A total of 342 respondents filled out the survey completely. Respondents who did not finish the survey (272 incomplete responses) were excluded from analysis. The high dropout rate might be explained by the relatively long questionnaire – the mode of completion time was 12.1 minutes. In total 614 participants started filling out the survey in the data collection period between May 18th 2020 and June 10th 2020. Three outlier responses (respondents with inconsistent answers considering their demographics) excluded after considering individual responses and histograms thus resulting in 339 remaining responses.
The age variable was re-coded from continuous variable to categorical variable to make analyses more natural to read. The age groups were defined as 1 – 18-24, 2 – 25-34, 3 – 35-44, 4 – 45-54, 5 – 55-64, and 6 – 65+.
Table 2. Descriptive statistics. (N = 339)
Variables Frequency Percentage
(n) (%)
‧
Bachelor’s degree 119 35.1
Master’s degree 127 37.5
Doctorate 24 7.1
Of the 339 participants 177 (52%) were male, 161 (42%) female, 1 (0,3%) other with the average age of 39 ( SD = 12.53). Respondents were of 22 different nationalities, with 131 (38%) respondents from Hungary, 96 (28%) from the United States of America, 20 (6%) from the United Kingdom, 15 (4.5%) from Taiwan, and 11 (3.2%) from Australia.
Looking at highest obtained degree of education, 79.7% of respondents was college educated with 37.5% holding a Master’s degree, 35.1% a Bachelor’s degree, and 7.1% a Doctorate. 13.3% had some college education but no degree, and high school diploma was the highest degree of education for 7.1% of respondents.
4.2 Media consumption habits
To better understand participants’ media consumption, the survey asked them about their main news source, the frequency and time they spend with online news media, and international news media specifically. They were also asked about their social media use (i.e., time spent on social media, frequency of using social media, platforms used).
Table 3. Main news source (N = 339)
Variables Frequency Percentage
(n) (%)
Main news source
Traditional news (e.g. TV, print
newspapers, radio) 44 13.0
‧
source, and 23.1% of all primary traditional news media users coming from this age group.Those between 25 and 34 are the biggest age category in both news sites, news aggregators, and social media, with news sites (main news source for 54.7% of this age group) being the most popular. Social media as a primary news source is the most popular amongst the youngest age group (18–24) with 42.6% saying it’s their primary source for news. This preference for social media decreases with each age group, with only 4.8% of those between 55–64, and 0% or those above 65 calling it their main source of news.
Table 4. Main news source per age groups. Crosstabulation.
Main news source
‧
Variables Frequency Percentage
(n) (%)
Frequency of online news consumption
Never 2 0.6
Less than twice a month 6 1.8
Two or three times a month 5 1.5
About once a week 16 4.7
Two or three times a week 35 10.3
Almost everyday/Everyday 272 80.2
Daily time of online news consumption
Less than 30 minutes 101 29.8
30-60 minutes 140 41.3
1-2 hours 57 16.8
2-3 hours 20 5.9
3+ hours 15 4.4
Frequency of international news consumption
‧
‧
Almost everyday/Everyday 212 62.5
Daily time of international news consumption
Less than 30 minutes 181 53.4
30-60 minutes 106 31.3
1-2 hours 29 8.6
2-3 hours 11 3.2
3+ hours 3 0.9
Respondents have proven to be avid news readers with 80.2% saying they read news online most every day and 62.5% reading international news stories most every day. 71.1%
says they spend between 0–60 minutes a day reading online news, and 53.4% claims to spend less than 30 minutes a day consuming international news stories.
As for SNS platforms used, participants used an average of 3.38 platforms, with Facebook being the most widely adapted (95.6%), followed by Youtube (68.1%) and Instagram (60.5%), Linkedin (37.2%), Twitter (36.6%), Line (12.4%), Snapchat (10.3%), Tiktok (5.9%) and other unlisted platforms such as Tumblr or Reddit (10.6%).
