Professional Documents
Culture Documents
The Effects of Television News Propaganda: Experimental Evidence From China
The Effects of Television News Propaganda: Experimental Evidence From China
The Effects of Television News Propaganda: Experimental Evidence From China
Abstract
More than half of the world’s population live under authoritarian rule, where pro-
paganda is deployed to manipulate attitudes and behaviors at scale. Television re-
mains one of the most powerful forms of propaganda because audiences continue to
regard television as an authoritative source of information. We create videos that re-
alistically resemble Chinese state news propaganda and experimentally assess their
effects on the policy preferences of the Chinese public. We find that propaganda
moves respondents to adopt policy positions espoused in the video up to 48 hours
after exposure and appears to work through persuasion. In contrast to prior research,
effects do not vary based on individual predispositions or characteristics, which may
be due to the content and format of propaganda.
∗
Our thanks to Andrew Chadwick, Lance Holbert, Rasmus Nielsen, and seminar participants at Washington
University in St. Louis for many helpful comments and suggestions.
†
Assistant Professor, Department of Communication, Stanford University. Email: jp1@stanford.edu.
‡
Associate Research Fellow and Postdoctoral Researcher, School of Government, Sun Yat-sen University.
Email: shaozijie@vip.sina.com.
§
Assistant Professor, Department of Political Science, Stanford University. Email: yiqingxu@stanford.edu.
More than half of the world’s population live under authoritarian rule,1 where propaganda
is deployed to manipulate attitudes and behaviors at scale. Television—in traditional and
online streaming forms—remains one of the most powerful forms of propaganda available
to autocrats despite the increasing complexity of media ecosystems and fragmentation of
audiences (Chadwick 2017; Webster and Ksiazek 2012). Audiences continue to regard
television as an authoritative source of information. In China, a nationally representative
survey from 2015 showed the primary way Chinese people consume news is through tele-
vision (CUGS 2015). In Russia, one of the first things Putin did after assuming the presi-
dency in 2000 was to impose state control on all major television channels by co-opting,
imprisoning, or exiling owners of independent stations (Enikolopov et al. 2011). The Rus-
sian regime uses television to influence the political sphere (Nisbet et al. 2017; Oates and
McCormack 2010; Stoycheff and Nisbet 2017), and as of 2018, most Russians say that
television, which reaches more than 70% of the urban population, is their most important
source of information (Zakem et al. 2018). In Azerbajian, AzTV, the state-controlled na-
tional television channel, disseminates pro-government propaganda reaching more than
99.9% of the population.2 In Iran, more than 80% of the population considers the state-
controlled Islamic Republic of Iran Broadcasting to be the most authoritative source of
information, and Iran recently redoubled investment to upgrade television content and
technology.3
Given its continued prominence, does television propaganda work? If yes, how does
it work and how well does it work? In this paper, we examine these questions by experi-
mentally assessing the effects of Chinese television propaganda. We create short segments
of news videos in collaboration with industry experts that realistically resemble Chinese
state propaganda. We embed these videos in an online survey in China to examine effects
on policy preferences. Our results show that exposure to television propaganda moves
1
Regime classification based on Geddes et al. (2014) and population data from the World Bank
(https://data.worldbank.org/indicator/SP.POP.TOTL).
2
AzTV penetration rate from Azerbaijan’s Ministry of Communications and Information Technologies; see
https://bit.ly/2VNqB51 (Accessed April 16, 2019).
3
See https://bit.ly/2ICcjk3 (Accessed December 8, 2019).
Theories of Propaganda
Research on media effects and propaganda has generally found that consuming propa-
ganda leads to changes in attitudes, opinions, and behaviors (Adena et al. 2015; Bleck and
Michelitch 2017; Geddes and Zaller 1989; Gerber et al. 2009; Iyengar and Hahn 2009;
Iyengar et al. 1982; King et al. 2017; Lasswell 1927; Nisbet and Myers 2011; Stockmann
and Gallagher 2011; Szostek 2017a; Yanagizawa-Drott 2014). There is a greater diversity
of views, however, about how propaganda changes attitudes and what attitudes it changes.
Propaganda has been found to work through persuasion, where the provision of informa-
tion about an issue changes opinions and attitudes toward that issue (Chaffee 1977; Lass-
well 1927; Jowett and O’Donnell 2018). Propaganda has also been shown to work through
learning (Prior and Lupia 2008), where information provided by propagandists is novel
so that the public learns and forms new opinions as a result. Instead of changing attitudes
about the particular topic contained in propaganda, researchers have also shown that pro-
paganda can generate more generalized feelings of fear and guilt (Huang 2015; Wedeen
H1: People exposed to the regime’s position on a domestic policy issue will have policy
preferences closer to the regime’s positions on that issue than respondents in the
control condition.
