Social Media and Rigid Beliefs: Evidence from - Impeachment of the President Yong Suk Lee

 
CONTINUE READING
Social Media and Rigid Beliefs: Evidence from - Impeachment of the President Yong Suk Lee
Social Media and Rigid Beliefs: Evidence from
         Impeachment of the President

                 Yong Suk Lee

                    August, 2018
               Working Paper No. 1021
Social Media and Rigid Beliefs: Evidence from - Impeachment of the President Yong Suk Lee
Social Media and Rigid Beliefs: Evidence from Impeachment of the President

Yong Suk Leea1,
a
          Freeman Spogli Institute for International Studies, Stanford University, Stanford, CA 94305

Abstract
Controversial news with little to no factual support has been proliferating in recent years. Such
news has spread widely via social media, where individuals can easily create and distribute news.
Using the news and events surrounding the impeachment of the former South Korean president
Park Geun-hye, I examine (1) the relationship between social media use and beliefs in
controversial news and (2) the relationship between protest participation and rigid beliefs based
on social media use. The results indicate that individuals who use social media as their primary
source of news are significantly more rigid in their beliefs in controversial news. Further, those
who strongly believe in controversial news and are more consistent or extreme in their
ideological views are more likely to participate in (more) protests. Moreover, this pattern is
significantly stronger for those who primarily get their news from social media. Protests can
have real-world impacts on public opinion and the policy-making process. Furthermore, a
relatively small number of people can effectively mobilize political protests. Though protest
participation is an individual choice, it is influenced by others and initially coordinated by a
small number of people. The findings of this paper suggest that social media may allow a small
group of people, potentially with more polarized and rigid beliefs, to have a disproportionately
large impact on public opinion and policy-making in modern democracies.

Keywords: social media, news, rigid beliefs, polarization, protests, impeachment

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
1	
  To	
  whom	
  correspondence	
  should	
  be	
  addressed.	
  Email:	
  yongslee@stanford.edu.	
  

	
                                                                                                                                                                                                                                 1	
  
Social Media and Rigid Beliefs: Evidence from - Impeachment of the President Yong Suk Lee
1. Introduction
Controversial news with little to no factual support has been proliferating in recent years (Allcot
and Gentzkow 2017; Bradshaw and Howard 2017; Lazer et al. 2018). Such news has spread
widely via social media, where individuals can easily create and distribute news. Social media
enables people to share news with like-minded people, potentially creating an echo-chamber
effect (Sunstein 2001; Pew Research Center 2014; Bakshy et al. 2015). Moreover, individuals
with extreme political views can use social media to not only consume, but also distribute
selective news. The first objective of this paper is to examine whether people who use social
media as their primary source of news believe more strongly in controversial news. Studies have
found that people who hold strong convictions are less likely to find a middle ground, reconcile
on contentious issues, and solve problems creatively (Carnevale and Probst 1998; Mutz 2006;
Montpetit 2012). Moreover, studies have found that false information travels faster than truthful
information on social media (Vosoughi 2018) and that increased exposure to fake news increases
people’s tendency to believe in such news (Pennycook et al. 2018). The second objective of this
paper is to examine social media users’ protest participation based on the rigidity or ideological
consistency of their beliefs. Studies have found that social media facilitate political protests
(Bartels 1993; Shirky 2001; Bond et al. 2012; Tufekci and Wilson 2012; Enikolopov, Petrova,
and Macarin 2017) and that protests can shape public opinion and political outcomes (Madestam
et al. 2013; Acemoglu et al. 2018). The public and the government have become increasingly
concerned about the effects of social media on public opinion and behavior and, ultimately,
political outcomes (Pew Research Center 2016, 2018). Since the decision to select one media
source over another is an individual choice, understanding the beliefs and protest behaviors of
individuals who use social media as their primary source of news could shed light on how the
prevalence of social media might shape public opinion and political outcomes in modern
democracies.
       This paper examines these questions using the news and events surrounding the
impeachment of Park Geun-hye, South Korea’s previous president. Park was impeached due to
influence peddling by her close friend who had no role in the government. News related to the
scandal was widely covered in South Korea. However, there were also many controversial news
stories related to Park Geun-hye. As these news reports came out, people took to the streets and
demanded that Park step down. The opposition party eventually put forward an impeachment

	
                                              2	
  
Social Media and Rigid Beliefs: Evidence from - Impeachment of the President Yong Suk Lee
proposal, and the Constitutional Court unanimously voted for impeachment. However, the
impeachment ruling did little to confirm the factual accuracy surrounding most of the circulating
news. Other than confirming one account of abuse of power, the Court explicitly stated that there
was insufficient evidence of the accusations brought forward against Park (Lim 2017).
Consequently, some of the people who initially believed the controversial news may have
become more ambiguous after the impeachment rulings.
        I conducted an online survey one week before the impeachment ruling and followed up
with a post-impeachment survey after one month. To measure the rigidity of people’s beliefs, I
collected their pre- and post-impeachment beliefs on four news items related to Park Geun-hye. I
focused on news with little factual support and on which people may have adjusted their beliefs
following the impeachment ruling. In the pre-impeachment survey, I asked whether the
participants believed each news item was “True” or “False.” In the post-impeachment survey, I
also allowed respondents to choose “Uncertain.” I find that people who use social media as their
primary source of news (hereafter: social media users) are significantly more rigid in their
beliefs—that is, they are less likely to change their beliefs concerning the news. Furthermore,
social media users are ideologically more polarized—their ideological views are more consistent
across issues and they have more extreme views. The relationship between social media and
rigid beliefs holds even when controlling for measures of ideological consistency, extreme
ideological views, political spectrum, individual characteristics, and personality traits. Focusing
on protests as the outcome, social media users are more likely to have participated in (more)
candlelight protests. Furthermore, social media users with more rigid beliefs or polarized views
are significantly more likely to participate in street protests.