Table 6. Sample’s social media use habits (N = 339)
Variables Frequency Percentage
(n) (%)
Frequency of social media use
I don't use social media 2 .6
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Less than 30 minutes 27 8.0
30-60 minutes 88 26.0
1-2 hours 95 28.0
2-3 hours 77 22.7
3+ hours 45 13.3
Table 7. Sample’s social media platform use (N = 339)
SNS platforms Frequency Percentage
(n) (%)
Facebook 324 95.6
Instagram 205 60.5
Youtube 231 68.1
Twitter 124 36.6
Linkedin 126 37.1
Snapchat 35 10.3
Tiktok 20 5.9
Line 42 12.4
Other 36 10.6
4.4 Generalized media trust
Generalized media trust (M = 3.05, SD = .786) adapted from Kohring and Matthes (2007) measured trust in the selectivity of topics ( M = 3.13, SD = .97), selectivity of facts (M = 2.91, SD = .92), accuracy of depictions ( M = 3.26, SD = .87), and journalistic assessment (M = 2.90, SD = .98) in a total of 16 items, with answers ranging on a 5-point Likert scale (1 – Strongly disagree to 5 – Strongly agree)
Running One-way ANOVA shows there is a significant difference between preferred main news sources in terms of generalized media trust, F(4, 334) = 4.284, p < .01. Post hoc analysis using Tukey HSD suggests that respondents in whose main news source is news websites (M = 3.24, SD = .73) have higher generalized media trust than those whose main
‧
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧
Running One-way ANOVA shows there is a significant difference between age groups in terms of perceived credibility of Facebook as a news source, F(5, 333) = 4.70, p <
.001. Post hoc analysis using Tukey HSD suggests that respondents in the 45-54 age group (M = 1.89, SD = .69) perceived Facebook as a less credible news source than those in the 18-24 (M = 2.58, SD = .79), and in the 25-34 (M = 2.46, SD = .80) age groups.
Running One-way ANOVA shows there is a significant difference between age groups in terms of perceived credibility of the stimulus material, F(4, 334) = 7.48, p < .001.
Post hoc analysis using Tukey HSD suggests that respondents in the 18-25 ( M = 3.07, SD = .66) and 25-34 (M = 2.99, SD = .78) age groups perceived the stimulus material as a more credible than those in the 45-54 ( M = 2.32, SD = .98), and the 65+ (M = 2.17, SD = 1.02) age groups.
Running One-way ANOVA shows there is a significant difference between main news sources in terms of perceived credibility of the main news source, F(4, 334) = 5.29, p <
.001. Post hoc analysis using Tukey HSD suggests that respondents whose main news source is news websites (M = 3.32, SD = .97) perceive their main news source as more credible than those whose main news source is social media (M = 2.74, SD = .94).
Running One-way ANOVA shows there is a significant difference between main news sources in terms of perceived credibility of the stimulus material, F(4,334) = 3.08, p <
.05. Post hoc analysis using Tukey HSD suggests that respondents whose main news source is news websites (M = 2.95, SD = .79) perceived the stimulus material as more credible than
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4.6 Partisanship
To assess partisanship (M = 3.81, SD = 1.86) participants gave their opinion on the 2020 U.S.-Iran military conflict on a 7-point Likert scale adapted from Gunther and Schmitt (2004). Moderates (Likert scale 3–5, N = 170), anti-partisans (Likert scale 1–2, N = 101), and pro-partisans (Likert scale 6–7, N = 68) are differentiated.
Running One-way ANOVA shows there is a significant difference between groups in terms of time spent consuming online news, F(2, 330) = 7.25, p < .001. Post hoc analysis using Tukey HSD suggests that pro-partisan respondents (M = 2.65, SD = 1.16) spend more time consuming online news than moderates (M = 2.02, SD = 1.05) and anti-partisans (M = 2.00, SD = .92).