H2: People exposed to the regime’s position on a foreign policy issue will have policy
preferences closer to the regime’s positions on that issue than respondents in the
control condition.
H3: People exposed to the regime’s position on a domestic policy issue will not have
policy preferences closer to the regime’s positions on related issues than respon-
dents in the control condition.
4
All hypotheses and analysis presented in this paper are pre-registered.
We are also interested in the questions of how television propaganda works and how
well it works. To examine these questions, we conduct exploratory analysis of the effects
of propaganda given different pre-treatment measures of pre-disposition as well as for
different socioeconomic and demographic sub-groups.
The explicitly stated goal of propaganda in China is to shape public opinion in favor
of the Chinese Communist Party (CCP). The latest government document pertaining to
propaganda—the 2019 “Regulations of the Communist Party of China on Propaganda
Work”—describes propaganda as foundational to the strength of the CCP, and an impor-
tant way to shape the opinions of party leaders and party members as well as those of the
masses.5 The Propaganda Department of the CCP is the organization tasked with pro-
ducing and disseminating information to shape public opinions and behaviors. The CCP
Propaganda Department oversees a vast system of control that encompasses media, educa-
tion, culture, sports, and even non-governmental organizations and research (Cantoni et al.
2017; Clayton et al. 2006; King et al. 2013). Television broadcasting has been part of this
system since the late 1970s (Brady 2009; Lull 2013; Zhao 1998). The National Radio and
Television Administration, which operates the country’s predominant public broadcaster,
China Central Television (CCTV), reports directly to the Propaganda Department.
Television news is used by the Communist Party to shape public preferences on policy
issues. Shaping public opinion on policy issues is narrower than the broader goal of pro-
paganda in garnering public support and suppressing opposition. However, shaping public
attitudes on policy issues plays an important role in autocracies. Public opinion is gener-
ally considered less consequential in authoritarian regimes than in democracies because
the public cannot vote or easily sanction politicians for their policy decisions. However,
5
From “Regulations of the Communist Party of China on Propaganda Work” promulgated on August 2019
(中国共产党宣传工作条例). For details, see https://bit.ly/2R4ebVx (Accessed January 6, 2020).
Methods
We worked closely with media professionals in China to create six short news video seg-
ments that closely resemble the content and style of Xinwen Lianbo. We embedded these
news segments in an online survey experiment to measure the effects of television propa-
ganda on policy preferences. We first describe the video segments, and then describe the
survey in which the videos were embedded.
Treatment conditions: The six news segments were all one to two minutes long. Three
of the segments were focused on the domestic policy issue of state-owned enterprise
(SOE) reform. Three of the video segments are focused on the foreign policy (FP) issue of
disputes in the South China Sea between China and the Philippines. We wanted to include
a domestic and a foreign policy issue, and we chose SOE reform and China-Philippines
relations because they are areas where the Chinese government has expressed different
policy positions in recent times. This diversity in the positions the government has taken
allowed us to assess the effect of television propaganda rather than the particularities of
8
These channels are (listed in order of Xinwen Lianbo viewership): CCTV-6, CCTV-3, CCTV-8, CCTV-
News, Hunan TV Cartoons, CCTV-14 (Kids), Beijing TV Kids, CCTV-4, CCTV-Music, CCTV-10, Hunan
TV, CCTV-12, CCTV-5, Beijing TV, CCTV-7, CCTV-2, Anhui TV, Jiangsu TV, Jiangxi TV, Shanghai
Oriental TV, Guizhou TV, Shandong TV, Guangdong TV Kids, Hubei TV, CCTV-9, Hubei TV, Shanghai
Toonmax, Guangdong TV, Sichuang TV, Shanxi TV, Jiangsu Youman Cartoon, Heilongjiang TV, Zhejiang
TV, Chongqing TV, Southeast TV, Henan TV, Liaoning TV, Tianjin TV, Qinghai TV, CCTV-10, Guangxi TV,
Shenzhen TV, Xinjiang TV, Liaoning TV, Inner Mongolia TV, Bingtuan TV, Yunnan TV, Jilin TV, Shandong
Education TV, Shaanxi TV, Xiamen TV, China Education, Amdo TV, Tibet TV, Travel TV, Guandong
Southern, China Education 2.