2. The Context
2.1 The news and scandals
Park Geun-hye assumed office as South Korea’s first female president in February 2013. One of
Park’s appeals was her image as a clean leader. Almost all previous South Korean presidents or
their direct family members had been involved in some form of bribery or influence peddling
and were jailed. Park was never married; both of her parents were assassinated decades ago; and,
other than an estranged brother, she had no close family. Despite being elected the first female
president of East Asia, Park experienced a less popular presidency than initially hoped. Her

	
                                                 3	
  
Social Media and Rigid Beliefs: Evidence from - Impeachment of the President Yong Suk Lee
governing style was reclusive, and she tended to hire and consult with only a small number of
people (Doucette 2017; Kim, H. 2017). Her presidency experienced a major crisis with the
Sewol ferry disaster of April 16, 2014. More than 300 people, most of them young students, died
while the nation helplessly watched the ferry sink live on TV. Park made her first appearance
after seven hours of the sinking, and people wondered what the president was doing that day and
how she could have let such disaster unfold in front of everyone’s eyes (Fermin-Robbins 2018).
Her popularity declined substantially.
       In September 2016, news broke about the influence of Choi Soon-sil, Park’s long-time
friend, who had no official governmental position, over Park. Several news media reported that
Choi had access to confidential government documents and information. Evidence that Choi
edited and provided feedback on presidential addresses emerged. Continued news investigation
found that Choi established several foundations through which she yielded political and financial
influence. Taking advantage of her close ties with Park, Choi requested donations from major
conglomerates to fund foundation activities. Choi's foundations used those funds to buy horses
and fund her daughter's equestrian activities. Park was accused of being involved in this process,
as she met with many of the conglomerate owners around the same time. Furthermore, it was
revealed that Choi used her influence not only to send her daughter to a prestigious women’s
university, but also to reprimand a professor who gave her daughter low grades due to poor
attendance and performance. Many people were shocked and infuriated that someone with no
official government position could hold so much political and financial influence. In late October
2016, Park publicly acknowledged her close ties with Choi, and her approval ratings fell to an
all-time low of 5 percent (Harris 2017; Fendos 2017).
       Various news stories emerged around this time, especially on the internet and social
media. For instance, it became known that there was a private meeting between Park Geun-hye
and Jay Lee, Samsung’s vice chairman and heir apparent. News concerning the motives of the
meeting quickly spread. Some news stories reported that Lee asked Park to help in his succession
of Samsung and, in return, offered to donate to one of Choi’s foundations. Other news stories
proposed the opposite angle: that Park first asked Lee to support Choi’s foundation and
equestrian activities and, in return, offered to push the National Pension Fund to support Lee’s
cause at Samsung (Seo 2016; Kim, S. 2017; Fermin-Robbins 2018). Park’s response to the
Sewol ferry disaster was also the subject of many news stories. The ferry’s crash and sinking

	
                                              4	
  
Social Media and Rigid Beliefs: Evidence from - Impeachment of the President Yong Suk Lee
were televised in the morning, but Park did not make her first remark or appearance regarding
the disaster until seven hours after the disaster began. News stories speculated on what Park had
been doing during that “lost seven hours.” Some reported that Park was receiving a beauty (i.e.,
Botox) treatment. Others reported that Park had brought in a shaman to perform shamanistic
rituals, was doing her hair, or had simply slept in (Fermin-Robbins 2018, Lee 2018). Overall,
there was no shortage of news related to Park during this period, but most lacked concrete
evidence.

2.2 The candlelight protests
The public eventually took to the streets, and on October 29, 2016, the first candlelight protest
demanding Park’s resignation was held in downtown Seoul. The candlelight protests became a
weekly event and grew larger by the week, with crowd estimates ranging from 0.5 to 1.5 million
in the late November protests (Campbell 2018). Park eventually offered to resign as president on
November 12, 2016, and to let the National Assembly decide when she should step down to
ensure an orderly transfer of power (Choi 2016). However, many considered Park’s offer of
resignation as an easy way out to avoid impeachment. Thus, the candlelight protests continued to
grow into December of 2016, when the opposition party submitted the “President impeachment
proposal” to the National Assembly for violations of the constitution and the law. Six days later,
the National Assembly voted 234 to 56 to impeach President Park, and she was immediately
suspended from her executive powers. The Constitutional Court then had six months to decide
on a ruling.
        However, as the scandal headed towards the impeachment ruling, conservatives and pro-
Park supporters rallied against the candlelight protests supporting Park’s case. The counter-
protests by Park supporters—the so-called the Taegeukgi protests—argued that Park was not
guilty because she did not personally take any bribes. Though she made bad judgments regarding
Choi, they argued that her actions were not cause for impeachment. Moreover, the counter-
protests pointed out the lack of evidence for the accusations against Park. In the midst of
continued weekly protests on both sides, the Constitutional Court set March 10, 2017, as the date
for ruling.

2.3 The impeachment of Park Geun-hye

	
                                              5	
  
Social Media and Rigid Beliefs: Evidence from - Impeachment of the President Yong Suk Lee
The Court’s ruling was televised live on all major TV and cable stations and was streamed live
on media websites. The TV viewing rating was 37.7% (Yoon 2017). The Court first announced
that Park’s actions violated the constitution and the law and that the benefits of dismissing her
were overwhelming. It then ruled for Park’s impeachment with a unanimous 8-0 vote. As many
of the Constitutional Court Justices held conservative and right-of-center views, the unanimity of
the vote came as a surprise. The Court then discussed the four central claims of the impeachment
bill: (i) the abuse of power in the appointment of civil servants, (ii) the infringement of the
freedom of speech and press, (iii) the failure to protect the right of life and to faithfully carry out
presidential responsibilities during the sinking of Sewol ferry, and (iv) the abuse of power in
granting political power to Choi Soon-sil. The Court dismissed the first two claims for
insufficient evidence. It also dismissed the third claim because the constitutional duty to
faithfully carry out presidential responsibilities is a relative and abstract notion on which
impeachment cannot be adjudged. The Court did find Park guilty of the last claim: that Park
violated the constitution by allowing Choi to extort tens of millions of dollars from companies
and by sharing official documents with sensitive information (Lim 2017). In short, the Court
dismissed all claims except for those concerning Park’s involvement in Choi Soon-sil’s influence
peddling.
       How might the Constitutional Court’s rulings have affected people's beliefs about the
various news? Other than confirming Choi Soon-sil’s influence peddling, the court rulings did
not confirm or disprove most of the news that had been proliferating on social media. The rulings
could have rendered people less certain about their beliefs in some of the news.

