Unpacking the Relationship between Mobile Phone Usage Patterns, Network Size, and Civic Engagement

This study, based on a survey of 1,260 mobile phone users, investigates the relationships among patterns of mobile phone use, network size, and civic engagement. It finds that informational uses of mobile phones are positively associated with civic engagement – engagement in informal socializing and engagement in organized groups or clubs. On the other hand, relational and recreational uses have a neutral or negative association with civic engagement. Most important, the network size in mobile communication moderates the impact of relational and recreational uses on civic engagement while it mediates the impact of informational uses on civic engagement. This research indicates that the mobile phone, depending on its usage patterns and communication network size, can have different effects on civic engagement.

H4a: The size of contacts in mobile communication will have a significantly positive relationship with informal civic engagement.
H4b: The size of contacts in mobile communication will have a significantly positive relationship with formal civic engagement.
In addition to exploring the association of the mobile phone with civic engagement, this study investigates the mediating and moderating role of network size. Network size may interact with influences of media use on individuals' behavioral consequences, and network size has been shown to account for different behaviors (Rojas, 2008;Lake & Huckfeldt, 1998).
Research shows a positive association between informational use of the mobile phone and civic engagement (Kwak & Campbell, 2010b). However, information-seeking behaviors without enough connections with others may end up satisfying individuals' own informational demands, rather than developing skills and willingness to engage in issues at a broader level. Therefore, we need to take a close look at the mechanism that connects information seeking and civic engagement.
According to the cognitive mediation model, interpersonal discussion is a reasoning behavior that is a critical condition in producing outcomes of information exposure (Cho et al., 2009).
Individuals who are engaged in interpersonal discussion become able to use complex concepts, make logical connections among them, and create consistent and reasoned argumentations (Cappella, Price, & Nir, 2002). Elaborative and collective thinking yields interest in civic issues that subsequently leads to civic engagement (Eveland, 2004). And the possibility to engage in interpersonal discussion is closely related to network size.
Based on the above reasoning, this study predicts that informational uses of mobile phones indirectly influence civic engagement through network size. A large network can create opportunities to participate in civic affairs (McLeod, Sheufele, & Moy, 1999;Rojas et al., 2005) because people can elaborate on problems facing the community and learn about civic issues through their wide networking. To put it another way, when network size is small, people who are attentive to obtaining information from their mobile phones may lose opportunities to elaborate on the information and, as a result, fail to engage in civic affairs.
H5: The network size in mobile communication will mediate the impact of Online Journal of Communication and Media Technologies Volume: 7 -Issue: 2 April -2017 informational uses on civic engagement.
If people have many partners with whom they communicate regularly, they are more likely to be exposed to expanded connectedness. For instance, if a person communicates with only one friend, the boundary of interactions is constrained to the people that the friend is connected to.
But if a person communicates with three friends on a regular basis, the possible interaction boundary gets tripled. This process is likely to lead to encounters with more weak-tie people and diverse viewpoints.
Expanded connection to weak ties, in turn, provides information and resources that individuals do not find in their immediate environment of relatives and close friends. In other words, expanded weak ties can provide nonredundant, diverse information that stimulates learning about civic issues (Granovetter, 1973). In addition to have access to more information, individuals who have more frequent contact with weak ties have more probabilities of being recruited to participate (Kotler-Berkowitz, 2005). The positive relationship between weak ties and civic and community engagement has been demonstrated by a number of studies (Eveland & Hively, 2009;Kavanaugh et al., 2005;Son & Lin, 2008).
The possibility to expand interpersonal network via the mobile phone is greater than in any other media (Park & Karan, 2014). In mobile communication, even those who usually use their mobile phones to interact with intimates cannot avoid encountering diverse opinions or people beyond their strong-tie networks. Therefore, this study predicts that a large network size may provide a significant opportunity to develop civic-mindedness and to engage in civic affairs even to the people who use their mobile phones primarily for relational or recreational purposes. Thus, the current study hypothesizes: H6: The network size in mobile communication will moderate the impact of relational or recreational uses of the mobile phone on civic engagement, with a larger network size increasing the likelihood of influencing civic engagement.

