA model of each variable, and thereby corrected the measurement errors before they were carried on to the omnibus measurement model.
4. The adoption of established measures to operationalize the TPB constructs in the current study was dependent on (a) whether the measure provided a valid and reliable operationalization of the target variable, (b) whether the measure was clearly described in the manuscript, and (c) the applicability of the measure to operationalize the current behavior. Following this rule, Park and Yang (Citation2012) and Ho et al.’s (Citation2015) scales were used in the current study to measure attitude and PBC, because they also predicted intention toward participating in environmental activities, which is proximal to the behavior of wearing PM2.5 mask to protect themselves from air pollution and therefore is more applicable in the current study. Park and Smith (Citation2007) and Ho et al.’s (Citation2015) studies both provided detailed items of both descriptive and injunctive norms, and therefore were adapted to measure these two constructs in this study.
5. Although the statistics suggested the CFA model did not fit the data for descriptive norm at T1, all of the other fit indices supported a good model fit.
6. Chi-square values to degrees of freedom ratio smaller than 3 show an adequate fit (Hoelter, Citation1983). Values smaller than .05 for both the standardized root-mean-square residual (SRMR) and root-mean-square of approximation (RMSEA) suggest close model fit while values between .05 and .08 for RMSEA shows reasonable fit (Browne & Cudeck, Citation1993; Hu & Bentler, Citation1998). Values greater than .93 for comparative fit index (CFI), and greater than .95 for Tucker–Lewis Index (TLI) indicate a good fit (Hu & Bentler, Citation1998).