3. Data and Key Variables
3.1 The pre- and post-impeachment surveys
I used a survey firm to recruit respondents and conduct the two online surveys. The first survey
was conducted between March 1 and March 7, 2017, shortly before the Constitutional Court’s
ruling on March 10, 2017. The post-impeachment survey was conducted between April 3 and
April 7, 2017. The utilized survey firm maintains a pool of panelists, and 2,000 individuals
across four age groups (20s, 30s, 40s, and 50s) and 1,000 college students were randomly
recruited from this pool. Excluding respondents who did not complete either survey or completed
in a time too short to be considered reliable yielded a panel of 2,749 respondents. Approximately

	
                                                6	
  
Social Media and Rigid Beliefs: Evidence from - Impeachment of the President Yong Suk Lee
46% were in their twenties due to the oversampling of college students, and the rest were
relatively evenly split between their thirties, forties and fifties. The gender split was pretty even,
with about 50.5% of the respondents being female. In both surveys, I asked about people’s
beliefs concerning news related to Park Geun-hye. In the pre-impeachment survey, I collected
information related to primary news sources, political leanings and worldviews, personal
characteristics, and protest participation. In the post-impeachment survey, I quizzed respondents
on the reason the Court upheld impeachment and collected additional personal characteristics,
including personality traits. The Appendix provides more detail on the survey implementation
and the other control variables used in the analysis. Appendix Table S1 presents the summary
statistics. I discuss the key variables below.

3.2 Beliefs in controversial news and belief rigidity
To measure the rigidity of people's beliefs in controversial news in the pre-impeachment survey,
I first asked whether the respondents believed whether each of the below news items was “True”
or “False.” These were news items that were widely circulated and familiar to most people, but
that lacked concrete evidence.
       News item 1. Park Geun-hye asked Samsung’s Jay Lee to support Choi Soon-sil's equestrian
       and winter sports efforts, and in return offered to help him with his succession of Samsung.
       News item 2. Park Geun-hye was receiving a beauty treatment on the day of the Sewol Ferry
       Disaster.
       News item 3. Park Geun-hye directly ordered the creation of a blacklist of artists.
       News item 4. Samsung's Jay Lee asked Park Geun-hye to help with his succession and
       provided a bribe amounting to tens of billions of dollars.
News item 1 concerned the private meeting between Park and Lee and whether Park directly
offered to help Lee with his succession. News item 4 was similar, but positioned Jay Lee as the
protagonist. The meeting between Park and Lee and Samsung’s donation to Choi's foundation to
support equestrian activities were facts; however, there was uncertainty concerning whether there
was a quid pro quo agreement and, if so, who initiated it (Seo 2016; Kim, S. 2017; Fermin-
Robbins 2018). News item 2 was one of the most widely circulated news items related to Park's
whereabouts during the day of the Sewol ferry sinking (Fermin-Robbins 2018; Lee 2018). News
item 3 concerned the Park administration's blacklisting of many left-leaning artists. Whether or

	
                                                  7	
  
Social Media and Rigid Beliefs: Evidence from - Impeachment of the President Yong Suk Lee
not Park directly ordered these artists blacklisted was unknown (Choi 2017). The Constitutional
Court's impeachment ruling did not resolve the uncertainty in any of the above news items.
       In the post-impeachment survey, I again asked whether people believed each news item
was “True” or “False,” but additionally allowed the option “Not certain whether the news is true
or false.” If the Constitutional Court's ruling had any effect, it would have been to reduce
people’s confidence in their beliefs. This seems to have been the case. Approximately 15% of
respondents changed their views on News item 1, and 20% of respondents changed their views
on the other three news items. As Figure 1 indicates, 10 to 20% of those who initially believed
that the news was “True” changed their beliefs to “Uncertain” after the impeachment ruling. In
general, more than 80% of those who initially believed the news was “False” changed their
beliefs to “Uncertain” or “True.” I measured the rigidity of people's beliefs by counting the
number of news items about which individuals held the same beliefs post-impeachment. About
8% of sample changed their beliefs on all four news items, 10% changed their beliefs on three
news items, 15% changed their beliefs on two news items, and 22.5% changed their beliefs on
one news item. A further 44.6% did not change their beliefs on any one of the four news items.
The majority of people who did not change their beliefs believed that the news items were true
(Appendix Figure S1).

3.3 Ideological consistency and extreme ideological beliefs
Many people have a mix of progressive and conservative viewpoints on different issues.
However, there are also people who have consistently liberal or conservative views on all
matters, and such people tend to be more involved in the political process (Pew Research Center
2014). The increase in the number of people with consistent ideological views characterizes the
recent rise in political polarization (Gentzkow 2016). I measured the respondents’ ideological
consistency using four questions that gauged their ideological worldview. I asked about the
extent to which they agreed or disagreed with the below statements on a scale of 1 to 10, where 1
corresponded to fully agreeing to the left and 10 corresponded to fully agreeing to the right. For
each question, the left represented the progressive viewpoint, and the right represented the
conservative viewpoint in contemporary South Korea (Appendix Figure 2 presents the
distribution for each question).

	
                                              8	
  
1. Luck and connection determine success vs. If you work hard you will eventually live a
          good life
       2. Income should be more equal vs. Income inequality is necessary to motivate people
       3. The government should guarantee the people’s livelihood vs. Each individual should
          support himself/herself
       4. Small and medium enterprises will drive future growth vs. Conglomerates will drive
          future growth
I constructed a measure of ideological consistency by adding the responses across the four
questions and then normalizing so that the midpoint was equal to zero. Appendix Figure 3
presents the histogram, which resembles a normal distribution, but with a concentration of
people at the left. The higher density at the left reflects of the young being more progressive in
Korea and the survey oversampling people in their 20s.
          I also constructed a measure of extreme ideological beliefs by summing the instances in
which each individual responded to the above questions with a 1 or 10. Though more than 65%
did not choose the most extreme views on any of the questions, approximately 5% chose an
extreme response to all four: mostly the most progressive response.