Running One-way ANOVA shows there is a significant difference between groups in terms of frequency of consuming international news, F(2, 331) = 3.89, p < .05. Post hoc analysis using Tukey HSD suggests that anti-partisan respondents (M = 5.58, SD = .86) consume international news more frequently than moderates (M = 5.22, SD = 1.21).
Running One-way ANOVA shows there is a significant difference between groups in terms of time spent consuming international news, F(2, 327) = 4.10, p < .05. Post hoc analysis using Tukey HSD suggests that pro-partisan respondents (M = 1.89, SD = 1.07) spend more time consuming international news than moderates (M = 1.55, SD = .80).
Running One-way ANOVA shows there is a significant difference between groups in terms of frequency of looking at social media, F(2, 336) = 3.48, p < .05. Post hoc analysis using Tukey HSD suggests that anti-partisan respondents (M = 5.10, SD = .89) look at social media more often than moderates (M = 4.75, SD = 1.14).
4.7 Testing the Hostile Media Effect
For H1a and H1b respondents who were shown the objective journalist role condition stimulus material are considered, as this serves as a balanced news story.
Hypothesis 1a states “Pro-attack participants reading a balanced news story on the 2020 U.S.-Iran military conflict will perceive the content to be more hostile toward their personal position than moderates, those who are neither pro-attack or anti-attack.”
‧
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧
Running an independent-samples t-test shows that there is significant group difference in terms of hostile media perception between pro-attack participants and moderate participants when they were shown a balanced news story, t (112) = 2.691, p < .05).
Specifically, pro-attack partisan respondents (M = 2.30, SD = 1.03) perceived the objective stimulus material as more anti-attack than moderate respondents did ( M = 2.81, SD =.84).
Therefore, H1a is supported.
Hypothesis 1b states “Anti-attack participants reading a balanced news story on the 2020 U.S.-Iran military conflict will perceive the content to be more hostile toward their personal position than moderates, those who are neither pro-attack or anti-attack.”
There is a significant group difference in terms of hostile media perception between anti-attack participants and moderate participants when they were shown a balanced news story, ( t(108.59) = -2.00, p < .05). Specifically, anti-attack respondents (M = 3.11, SD = .88) perceived the objective stimulus material as more pro-attack than did moderate respondents (M = 2.81, SD = .84). H1b is supported.
For H2a and H2b we consider respondents who were shown the advocate journalist role condition stimulus material, as this serves as a slanted news story.
Hypothesis 2a states “Anti-attack participants will perceive a slanted news story to be more biased in favor of the pro-attack position than pro-attack participants.”
There is no significant partisan group difference in terms of perceived pro-attack bias, therefore H2a is not supported.
Hypothesis 2b states “Pro-attack participants will perceive a slanted news story to be more biased in favor of the anti-attack position than anti-attack participants.”
Running an independent-samples t-test shows there is a significant partisan group difference in terms of perceived anti-attack bias, t (87.89) = 4.20, p < .001. Specifically, pro-attack partisans (M = 1.46, SD = .83) perceived the post to be significantly more in favor of the anti-attack position than anti-attack partisans (M = 2.26, SD = .941). H2b is supported.
When asked to estimate the percentage to which the anti-attack slanted post was biased against the military actions, pro-attack participants (M = 71.05, SD = 32.94) answered with higher percentages than anti-attack participants (M = 60.87, SD = 27.87), but the difference did not reach significant level (p = .135).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4.8 Testing general media trust in partisans
Hypothesis 3 states “Participants who have a stronger opinion on the 2020 U.S.-Iran military conflict will show lower levels of general media trust than participants who do not have a stronger opinion.”
Running an independent-samples t-test shows there is a significant partisan group difference in terms of generalized media trusts, t (337) = 2.16, p < .05. Specifically, moderates (M = 3.14, SD = .58) had higher generalized media trust than partisans (M = 2.96, SD = .94). H3 is supported.