9
For comparison, the highest rated show broadcast on a single channel on weekday evenings in March
2019 is Wolf Warrior II, with 1.83% of households with TV access and 7.17% of TV viewers tuned in
during weekday evenings.
Survey design: The overall flow of our survey is illustrated in Figure 1. After screening,
we first asked respondents to answer two sets of questions to measure their predisposition
11
We plan to anonymize these videos if the paper is accepted to protect the identities of the interviewees.
12
Rich provinces include Beijing, Tianjin, Inner Mongolia, Liaoning, Shanghai, Jiangsu, Zhejiang, Anhui,
Fujian, Shandong, Hubie, Guangdong, Hainan, and Chongqin.
10
Start Start
Pre-disposition
Randomized order
Randomized order
Market State Control Dove Hawk Control SOE Reform
Preferences
Manipulation Check Manipulation Check
SOE Reform Preferences Foreign Policy Preferences Foreign Policy
Preferences
Demographics
Other Outcomes
End 48 hours later End
on economic and foreign policies. One set contained seven questions related to pref-
erences on market economy—for example, whether private ownership and sale of land
should be allowed—where responses could range from preferences for free market and
private capital to preferences for state control of the economy and limits on the influence
of private capital (see Appendix Table A1 for question details). The other set contained
seven questions related to preference on national sovereignty, such as whether it is ap-
propriate to use military force to unify Taiwan with China. Here, responses ranged from
highly nationalistic to not nationalistic. Based on these two sets of questions, we con-
structed two simple additive indices as measures of respondent predisposition in the two
dimensions.13
After we measured predisposition, respondents watched two videos, one on each pol-
icy issue. The order in which the videos were watched was randomized, and randomiza-
tion for each video was independent. In other words, the assignments of the two treat-
ments were independent of each other, such that if a respondent was assigned the video
espousing market orientation for SOE reform, the respondent could have been assigned to
any of the three treatment conditions for the China-Philippines relations video with equal
probability. The order of the two policy areas was also randomized. Respondents were
13
Questions were drawn from a recent studyf that shows that policy preferences of the Chinese public mea-
sured this way have high inter-temporal stability (Pan and Xu 2020).
11
12
13
Research ethics: The primary ethical concern in this paper, as with any experimental
study of media effects, was that the experiments would alter policy preferences in politi-
cally meaningful ways. Our substantive focus helped us address this concern because we
were interested in focusing on specific policy issues rather than on propaganda that aimed
to promote support for the regime, to strengthen state legitimacy, or to promote a cult of
personality. In addition, because we chose policy issues where the position of the state
had changed and because our experimental treatments contained these divergent policy
positions, our experiment does not promote any hegemonic political position.
A secondary ethical concern was that the respondents who participated in the survey
would experience retaliation if their responses deviated from the preferred policy positions
of the Chinese regime. We guarded against this threat by not collecting any personally
identifiable information, and by securely storing the data. We obtained approval from our
university IRBs for all aspects of the research.
Results
In line with H1 and H2, we find that exposure to video news content moves respondents
toward the espoused policy position of the state, regardless of what that position is and
regardless of the prior disposition of the respondent. Many of these effects persist up to 48
hours after exposure. Contrary to H3 and H4, video exposure also moves the expressed
policy positions of respondents on related issues. We begin by showing the manipulation
check before moving to the main results, followed by analysis of heterogeneous effects
17
Speeders are defined as those who spent less than 10 minutes completing the survey. The median finish
time was 17 minutes.
18
Not surprisingly, the estimated effect became stronger when we focused on the sub-sample of respondents
who were more attentive when answering the survey.
14
Manipulation checks: The video treatments have high construct validity. Most respon-
dents, 86%, thought our video segments came from CCTV. The manipulation checks show
that respondents were able to correctly answer actual questions about the videos. Columns
(1) and (4) of Table 1 show that most of the respondents (83.4% for SOE reform, 97.5%
for China-Philippines relations) correctly answered the factual questions after watching
the two videos.19
In terms of comprehending the policy position espoused in the video, the vast majority
of respondents were able to correctly identify the policy position of most of the videos.
Columns (2) and (3) of Table 1 show that 75.5% of respondents who were assigned to
the pro-market condition identified the SOE media content as pro-market (as opposed to
24.5% under the neutral condition), and 61.2% of those who were assigned to the pro-
19
Because treatment was assigned at the individual level, we use robust standard errors throughout this paper.