Data Collection
This article relies on online survey data collected from the residents of South Korea between May 20 and June 20, 2013. In order to assure the representativeness of the sample, data were collected via a stratified quota sampling method. Before doing stratification, this study constructed a sampling frame of 90,000 using a random number generator from the 2012 voter registration data of the Korea Election Management Commission (2013). Out of the frame, a total of 3,500 adults were selected using a four-way cross-classification system (age x gender x education x income). This stratified quota sampling is differs from conventional probability sampling procedures, yet produces highly comparable data (Eveland & Shah, 2003;Putnam, 2000).
The target participants were provided the present survey's URL via emails. To increase the response rate, subjects were entered into a raffle for 10 $20 gift cards, and two reminder emails were sent out. The invitation obtained 1,260 responses. The response rate was 36% (AAPOR, 2009).
Online Journal of Communication and Media Technologies Volume: 7 -Issue: 2 April -2017
The original items were revised to fit the South Korean context. Responses were coded on a 5-point scale ranging from 1 (never) to 5 (very frequently). For informational use, respondents were asked during the last one month over their mobile phones how often they (1) visited online to read opinions of others; (2) listened to radio news or watched TV news; (3) visited online to share news items; (4) called others to discuss social matters; and (5) visited online to express their opinions about issues. After a test of indicator reliability, (4) and (5) were dropped in the model, remaining the first three items (α = .79, M = 2.27, SD = 1.01). For relational use, two items were employed: (1) called friends or family to be social; and (2)   To measure formal engagement, the study asked respondents how frequently they participated in the following activities during the past three months on a 5-point scale ranging from 1  et al., 2000). Convergent validity is the extent to which a measure correlates positively with alternative measures of the same construct. Convergent validity is assessed using three criteriaindicator reliability, composite reliability, and average variance extracted (AVE). To test the discriminant validity, item cross-loadings and the Fornell-Larcker criterion must be examined. Table 4 shows that most outer loadings were above the .70 for their respective construct, suggesting good indicator reliability. Three items from usage patterns were dropped from subsequent analyses since their loadings were lower than .50. The composite reliabilities of all factors ranged from .70 to .89, which exceeded the recommended the threshold value of .70. In addition, all latent variables indicated good AVE scores. Hence, all three conditions of convergent validity were met (Table 2).

Scale Validation in the Measurement Model
For the discriminant validity, cross-loadings and inter-correlations were calculated. As shown in Table 3, all AVE scores in the matrix diagonals were larger than the off-diagonal correlations, satisfying the Fornell-Larcker criterion. In the test of cross-loadings, the magnitude of the factor loading of each item on its corresponding construct exceeded the magnitude of its cross-factor loadings. Hence, the discriminant validity criteria were also satisfied.
[ For discriminant validity, diagonal elements should be larger than off-diagonal elements. For instance, in a PLS path model without the mediator variable, a positive direct effect would become smaller or insignificant after the inclusion of the mediator variable. Figure 3 shows that when network size is introduced as a mediator, the direct impact of informational use on informal engagement becomes significantly smaller (from β = .15 to β = .11), whereas its indirect effect via network size achieves a point estimate of .10 (a*b). Since all CIs do not contain zero (Table 5), the indirect effect is significant. This means that network size partially mediates the influence of informational use on informal engagement. When network size is introduced as a mediator, the direct association between informational use and formal engagement becomes nonsignificant (from β = .11 to β = .08), whereas its indirect effect via network size achieves a point estimate of .09 (a*c). Since all CIs do not contain zero, the indirect effect is significant.
This means that network size fully mediates the influence of informational use on formal engagement. Consequently, H5 was supported.
[ Figure 3 and Table 5] Mediating effects of network size between informational use and civic engagement.
Note: Cell entries are standardized path coefficients. The effects of control variables (age, gender, education, income, political interest, and ideological extremity) on endogenous and exogenous variables have been controlled. The coefficients in parentheses indicate direct effects. ns = not significant, * p < .05, ** p < .01, *** p < .001. Note: * p < .05, ** p < .01, *** p < .001; t values in parenthesis. Point estimates were obtained by multiplying the direct effect of Informational use on network size and the direct effect of network size on informal engagement and formal engagement (.24 x .40 = .10; .24 x .36 = .09).
This study also investigated whether network size moderates the impact of mobile phone use on civic engagement. To this end, interaction terms were formulated by multiplying the corresponding indicators of the predictor and the moderator construct (Chin, Marcolin, &