3.4 Protest participation
I asked whether people participated in the candlelight protests and, if so, how often. Over 31% of
the respondents had participated in the candlelight protests: 18% participated once, 6.3%
participated twice, 3.7% participated three times, 0.7% participated four times, and 2.6%
participated five or more times. Only 2.6% participated in the pro-Park (Taegeukgi) protests.

3.5 The reason for impeachment
Some of the respondents may not have updated their beliefs because they were not paying
attention to or interested in the rationale behind the Constitutional Court's ruling. To control for
this factor, I directly asked respondents about the reason the Constitutional Court decided to
impeach Park Geun-hye in the post-impeachment survey. Specifically, I asked:
Which of the below did the constitutional court adjudge as a violation of constitutional law and,
thus, use as a basis for impeaching Park Geun-hye? Select all that apply.
       A. Collusion with Choi and abuse of authority

	
                                               9	
  
B. Violation of duty to protect lives in the Sewol ferry disaster
       C. Abuse of authority in the appointment of civil servants
       D. Infringement upon the freedom of the press
The correct answer was A only, and 22.6% of the responses were correct. I used the results of
this quiz as a proxy for people’s attentiveness to the facts surrounding Park's impeachment.

3.6 Primary source of news
Respondents primarily received their news from TV (41.6%) and online websites (41%). Social
media (Facebook, Instagram, Twitter, Kakaostory, Band, etc.) was the primary source of news
for 8.7% of the respondents. Furthermore, 5.5% primarily got their news from friends and
family, 2.36% got their news from print media, and 0.8% got their news from the radio
(Appendix Figure S4). Though social media ranked third, TV and online websites were still the
respondents’ primary sources of news. The survey also asked how much the participants believed
in their primary news source. Those who chose TV had the highest level of trust, at 55%, while
those who chose social media had the lowest, at 40%.
          Those whose primary news source was social media were more consistent in their
ideological views and more likely to hold extreme ideological views. They identified as
progressive more than any other media users, were younger, and were less likely to correctly
answer the quiz on impeachment. They were also most likely to have participated in the
candlelight protests. Figure 2 illustrates how several key characteristics varied by primary news
source (Appendix Figure S5 presents additional characteristics). Figure 3 examines the
relationship between protest participation and rigid beliefs or ideological consistency by social
media use. As the different slopes indicate, protest participation by social media users was more
strongly associated with ideological consistency or rigid beliefs.

4. Results
4.1 Social media and rigid beliefs
Table 1 presents the relationship between social media and rigid beliefs. All regressions in Table
1 control for a rich set of variables using fixed effects for gender, birth year, education level,
region, income level, employment status, and occupation category (see the Appendix for details).
Even with these controls, column (1) indicates that social media users have significantly more

	
                                                  10	
  
rigid beliefs (i.e., they are more likely to hold the same beliefs on the news items post-
impeachment). Some people may have not changed beliefs because they were unaware of the
details behind the Constitutional Court’s decision. Hence, I control for whether participants
correctly answered the reason behind the Court’s impeachment ruling in column (2). Those who
answered the quiz correctly are significantly more likely to adjust beliefs post-impeachment;
however, the estimate on social media users remains positive and significant. I next control for
ideological consistency, extreme ideological views, and political spectrum in column (3).
Ideological consistency and extreme ideological views are both positively associated with more
rigid beliefs, and progressives are more likely to have rigid beliefs (i.e., to more strongly believe
that the news was true before and after impeachment). The estimate on social media drops to
0.213 but is still significant at the 5% level. I next control for people's pre-impeachment beliefs
on each item and how much they believed in their primary news source (column 5), as well as
for risk preference and personality traits (column 6). Controlling for these reduces the estimate
on social media users to 0.18, but the estimate remains statistically significant. Among the
personality traits, openness is significantly associated with less rigid beliefs, and
conscientiousness is associated with more rigid beliefs (Appendix Table S2). The estimates
indicate that social media users do not revise their beliefs concerning controversial news post-
impeachment by about 0.2 to 0.3 counts compared to others with similar observable
characteristics. This relationship is stronger than the effects of a standard deviation change along
the political spectrum, ideological consistency, or extreme ideological beliefs.

4.2 Social media and protest participation
I examine how protest participation relates to social media use and ideological polarization in
Table 2. Social media users are 13% more likely to have participated in the candlelight protests
(column 1). In the next three columns, I examine the relationship between protest participation
and ideological polarization or rigid beliefs by social media use. People who are more
ideologically polarized (i.e., more consistent or extreme in their ideologies) are significantly
more likely to participate in protests. Furthermore, this relationship is significantly stronger for
social media users (columns 2 and 3). The same pattern holds between protest participation and
rigid beliefs (column 4). The magnitudes of the estimates on the interaction terms indicate that
the relationships between protest participation and beliefs are twice as large for those who

	
                                               11	
  
primarily get their news from social media. The results from the logit regressions in Table
Appendix S3 show similar patterns.
       In addition to whether or not a respondent participated in the candlelight protests, I
examine the frequency of the participation (Columns 5-8). Since the survey capped protest
frequency at five, I present the tobit regressions. Social media users participated in more protests.
The relationship between ideological consistency and the number of protests is positive, and the
relationship is stronger for social media users. Again, similar patterns emerge for those with
more extreme or rigid ideological beliefs.

4.3 Is social media really different from other media?
Finally, I examine whether the patterns between social media and rigid beliefs or protest
participation are a unique social media effect or related to other media sources. Table 3 examines
all media source, with TV as the base omitted category. Relative to those who primarily get their
news from TV, social media users have significantly more rigid beliefs. Those who primarily get
their news from websites also exhibit higher levels of rigid beliefs, but the effect is much smaller
and statistically weaker. In terms of protest participation, only social media exhibits a positive
and significant relationship.

5. Conclusion
The results indicate that individuals who use social media as their primary source of news are
significantly more rigid in their beliefs in controversial news. Those who strongly believe in
controversial news and are more consistent or extreme in their ideological views are more likely
to participate in protests and in more protests. Moreover, this pattern is significantly stronger for
social media users. Research has found that protests have real-world impact on shaping public
opinion and the policy-making process (Madestam et al. 2013). Furthermore, a relatively small
number of people can more effectively mobilize political protests (Olson 1971). Though protest
participation is an individual decision, the choice to participate is influenced by others and
initially coordinated by a small number of people. Social media may facilitate a small group of
people, potentially with more polarized and rigid views, to have a disproportionately large effect
on public opinion and policy making in modern democracies.