Clustering standard errors at the provincial level gives almost identical substantive results.
15
Main effects: Respondents adopted policy positions closer to the policy position es-
poused in the propaganda video, regardless of what that position was. Figure 2 presents
the estimated treatment effects on preferences toward SOE reform. The outcome in the
Pro−market Pro−market
0.6
0.6
0.4
0.4
● ●
0.2
0.2
Coefficients
Coefficients
0.0
0.0
● ● ●
●
−0.2
−0.2
●
−0.4
−0.4
−0.6
−0.6
Pro−state Pro−state
T1: Market T1: State T2: Dove T2: Hawk T1: Market T1: State T2: Dove T2: Hawk
left panel is a three-value indicator of direct policy preference (a bigger value means a
more pro-market approach), and the outcome in the right panel is a standardized index
of market-oriented reform (with 0 mean and 1 standard deviation) based on nine related
policy questions. The thick and thin bars in both plots represent 95% and 99% confidence
intervals for the point estimates, respectively (same for subsequent Figures 3 and 6). We
estimated the effects using regressions, controlling for a full set of individual character-
istics and provincial fixed effects (see Table A6 in the Appendix for the full regression
table).20 In line with H1, we see that the pro-market video content on average shifted the
20
The control variables include the respondent’s gender, age, age squared, level of education, ethnic minority,
religiosity, marital status, working experience, Communist Party membership, experience using a virtual
16
Dovish Dovish
0.6
0.6
0.4
0.4
0.2
0.2
●
●
Coefficients
Coefficients
●
0.0
0.0
● ● ●
−0.2
−0.2
●
●
−0.4
−0.4
−0.6
−0.6
Hawkish Hawkish
T1: Market T1: State T2: Dove T2: Hawk T1: Market T1: State T2: Dove T2: Hawk
ences (a bigger value means a more dovish approach), and the right panel shows a stan-
dardized index of dovish foreign policy (with 0 mean and 1 standard deviation) based on
nine related policy questions. The full regression results are presented in Table A7 in the
Appendix. From Figure 3, we see that the dovish message on China-Philippines relations
private network, English proficiency, income level, and self-reported socioeconomic status.
17
Treatment effects and predisposition: Predisposition does not appear to moderate the
effects of television news propaganda. We measured predisposition toward economic
policies and national sovereignty before exposing respondents to the video treatments.
We find strong correlation between predisposition and post-treatment policy preferences
in both the economic and the sovereignty dimensions. This suggests, first, that many re-
spondents have relatively coherent policy preferences. Second, this suggests respondents
are not simply answering questions to satisfy the researchers. Strikingly, regardless of
predisposition, exposure to the video segments changes policy preferences.
Figure 4 shows the predicted outcomes of three sub-groups: the group that viewed
the pro-market video (left panel), the control group (central panel), and the group that
viewed the pro-state video (right panel). In each group, we see a strong, positive corre-
lation between predisposition for a free market economy (x-axis values) and support for
a more market-oriented approach to SOE reform (y-axis values). However, regardless of
predisposition, we see that the pro-market video increases support for a market-oriented
approach to SOE reform. We can see this by comparing the left and central panels of
Figure 4, where average support for market reform is higher at all values of predisposition
in the market treatment (left panel) than in the control group (central panel). The pro-state
video decreases support for a market-oriented approach to SOE reform across the board.
We observe this by comparing the central and right panels of Figure 4, where average
support for market reform is lower at all values of predisposition in the state treatment
18
Note: The histogram at the bottom of each plot shows the number of respondents under each treatment
condition at different values of predisposition (in percentile).
group (right panel) than in the control group (central panel).21 There are no interactive
effects between predisposition and exposure to video propaganda on economic policy.
Figure 5 shows predicted outcome of the group that viewed the dovish video (left
panel), the control group (central panel), and the group that viewed the hawkish video
(right panel) on China-Philippine relations. Again, we see a positive correlation between
predisposition for more nationalistic/hawkish views toward sovereignty (x-axis values)
and support for hawkish foreign policy in all three groups (y-axis values). Again, regard-
less of predisposition, the hawkish video increases support for hawkish foreign policy
(comparing the right and central panels of Figure 5) and the dovish video increases sup-
port for dovish foreign policy across the board (comparing the left and central panels of
Figure 5). This means that in the foreign policy domain, we also do not observe interactive
effects between predisposition and exposure to video propaganda.
