Chapter 3. Methodology
3.3 Measurement
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
3.3 Measurement
3.3.1 Outcome Variable
Adapted from the past research, attitude certainty was measured by asking the respondents how certain they were of their attitudes toward an issue or objects (Fazio
& Zanna, 1978; Tormala & Petty, 2002). In Petrocelli, Tormala and Ruckers’ research (2007), they conducted an experiment by using a single item to measure attitude certainty on specific issues. Respondents were asked to rate the extent to which they are certain about a specific issue with a nine-point scale (1 = not certain at all; 9 = very certain). Previous research operationalized the degree of certainty to measure attitude certainty. However, the current research aimed to examine whether individuals can develop a certain attitude. Therefore, this study employed the conceptualization of attitude certainty by asking respondents which statements do they agree with the selected issues.
Attitude Certainty
Attitude Certainty (Kr20 = .557, M = 3.55, SD = .82) can be evaluated by implicit
or explicit measures. An explicit method was implemented by asking respondents how certain they are with their attitudes (Clarkson et al., 2011). In comparison, an implicit method was measured with structural parameters (Ryffel et al., 2014). In the current study, attitude certainty was measured by to what extent the participants can develop a certain attitude. Participants were asked which statement they consider to agree or disagree with four controversial social issues. Issues included nuclear power generation, air pollution led by the increasing coal-fueled power generation, marriage for homosexuality, pension reform for retired teachers and veterans. The response categories were: 1=agree with the issue, 2=disagree with the issue, 3= agree with both
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
sides, 4=disagree with both sides, 97=I don’t know. Responses for agreement and disagreement were coded as “1”, which referred to a certain attitude. On the contrary, ambivalent responses such as “both agree,” “both disagree,” “don’t know” were recoded as “0”, which represented those attitudes were uncertain. Attitude certainty was then measured by summing those four items.
3.3.2 Independent variable
Heterogeneous Information Exposure
Heterogeneous information exposure on SNS (M = 2.82, SD = .88) served as the
only independent variable in this study. Respondents were asked, “How often do you see the information of political or public issues on social media that differ from your position.” A five-point scale (1=never, 2=seldom, 3=sometimes, 4=usually, 5=always, 94=it depends, combined with 3) was used to indicate the level of heterogeneous information exposure on SNS.
3.3.3 Mediator
Heterogeneous discussion
To measure heterogeneous discussion (M = 1.54, SD = .75), this study adopted the operational measurement of cross-cutting discussion based on the previous research (Shih, Scheufele & Brossard, 2013; Kim, Hsu & de Zuniga, 2013). One critical question was chosen by asking, “During the past six months, when you saw your friend’s Facebook post or comment have different positions on political or public issues, how often do you raise your own opinions?” Responses were on a five-point scale, with higher points indicating higher frequency (1=never, 2=seldom, 3=sometimes, 4=usually, 5=all the time, 94=it depends, combined with 3).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
3.3.4 Moderators
Selective avoidance on SNS
According to the previous research, selective avoidance (M = 1.3, SD = .57, r = .4) was assessed by unfriending and hiding others’ posts and comments (Zhu, Skoric, &
Shen, 2016). The former referred to aggressively removing the online connection between the dyad without informing the other person. Simultaneously, the latter served as a passive act to hide content from Facebook News Feed without terminating the relationship. These two items were thus combined to a scale. To measure these concepts (John & Gvirsman, 2015; Yoo, Ng & Johnson, 2018), participants had to rate the following statements respectively on a five-point scale (1 = never to 5= all the time).
Questions were listed below.
(1) “During the past six months, when you saw your friend’s Facebook post or comment have different positions on political or public issues, how often do you hide the post or comment?”
(2) “During the past six months, when you saw your friend’s Facebook post or comment have different positions on political or public issues, do you unfriend or delete that friend?”.
3.3.5 Control variables
To avoid a confounding effect, demographic variables such as age, gender, and educational level were included as controls in the analysis. Firstly, to measure age, respondents were asked to indicate their birth year. In terms of gender, male was coded as 1, female as 0. The ratio of men to women was near 1:1, with slightly more women (52.7%). Educational level was calculated with a seven-point scale (1=none or self-study, 2=elementary school, 3=vocational/ junior high school, 4=senior high school and cadet school, 5=open university, junior college of military/police, and junior college,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
6=university, college of military/police, 7=master’s degree and above) (M=5.2, SD=1.2).
Besides, drawing from the observation in Chen’s research (2019), it is evidenced that people who do not hold any religious beliefs and whose partisanship falls to the
“pan-green” (DPP) revealed higher supportiveness to controversial social issues such as same-sex marriage. Therefore, party identification and religion were included for control purposes in this research. Party identification was distributed in three categories:
“pan-green”, “independent”, and “pan-blue”. Since this variable only consisted of three categories, which were not suitable to view as a continuous variable. For the purpose of data analysis, party identification was created as two dummy variables, with “pan-green” as the reference group. Religious belief was recoded as a nominal variable, with 0 indicating no religion, 1 referring to any kind of religious beliefs.