Results are also significant differentiating for partisan groups. Running ANOVA shows there is a significant difference between groups in terms of generalized media trust, F(2, 336) = 15.80, p < .001. Post hoc analysis using Tukey HSD suggests that pro-partisan respondents (M = 2.59, SD = 1.05) have lower generalized media trust than moderates (M = 3.14, SD = .58) and anti-partisans (M = 3.20, SD = .78).
4.9 Testing the impact of journalistic roles
Hypothesis 4 states “Advocacy journalism will result in higher perception of bias in a news post on Facebook than objective-neutral journalism.”
Running an independent-samples t-test shows that there is a significant group difference in terms of hostility perception ( t = 7.23, p < .001). Specifically, those who were shown the advocacy journalism stimulus material (M = 2.05, SD = 1.01) perceived more bias than those who were shown the objective-neutral stimulus material (M = 2.81, SD = .92). H4 is supported.
4.10 Partisanship and Facebook’s news credibility
Research question 1 seeks to answer what relationship is there between partisanship and Facebook’s perceived credibility as a news source. It asks: “What is the relationship between participants’ strength of opinion about the 2020 U.S.-Iran military conflict and their perceived credibility of Facebook as a news source?”
‧
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧
Running an independent-samples t-test shows significant results. There is a significant group difference in terms of perceived credibility of Facebook as a news source, t (328.17) = 3.43, p < .001. Specifically, participants with a stronger opinion on the U.S.-Iran military conflict (M = 2.18, SD = .90) perceive Facebook to be a less credible news source than moderate participants (M = 2.50, SD = .77).
Differentiating between partisan groups also shows significant results. Running ANOVA shows there is a significant difference between groups in terms of perceived credibility of Facebook as a news source, F(2, 336) = 12.67, p < .001). Post hoc analysis using Tukey HSD suggests that pro-partisan respondents (M = 1.90, SD =.87) perceive Facebook as a news source as less credible than moderates ( M = 2.50, SD = .77) and anti-partisans (M =2.37, SD = .88).
4.11 Social news content segments and perceived bias
Research question two asks, “What segments of the news content shared as a social media post on Facebook contribute to participants’ perception of bias?” This question was addressed in a multiple choice item where participants were asked to check all segments of the social media news post that caused them to believe the content was biased. This question will be analyzed separately for the objective-normative journalist role and the advocacy journalist role, since the two condition groups saw a different social media news post.
For participants under the objective-normative role condition (N = 166), the average hostile media perception was MHMP= 2.8155. In the item asking about post segments, after reading an instruction (“Please check all segments of the post that caused you to believe that this post was biased in favor of either side. Choose as many as you think applies”) participants answered in the following way:
Table 8. Objective-normative condition. Segments contributing to bias perception. (N =166)
Selected segments Frequency Percentage
(n) (%)
The text written by the journalist. (“Iran’s top security and intelligence
‧
The title of the article. (“The US Orders Strike…”)57 14.2%
The description of the article. (“Suleimani was planning…”)
82 20.4%
The photo accompanying the article. 59 14.7%
The number of likes on this post. 18 4.5%
The number of comments on this post. 14 3.5%
The number of shares on this post. 15 3.7%
Something else (please specify) 19 4.7%
None of the above 19 4.7%
Total 401 100%
Categorizing these answers in Walther and Jang’s (2012) segmentation of social content we get the following:
Table 9. Objective-normative condition. Segments contributing to bias perception. (N =166)
Type of segment Frequency Percentage
(n) (%)
Proprietor content 316 78.7%
Aggregate user
For participants under the advocacy role condition ( N = 165), the average hostile media perception was MHMP= 2.0485, which means, their bias perception was slightly higher than that of participants in the other group. In the item asking about post segments, after reading an instruction (“Please check all segments of the post that caused you to believe that this post was biased in favor of either side. Choose as many as you think applies”) participants answered in the following way:
Table 10. Advocacy condition. Segments contributing to bias perception. (N =165)
Selected segments Frequency Percentage
(n) (%)
‧
The text written by Reuters.("Breaking News: The U.S.
confirmed...")