Treatment effects and demographics: The effects of propaganda do not vary in sta-
tistically significant ways for individuals with different socioeconomic and demographic
characteristics. Using the same regression specification controlling for individual char-
acteristics and provincial fixed effects, we conducted subgroup analyses and explored
21
The differences between curves in the left and central panels and between the central and right panels
of Figure 4 can be understood as the marginal effects of the treatments on the outcome conditioning on
predisposition. See Appendix Figure A2 for the marginal effects plot.
19
Note: The histogram at the bottom of each plot shows the number of respondents under each treatment
condition at different values of predisposition (in percentile).
20
Female ● Female ●
Male ● Male ●
−1.0 −0.5 0.0 0.5 1.0 −1.0 −0.5 0.0 0.5 1.0
Coefficients Coefficients
(a) SOE Treatments (T1) on Market Index (b) FP Treatments (T2) on Dove Index
Robustness checks: We conducted various robustness checks by pruning the data based
on different criteria. Table A8 presents the results. The outcome variables in Panel A and
B are the two indices of policy preferences. Column (1) shows the benchmark result
using the full sample. In columns (2) through (6), we limit the sample to respondents who
satisfied various conditions, including those who identified the source of the videos as
CCTV (86% of all respondents), those who passed the age consistency filter and were not
speeders (92% of all respondents), those who answered both factual questions correctly
(83% of all respondents), those who agreed the policy toward SOE reform is correct (94%
of all respondents), and those who agreed the policy toward the South China Sea is correct
(96% of all respondents). The results remained substantively the same, and in some cases,
the effects became even stronger.
21
23
The other respondents were part of our soft launch.
22
2.0
●
1.8
●
● ●
2.5
●
●
1.6
●
●
2.0
1.4
●
●
●
1.5
1.2
Control Control
Market Dove
State Hawk
1.0
1.0
Wave 1 Wave 2 Wave 1 Wave 2
Note: The hollow dots represent outcome variable means in respective treatment groups for the full sample
(wave 1 only); outcome variable means in respective treatment groups for the subset of respondents who
participated in both waves are represented by solid dots (wave 1) and arrows (wave 2), respectively.
of Figure 7 shows that the wave 1 views of respondents who also completed the 48-hour
follow-up (solid circle) persist in the 48-hour follow-up (solid triangle). Respondents
exposed to the pro-market SOE reform video (dashed blue line) remain strongly support-
ive of market-oriented reform of SOEs, while those exposed to the pro-state SOE reform
video (dot-dashed red line) remain less supportive of market-oriented SOE reform. The
right panel of Figure 7 shows similar trends of persisting views for the China-Philippines
relations videos, with those exposed to the dovish video remaining more supportive of co-
operation and those exposed to the hawkish video less supportive of cooperation. These
results are generally similar when we use the index of preferences for market-oriented
reform and South China Sea (see Table A10 for full regression tables).
Discussion
In this paper, we conducted an experiment to study the effect of television news propa-
ganda on the stated policy preferences of ordinary Chinese people. We created six short
23
24
24
Preference falsification may also make mobilization and revolution more unpredictable, but in the absence
of revolutionary bandwagons or information cascades (Kuran 1991; Lohmann 2000), mobilization against
the regime is in general more difficult.
25
26
27
28
29
30
Table of Contents
3. Predisposition questions
5. Summary statistics
6. Balance table
7. Main results
A-1
A-2
SOE: Market In 2014 Ministry of Finance data showed that there were 15 million
enterprises nationwide, and state-owned accounted for less than 1% of enterprises but
assets controlled by state-owned enterprises accounted for at least 30% of total assets. As
a crucial component to economic reform and development, the evolution of state-owned
enterprises plays a decisive role in the development of China’s national economy.
[SOE Expert Interview] The biggest problem for most SOEs is that they bear a lot
of policy burdens because they need to bear a lot of responsibilities and you don’t know
how much policy burden they bear. It is uncertain how much subsidies the government
should give them. Their optimal choice may be to ask the government for more subsidies,
but then they have no incentive to improve their overall performance and efficiency. If
SOE managers and employees do not perform well, you will find that SOEs performance
will be worse than that of private enterprises. Government departments can measure the
performance of SOEs using relative measures, and use this as a basis to provide incentives
to SOE managers.
The key to the SOE reform is to ensure it operates in a market-oriented manner, to
increase its competitiveness in the market such that the market can play a decisive role in
resource allocation. We must break barriers in institutional obstacles, improve the regu-
latory system, follow market forces, reduce burdens, strengthen health, improve quality,
and enhance efficiency in order to truly make SOEs competitive and independent market
players.
