Social Media Campaign Effects: Moderating Role of Social Capital in an Anti-Smoking Campaign

ABSTRACT

This study examined the effects of an anti-smoking campaign that employs a crowdsourcing method with a social networking service. Drawing upon social capital scholarship and the expression effect research paradigm in eHealth systems, the study also investigated the roles of social trust and community life satisfaction in the social media campaign that has a specific geographical boundary. To that end, we conducted an experiment using a two-group pretest–posttest design. We randomly assigned 201 participants to two conditions: “campaign message reception only” as a control group and “message reception and expression” as a treatment group in which participants fully engaged in the campaign process by sharing their own campaign ideas with other participants. Findings revealed that social trust and community life satisfaction interacted with the treatment condition to positively affect persuasive intentions, but in distinct ways. Social trust moderated the effect of the message reception and interaction condition on participants’ willingness to encourage community members to stop smoking. In contrast, community life satisfaction moderated the effect of the treatment condition on encouraging others to comply with the community’s anti-smoking policy. These results provide theoretical and practical implications related to the roles of social capital in geographically defined social media campaigns.

Notes

¹ We held 62 information sessions (31 sessions for each group) in a research lab on campus. During the sessions, we explained the purposes and procedures of our research and provided consent forms. Before leaving the lab, participants took the pre-intervention survey and met with us to “friend” their experimental group’s Facebook profile. Participants in both conditions received monetary rewards for attending an information session and completing the pre- and post-test surveys.

² The results of explanatory factor analysis (EFA) of 12 items used to measure moderating and dependent variables are available from the corresponding author.

³ Descriptive statistics and correlations between the moderating and dependent variables are available from the corresponding author.

⁴ We conducted post hoc power analyses to assess the degree of reliability of our findings (Cohen, 1988). Using a software package, Gpower3.1, we calculated power in multiple regression using alpha level (.05), sample size (192), and number of predictors (H1a and H1b = 7; RQ1 and RQ2 = 8). Based on Cohen’s (1988) criteria, we predetermined the weak (ƒ² = .02), medium (ƒ² = .15), and large effect size (ƒ² = .35; Faul, Erdfelder, Lang, & Buchner, 2007). The results are following: H1a and H1b: ƒ² = .02, power = 0.23; ƒ² = .15, power = .99; ƒ² = .35, power = 1.00, and RQ1 and RQ2: ƒ² = .02, power = .22; ƒ² = .15, power = .98; ƒ² = .35, power = 1.00.

⁵ Multiple comparisons (having four control and three independent variables with two interaction terms) in our models might increase the probability of the familywise (or cumulative alpha) error rates, and this leaves a possibility of finding the significant effects by chance. However, the significant interaction effects still remain when we employed a different statistical analysis. For example, we stratified the data by the level of social trust and campus life satisfaction (by a method of median split). In the high social trust stratum, the difference in persuasive intention for stop smoking is marginally significant between the control and the treatment group (F = 3.29, p = .073), although there is no significant difference in the low social trust stratum (F = .04, p = .844). In terms of the persuasive intention for community policy, the difference between the control and the treatment group is significant in the high community life satisfaction stratum (F = 4.17, p = .044), although there is no significant difference in the low community life satisfaction stratum (F = .60, p = .440).