	
                                               12	
  
Acknowledgements
I thank the team at Macromil Embrain for their valuable support. I thank Matthew Gentzkow,
David Yang, Emily Oster, Jesse Shapiro, and Munseob Lee for helpful input. Special thanks to
Joyce Lee for valuable suggestions throughout the project.

	
                                            13	
  
References

Acemoglu, D., Hassan, A., Tahoun, A. (2018). The Power of the Street: Evidence from Egypt’s
Arab Spring. Review of Financial Studies, 31(1): 1-42.
Allcott, H., & Gentzkow, M. (2017). Social Media and Fake News in the 2016 Election. Journal
of Economic Perspectives, 31(2): 211-236.
Bakshy, E., Messing, S., & Adamic, L. (2015). Exposure to Ideologically Diverse News and
Opinion on Facebook. Science, 348(6239): 1130–1132.
Bartels, L. (1993). Messages Received: the Political Impact of Media Exposure. American
political science review, 87(2): 267–285.
Bond R. M., Fariss C.J., Jones J.J., Kramer A.D.I., Marlow C., Settle J.E., and Fowler J.H.
(2012). A 61-Million-Person Experiment in Social Influence and Political Mobilization. Nature,
489: 295–298.
Bradshaw, S., & Howard, P. (2017). Troops, Trolls and Troublemakers: a Global Inventory of
Organized Social Media Manipulation. Oxford, UK: project on computational propaganda.
Campbell, C. (2018, November 28). Huge Numbers Demand the Ouster of South Korea's
President in a Fifth Week of Protests. Time. Retrieved from
http://time.com/4583033/south-korea-protest-demonstration-seoul-park-geun-hye-choi-
soon-sil/.
Carnevale, P.J., & Probst, T.M. (1998). Social Values and Social Conflict in Creative Problem
Solving and Categorization. Journal of personality and social psychology, 74(5): 1300-1309.
Choi, S. (2016, November 29 )	
  Park	
  Geun-­‐hye,	
  Embattled	
  South	
  Korean	
  President,	
  Says	
  She’s	
  
Willing	
  to	
  Resign.	
  Retreived	
  from	
  https://www.nytimes.com/2016/11/29/world/asia/park-geun-
hye-south-korea-resign.html	
  
Choi, S. (2017, July 27). 6 Ex-Officials in South Korea Are Sentenced for Blacklisting Artists.
The New York Times. Retrieved from https://www.nytimes.com/2017/07/27/world/asia/south-
korea-park-aides-artists-blacklist.html.
Doucette, J. (2017). The Occult of Personality: Korea's Candlelight Protests and the
Impeachment of Park Geun-hye. The Journal of Asian Studies, 76(4): 851-860.
Enikolopov, R., Petrova, M., & Macarin, A. (2017). Social Media and Protest Participation:
Evidence from Russia. Mimeo.
Fendos, J. (2017). The History of a Scandal: How South Korea's President Was Impeached. The
Diplomat. Retrieved from http://thediplomat.com/2017/01/the-history-of-a-scandal-how-south-
koreas-president-was- impeached/.
Fermin-Robbins, J. (2018). The Impeachment of South Korean President Park Geun-hye.
Carnegie Council for Ethics in International Affairs.
Gentzkow, M. (2016). Polarization in 2016. Mimeo.
Harris, B. (2017, March 9). Timeline: Downfall of Park Geun-hye. Financial Times.
Kim, S. (2017, June 12). There’s Still No Smoking Gun in the Samsung Bribery Trial. Bloomberg.
Retrieved from https://www.bloomberg.com/news/articles/2017-06-12/thirteen-weeks-into-
samsung-bribes-trial-there-s-no-smoking-gun.

	
                                                       14	
  
Kim, H. (2017). “Spoon Theory” and the Fall of a Populist Princess in Seoul. The Journal of
Asian Studies, 76(4): 839-849.
Lazer, D-M-J et al. (2018). The Science of Fake News. Science, 359, 1094-1096.
Lee, C. (2018, March 29). Was Park Geun-hye Asleep While Sewol Ferry Was Sinking. The
Korea Herald. Retrieved from http://www.koreaherald.com/view.php?ud=20180329000933.
Lim, H. (2017). A Closer Look at the Korean Constitutional Court’s Ruling on Park Geun-hye’s
Impeachment. Yale Journal of International Law.
Madestam, A., Shoag, D., Veuger, S., Yanagizawa-Drott, D. (2013). Do Political Protests
Matter? Evidence from the Tea Party Movement. Quarterly Journal of Economics, 128:1633–
1685.
Montpetit, E. (2012). Does Holding Beliefs with Conviction Prevent Policy Actors from
Adopting a Compromising Attitude? Political Studies, 60: 621-642.
Mutz, D.C. (2006). Hearing the Other Side: Deliberative versus Participatory Democracy.
Cambridge: Cambridge University Press.
Olson, M. (1971). The Logic of Collective Action: Public Goods and the Theory of Groups.
Cambridge, Massachusetts: Harvard University Press.
Pennycook, G., Cannon, T., & Rand, D. (2018). Prior Exposure Increases Perceived Accuracy of
Fake News. Forthcoming in Journal of experimental psychology: general.
Pew Research Center (2014). Political Polarization in the American Public.
Pew Research Center (2016). Many Americans Believe Fake News Is Sowing Confusion.
Pew Research Center (2018). Americans Favor Protecting Information Freedoms over
Government Steps to Restrict False News Online.
Seo, Y. (2016, November 23). Why Are South Koreans So Angry about Presidential ‘Choi-
Gate’? Here Are 4 Reasons. The Washington Post. Retrieved from
https://www.washingtonpost.com/news/worldviews/wp/2016/11/23/why-are-south-koreans-so-
angry-about-presidential-choi-gate-here-are-4-reasons.
Shirky, C. (2011). The Political Power of Social Media: Technology, the Public Sphere, and
Political Change. Foreign Affairs 90 (1): 28–41.
Sunstein, C. (2001). Echo Chambers: Bush v. Gore, Impeachment, and Beyond. Princeton
University Press.
Tufekci, Z., & Wilson, C. (2012). Social Media and the Decision to Participate in Political
Protest: Observations from Tahrir Square. Journal of communication, 62: 363–379.
Vosoughi, S., Roy, D., & Aral, S. (2018). The Spread of True and False News Online. Science,
359: 1146-1151.
Yoon, K. (2017, March 10). Impeachment Ruling TV Viewership Ratings. Yonhap News.
Retrieved from
http://www.yonhapnews.co.kr/bulletin/2017/03/10/0200000000AKR20170310091900033.HTM
L.