34 8.2%
The title of the article. (“The US Orders Strike…”)
The number of likes on this post.
27 6.5%
The number of comments on this post.
19 4.6%
The number of shares on this post.
Categorizing these data in Walther and Jang’s (2012) segmentation of social content we get the following:
Table 11. Advocacy condition, Content segments contributing to bias perception (N =165)
Type of segment Frequency Percentage
(n) (%)
Proprietor content 322 77.8%
Aggregate user representation 76 18.3%
Something else (please specify) 12 2.9%
None of the above 4 1.0%
Total 414 100%
4.12 Comparative analysis
A 2 by 3 comparative analysis regarding research condition (objective or advocacy journalist role) and partisanship (moderates, pro-partisans, anti-partisans) was completed on the hostile media effect, general media trust, and Facebook’s credibility as a news source in order to
‧
For the hostile media effect (measured on a Likert scale 1–5, from strongly biased against the military actions to strongly biased in favor of the military actions ), under the objective-normative journalist role condition moderates scored a mean of 2.81, with a standard deviation of .84; anti-partisans (those who did not support the military actions) scored a mean of 3.11 with a standard deviation of .88; and pro-partisans (those who did support the military actions) scored an average of 2.30 with a standard deviation of 1.03.
Under the advocacy condition moderates scored a mean of 2.20, with a standard deviation of 1.03; anti-partisans (those who did not support the military actions) scored a mean of 2.26 with a standard deviation of .94; and pro-partisans (those who did support the military actions) scored an average of 1.46 with a standard deviation of .83.
Running two-way ANOVA shows no significant interaction effects. Results are significant for the main effects. Simple main effects analysis showed that pro-partisans perceived the content as significantly more biased against the military actions than moderates (p < .001), and anti-partisans (p < .001).
For generalized media trust (measured on a Likert scale 1–5, from strongly disagree to strongly agree), under the objective-normative journalist role condition moderates scored a mean of 3.12, with a standard deviation of .57; anti-partisans (those who did not support the military actions) scored a mean of 3.15 with a standard deviation of .78; and pro-partisans (those who did support the military actions) scored an average of 2.69 with a standard deviation of 1.06. Under the advocacy condition moderates scored a mean of 3.17, with a standard deviation of .59; anti-partisans (those who did not support the military actions) scored a mean of 3.27 with a standard deviation of .78; and pro-partisans (those who did support the military actions) scored an average of 2.51 with a standard deviation of 1.04.
Running two-way ANOVA analysis shows no significant interaction effects. Results are significant for the main effects. Simple main effects analysis showed that pro-partisans had significantly lower generalized media trust than moderates ( p < .001), and anti-partisans (p < .001).
For Facebook’s credibility as a news source (measured on a Likert scale 1–5, from strongly disagree to strongly agree), under the objective-normative journalist role condition moderates scored a mean of 2.54, with a standard deviation of .80; anti-partisans (those who
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
a standard deviation of .58. Under the advocacy condition moderates scored a mean of 2.48, with a standard deviation of .75; anti-partisans (those who did not support the military actions) scored a mean of 2.35 with a standard deviation of .81; and pro-partisans (those who did support the military actions) scored an average of 1.91 with a standard deviation of 1.05.
Running two-way ANOVA analysis shows no significant interaction effects. Results are significant for the main effects. Simple main effects analysis showed that pro-partisans had significantly lower trust in Facebook as a news source than moderates ( p < .001), and anti-partisans (p < .001).
In conclusion this study observes that there are no significant interaction effects between condition and partisanship on the hostile media effect, general media trust, and Facebook’s credibility as a news source, however, in each case significant main effects of partisanship were observed, and it can be stated that pro-partisans, those who supported the U.S. military actions against Iran, perceived more hostility, has lower generalized media trust, and saw Facebook as a less credible news source than did moderates or anti-partisans, and this difference was significant.