	
                                          15	
  
Figure 1. Beliefs in each news item before and after impeachment

       Notes: The horizontal axis represents the share of people based on their beliefs in each news before the
       impeachment ruling. The vertical axis represents their beliefs in the same news after the impeachment ruling.
       All four news were widely circulated and familiar but did not have clear supporting evidence. The impeachment
       ruling did not resolve the uncertainty in any of the news. Each of the four news is described below.

       News 1. Park Geun Hye asked Samsung’s Jay Lee to support Choi Soon Sil's equestrian and winter sports
       efforts, and in return offered to help him with his succession of Samsung
       News 2. Park Geun Hye was receiving beauty treatment on the day of the Sewol ferry disaster
       News 3. Park Geun Hye directly ordered to create a blacklist of artists
       News 4. Samsung's Jay Lee asked Park Geun Hye to help with his succession and provided a bribe amounting
       to tens of billions of dollars

	
                                                       16	
  
Figure 2. Characteristics of respondents based on primary source of news

Notes: The diamonds indicate the mean and the bands indicate the 95% confidence interval. The ordering of the
news source is based on the number of people that report each medium as their primary source of news (Appendix
Figure S4). TV comprises 45.7%, websites 39.4%, social media 6.6%, family and friends 4.5%, print media 2.7%
and radio 1.1% of the sample.

	
                                                    17	
  
Figure 3. The relationship between protest participation and beliefs by social media use

	
                                             18	
  
Table 1. Social media and rigid beliefs
                                               (1)          (2)            (3)           (4)           (5)
                                              Number of news items on which individuals hold the same beliefs
                                                            pre and post impeachment ruling
                                             0.312***      0.298***       0.213**       0.182***       0.183***
       Social media is primary news source
                                              (0.093)       (0.091)        (0.085)       (0.070)        (0.070)
                                                           -0.410***     -0.383***      -0.154***     -0.156***
       Gets impeachment ruling correct
                                                            (0.067)        (0.062)       (0.054)        (0.054)
                                                                          0.111***      0.083***       0.089***
       Ideological consistency
                                                                           (0.029)       (0.025)        (0.025)
                                                                          0.074**        0.046*         0.043*
       Extreme ideological beliefs
                                                                           (0.030)       (0.025)        (0.025)
                                                                         -0.340***      -0.121***     -0.125***
       Political spectrum                                                  (0.025)       (0.022)        (0.022)
                                                                                                        (0.021)
       Gender control                           Yes           Yes           Yes            Yes           Yes
       Age fixed effects                        Yes           Yes           Yes            Yes           Yes
       Education fixed effects                  Yes           Yes           Yes            Yes           Yes
       Region fixed effects                     Yes           Yes           Yes            Yes           Yes
       Income fixed effects                     Yes           Yes           Yes            Yes           Yes
       Employment status fixed effects          Yes           Yes           Yes            Yes           Yes
       Occupation category fixed effects        Yes           Yes           Yes            Yes           Yes
       Belief in each news pre-impeachment                                                 Yes           Yes
       Belief in primary news source                                                       Yes           Yes
       Risk preference                                                                                   Yes
       Personality traits                                                                                Yes
 R-sq                                         0.085          0.103           0.196         0.477         0.480
Notes: Number of observations is 2,749. Robust standard errors are in parentheses. *** p
Table 2. Social media and protest participation
                                                (1)        (2)             (3)      (4)        (5)       (6)        (7)        (8)

                                                Participated in candlelight protests            Frequency of candlelight protests
                                                        (OLS regressions)                             (Tobit regressions)

                                             0.129**    0.093** 0.113***           -0.104    0.399*** 0.295*** 0.347***       -0.313
       Social media is primary news source
                                             (0.039)    (0.039)       (0.039)     (0.098)    (0.122)   (0.112)    (0.116)    (0.225)
       Social media primary news source *               0.072**                                         0.199*
       Ideological consistency
                                                        (0.032)                                        (0.106)
                                                       0.049***                                        0.146***
       Ideological consistency
                                                        (0.010)                                        (0.028)
       Social media primary news source *                             0.063**                                      0.207*
       Extreme ideological beliefs
                                                                      (0.032)                                     (0.107)
                                                                     0.031***                                     0.104***
       Extreme ideological beliefs
                                                                      (0.011)                                     (0.028)
       Social media primary news source *                                         0.072**                                    0.222***
       Rigid beliefs in news
                                                                                  (0.030)                                    (0.083)
                                                                                  0.030***                                   0.068***
       Rigid beliefs in news
                                                                                  (0.008)                                    (0.020)
       N                                       2746       2746             2746    2746       2749      2749       2749       2749
       R-sq                                   0.151      0.165         0.165       0.164

Notes: In each column one's political spectrum, belief in each news pre-impeachment and primary news source, risk preference,
personality traits, gender, age fixed effects, education fixed effects, region fixed effects, income fixed effects, employment status
fixed effects, and occupation category fixed effects are included as control variables. Because interactions effects are not readily
interpreted in logit regressions, I present OLS regress results in columns 1 to 4. Logit regression results are reported in Appendix
Table S3. Robust standard errors are in parentheses. *** p
Table 3. Social media versus other media sources
                                                  (1)                       (2)                 (3)                      (4)
                                         Number of news items      Participated in        Participated in      Frequency of candlelight
                                         on which individuals    candlelight protests   candlelight protests           protests
                                         hold the same beliefs    (OLS regressions)     (Logit regressions)       (Tobit regressions)

       Social media is primary news            0.228***                   0.116***           0.566***                 0.344***
       source                                   (0.075)                   (0.040)             (0.201)                  (0.123)
       Newspaper is primary news                -0.009                     0.085               0.456                   0.293*
       source                                   (0.122)                   (0.055)             (0.286)                  (0.175)
                                                0.086*                     0.016               0.093                    -0.007
       Web is primary news source
                                                (0.045)                   (0.021)             (0.114)                  (0.051)
                                                0.078                      -0.095             -0.680                    -0.122
       Radio is primary news source
                                                (0.245)                   (0.082)             (0.525)                  (0.251)
       Family/friends are primary news          0.010                     -0.079*             -0.436*                   -0.118
       source                                   (0.099)                   (0.043)             (0.259)                  (0.103)
       N                                         2749                      2749                2746                     2749
 R-sq                                        0.481                   0.167
Notes: In each column one's ideological consistency, extreme ideological beliefs, political spectrum, belief in each news pre-
impeachment and primary news source, risk preference, personality traits, gender, age fixed effects, education fixed effects,
region fixed effects, income fixed effects, employment status fixed effects, and occupation category fixed effects are included as
control variables. Column (1) additionally controls for whether one gets the quiz on impeachment ruling correct. Robust standard
errors are in parentheses. *** p
Appendix

A. Supplemental Information on the Survey and Data

Survey implementation
I used the marketing and survey company Macromill-Embrain, a Korean affiliate of Macromill
Inc., a multinational marketing survey, to conduct both the pre and post impeachment survey.
The pre-impeachment survey was conducted between March 1 and March 7, 2017, which was
shortly before the constitutional court’s ruling on March 10, 2017. The sample was randomly
targeted across four age group categories (20s, 30s, 40s, and 50s) and additionally sampled
college students, since the latter are more likely to use social media. The initial survey sample
was contacted by email to participate in the web-based survey on personal computers. In cases
respondents did not participate, another email was sent out after three to four days. The target
size for the first survey was 3,000 people with 3,114 ultimately participating in the first survey.

I survey the same respondents after the impeachment ruling. The post-impeachment survey was
conducted between April 3 and April 7, 2017. As before email invitations were sent out but
additional encouragements were made through text messages. Participants accrue points by
participating in the firm’s various surveys, and they can later use the accrued points to exchange
for cash or use for online purchases. Compensation is set by the expected time of completion and
translates to about 100 KRW (approximately 10 US cents) per minute. The survey firm’s
expected time for completion was between 6 and 7 minutes. The median time for actual
completion was 7 minutes and 44 seconds. After surveys that were incomplete or had
implausibly short response times were excluded, I ended up with a panel of 2,749 people that
fully responded to both surveys.

Other controls variables in the data and summary statistics
The survey asked the education level of the respondent across 7 categories: less than high school,
graduated high school, attend a 2 year college, graduated a 2 year college, attend a 4 year
college, graduated a 4 year college, and graduate school or above.

Respondent region was selected across the 12 provinces or province level cities. Respondents
selected their own or household (if still a student) income level across 9 bins.

Employment status was defined as employed, self-employed, unemployed looking for work, or
unemployed not looking for work.

There were a total of 13 occupation categories that the respondents could choose from. They
were: student, public service, professional, education, medical, military/police, finance, large
corporation, small or medium enterprise, foreign enterprise, entrepreneur, unemployed, or other.

Individual risk preference was based on the answer to a question that asks where the respondent's
life views lie along a scale from 1 to 10, where 1 is "I tend to avoid risk and choose the most safe
options" and 10 is "I appreciate risk and challenges."

	
                                               22	
  
Personality traits were measured using the Big Five personality traits tests. I control for each of
the five personality traits - extraversion, agreeableness, openness, conscientiousness, neuroticism
- in the regressions.

B. Supplemental Figures and Tables

Figure S1. Rigid beliefs – the number of news that individuals hold the same beliefs before and
after the impeachment ruling.

Notes: The above figure indicates that about 44.6% of the respondents did not change their beliefs on any of the four
news items and they predominantly (43.7% out of 44.6%) believed that the news were all true. The rest 55.4%
changed their views on at least one news item, and 8% changed their beliefs on all news items.

	
                                                      23	
  
Figure S2. Distribution of ideological worldviews

Notes: I ask the extent to one agrees or disagrees with the below statements in a 1 to 10 scale, where 1 corresponds
to fully agreeing to the left statement and 10 fully agreeing to the right statement. For each question the left
represents the progressive's viewpoint and the right the conservative's viewpoint in contemporary South Korea.

       1.   Luck and connection determines success vs. If you work hard you will eventually live a good life
       2.   Income should be more equal vs. Income inequality is necessary to motivate people
       3.   The government should guarantee the people’s living vs. Each individual should support himself/herself
       4.   Small and medium enterprises will drive future growth vs. Conglomerates will drive future growth

	
                                                        24	
  
Figure S3. Ideological consistency

Notes: Ideological consistency is measured by adding the responses across the four questions and then normalizing
so that the midpoint is equal to zero. The above figure presents the histogram, which resembles a normal
distribution, but with a concentration of people at the left. The higher density at the left is a reflection of the young
being more progressive in Korea and the survey oversampling the 20s.

	
                                                        25	
  
Figure S4. Primary source of news

Notes: The above represents the distribution of respondents based on their primary source of news.

	
                                                      26	
  
Figure S5. Characteristics of social media users

Notes: The diamonds indicate the mean and the bands indicate the 95% confidence interval. The ordering of the
news source is based on the number of people that report each medium as their primary source of news

	
                                                     27	
  
Table S1. Summary statistics
                                                               Std.
       Variable                                      Mean                Min     Max     Obs
                                                               Dev.
       Rigid beliefs-Number of news items on which
                                                     2.857     1.302      0       4      2,749
       individuals hold the same beliefs

       Social media is primary news source           0.087     0.282      0       1      2,749

       Gets impeachment ruling correct               0.226     0.418      0       1      2,749

       Ideological consistency                       0.000       1      -1.235   2.848   2,749

       Extreme ideological beliefs                   0.000       1      -0.605   2.898   2,749

       Political spectrum                            0.000       1      -2.248   2.981   2,749

       Belief in news from primary news source       0.000       1      -2.528   2.392   2,749

       Pre-impeachment belief of news 1              0.916     0.277      0       1      2,749

       Pre-impeachment belief of news 2              0.891     0.311      0       1      2,749

       Pre-impeachment belief of news 3              0.781     0.414      0       1      2,749

       Pre-impeachment belief of news 4              0.834     0.372      0       1      2,749

       Risk preference                               0.000       1      -1.513   2.557   2,749

       Personality trait: extraversion               0.000       1      -3.091   3.369   2,749

       Personality trait: agreeableness              0.000       1      -3.710   3.239   2,749

       Personality trait: conscientiousness          0.000       1      -4.121   2.971   2,749

       Personality trait: neuroticism                0.000       1      -3.690   3.315   2,749

       Personality trait: openness                   0.000       1      -3.804   3.615   2,749

       Female                                        0.505     0.500      0       1      2,749

       Age                                           35.088    11.850    20       59     2,749

	
                                                    28	
  
Table S2. Social media and rigid beliefs
                                                                  (1)                                      (2)
                                                 Number of news items on which individuals hold the same beliefs pre and post
                                                 impeachment ruling
                                                               0.182***                                  0.183***
       Social media is primary news source
                                                                (0.070)                                   (0.070)
                                                               -0.154***                                -0.156***
       Gets impeachment ruling correct
                                                                (0.054)                                   (0.054)
                                                               0.083***                                  0.089***
       Ideological consistency
                                                                (0.025)                                   (0.025)
                                                                0.046*                                    0.043*
       Extreme ideological beliefs
                                                                (0.025)                                   (0.025)
                                                               -0.121***                                -0.125***
       Political spectrum
                                                                (0.022)                                   (0.022)
                                                                -0.005                                    -0.006
       Pre-impeachment belief of news 1
                                                                (0.023)                                   (0.022)
                                                                 0.145                                    0.150
       Pre-impeachment belief of news 2
                                                                (0.094)                                   (0.094)
                                                               0.497***                                  0.491***
       Pre-impeachment belief of news 3
                                                                (0.076)                                   (0.076)
                                                               1.060***                                  1.050***
       Pre-impeachment belief of news 4
                                                                (0.062)                                   (0.062)
                                                               0.873***                                  0.872***
       Belief in news from primary news source
                                                                (0.070)                                   (0.070)
                                                                                                          0.010
       Risk preference
                                                                                                          (0.021)
                                                                                                          0.040*
       Personality trait: extraversion
                                                                                                          (0.022)
                                                                                                          -0.024
       Personality trait: agreeableness
                                                                                                          (0.021)
                                                                                                         0.043**
       Personality trait: conscientiousness
                                                                                                          (0.022)
                                                                                                          -0.011
       Personality trait: neuroticism
                                                                                                          (0.023)
                                                                                                        -0.069***
       Personality trait: openness
                                                                                                          (0.021)
       Gender control                                             Yes                                      Yes
       Age fixed effects                                          Yes                                      Yes
       Education fixed effects                                    Yes                                      Yes
       Region fixed effects                                       Yes                                      Yes
       Income fixed effects                                       Yes                                      Yes
       Employment status fixed effects                            Yes                                      Yes
       Occupation category fixed effects                          Yes                                      Yes
 R-sq                                                       0.477                                      0.480
Notes: Number of observations is 2,749. Robust standard errors are in parentheses. *** p
Table S3. Social media and protest participation (logit regressions)
                                                            (1)                  (2)                 (3)                (4)

                                                                           Participated in candlelight protests

                                                         0.621**              0.447**             0.540***            -0.664
       Social media is primary news source
                                                          (0.187)              (0.201)             (0.193)           (0.764)
       Social media primary news source *                                      0.358*
       Ideological consistency
                                                                               (0.186)
                                                                             0.245***
       Ideological consistency
                                                                               (0.055)
       Social media primary news source *                                                          0.317*
       Extreme ideological beliefs
                                                                                                   (0.166)
                                                                                                  0.142**
       Extreme ideological beliefs
                                                                                                   (0.057)
       Social media primary news source *                                                                             0.379*
       Rigid beliefs in news
                                                                                                                     (0.220)
                                                                                                                    0.169***
       Rigid beliefs in news
                                                                                                                     (0.054)

       N                                                   2746                 2746                2746               2746

Notes: In each column one's political spectrum, belief in each news pre-impeachment and primary source of news, risk and
personality trait, gender, age fixed effects, education fixed effects, region fixed effects, income fixed effects, employment status
fixed effects, and occupation category fixed effects are included as control variables. Robust standard errors are in parentheses.
*** p
Table S4. Social media and the counter (Taegeukgi) protests

                                               (1)       (2)              (3)     (4)        (5)        (6)       (7)       (8)

                                             Participated in the counter (Taegeukki)        Frequency of the counter (Taegeukki)
                                                    protests - OLS regressions                   protests - Tobit regressions

                                             0.053**   0.043**     0.046**       0.102     0.101***   0.070     0.086*     0.139
       Social media is primary news source
                                             (0.023)   (0.022)      (0.022)     (0.083)    (0.056)    (0.044)   (0.049)   (0.133)
       Social media primary news source *               0.030                                         0.099
       Ideological consistency                         (0.021)                                        (0.075)
                                                        0.003                                         0.006
       Ideological consistency
                                                       (0.003)                                        (0.010)
       Social media primary news source *                            0.032                                      0.087
       Extreme ideological beliefs                                  (0.026)                                     (0.079)
                                                                    0.007*                                      0.009
       Extreme ideological beliefs
                                                                    (0.004)                                     (0.008)
       Social media primary news source *                                        -0.017                                    -0.016
       Rigid beliefs in news                                                    (0.024)                                   (0.043)
                                                                                0.016***                                  0.055***
       Rigid beliefs in news
                                                                                (0.005)                                   (0.016)
       N                                      2746      2749         2749        2749       2749       2749      2749      2749
 R-sq                                       0.085     0.088      0.091      0.094
Notes: In each column one's political spectrum, belief in each news pre-impeachment and primary source of news, risk
preference, personality traits, gender, age fixed effects, education fixed effects, region fixed effects, income fixed effects,
employment status fixed effects, and occupation category fixed effects are included as control variables. Robust standard errors
are in parentheses. *** p
You can also read