SOE: State In 2014 Ministry of Finance data showed that there were 15 million enter-
prises nationwide, and state-owned accounted for less than 1% of enterprises but assets
controlled by state-owned enterprises accounted for at least 30% of total assets. As a
crucial component to economic reform and development, the evolution of state-owned
enterprises plays a decisive role in the development of China’s national economy.
[SOE Expert Interview] Now SOEs are responsible for some of the most important
tasks of the national economy. So they are indispensable to the entire national economy,
A-3
SOE: Control In 2014 Ministry of Finance data showed that there were 15 million
enterprises nationwide, and state-owned accounted for less than 1% of enterprises but
assets controlled by state-owned enterprises accounted for at least 30% of total assets. As
a crucial component to economic reform and development, the evolution of state-owned
enterprises plays a decisive role in the development of China’s national economy.
[SOE Expert Interview] At present, there are actually a lot of views on the reform of
SOEs. SOEs are responsible for some of the most important tasks of the national econ-
omy. So they are indispensable to the entire national economy. SOEs are enterprises
invested or controlled by the government, including not only the central government or
federal government, but also the local governments. As an organizational form of produc-
tion and management, SOEs have the characteristics of commercial and public welfare,
and their commercial performance is the pursuit of preserving and increasing the value of
state-owned assets. The establishment of SOEs is usually to achieve the goal of national
economic adjustment, and to play a role in reconciling all aspects of the national economy.
SOEs have a certain administrative nature. Due to historical reasons, the classification of
Chinese SOEs is quite complex. According to international practice, enterprises investing
in state-owned assets or holding more than 50% of their shares are SOEs. Therefore, en-
A-4
A-5
A-6
A-7
A-8
Panel A of Table A2 shows example questions on topics related to SOE reform, along
with their corresponding signs (1 or -1) used to construct the index. Panel B of Table A2
shows example questions on topics related to foreign policy.
A-9
A-10
Treatment1: Market 0.018 0.491 0.011 -0.020 0.013 0.014 0.010 -0.039 -0.016 -0.009 0.028 0.020 0.058 -0.077 -0.056 -0.045
(0.042) (0.733) (0.037) (0.028) (0.042) (0.016) (0.031) (0.042) (0.035) (0.031) (0.042) (0.029) (0.041) (0.097) (0.172) (0.175)
Treatment1: State -0.048 -0.201 -0.024 0.028 0.008 0.004 0.022 -0.001 -0.030 0.014 -0.035 -0.000 0.018 0.090 -0.060 0.238
(0.042) (0.745) (0.036) (0.030) (0.042) (0.014) (0.032) (0.043) (0.036) (0.032) (0.041) (0.028) (0.041) (0.097) (0.172) (0.174)
0.079* 0.462 0.017 0.008 0.052 0.004 -0.030 -0.066 -0.015 -0.024 0.021 0.014 0.018 0.137 -0.122 0.151
Treatment2: Dove (0.042) (0.758) (0.036) (0.028) (0.042) (0.015) (0.033) (0.042) (0.036) (0.031) (0.041) (0.028) (0.041) (0.097) (0.176) (0.175)
0.012 -1.244* 0.027 0.024 0.033 0.015 -0.060* 0.029 0.023 -0.020 0.016 0.034 -0.002 0.169* -0.322* 0.258
A-11
Treatment2: Hawk (0.041) (0.730) (0.036) (0.029) (0.042) (0.016) (0.032) (0.043) (0.035) (0.032) (0.041) (0.029) (0.041) (0.097) (0.170) (0.173)
Constant 0.386*** 31.311*** 0.230*** 0.120*** 0.391*** 0.023** 0.188*** 0.479*** 0.785*** 0.172*** 0.371*** 0.111*** 0.345*** 2.802*** 4.780*** 5.021***
(0.038) (0.693) (0.033) (0.026) (0.038) (0.010) (0.030) (0.039) (0.032) (0.029) (0.038) (0.026) (0.037) (0.090) (0.154) (0.164)
Observations 830 830 830 830 830 830 830 830 830 830 830 830 830 830 830 830
R-squared 0.008 0.008 0.002 0.004 0.002 0.002 0.005 0.007 0.002 0.002 0.003 0.003 0.003 0.007 0.004 0.006
F test p -value 0.147 0.142 0.790 0.503 0.796 0.666 0.377 0.173 0.735 0.845 0.612 0.711 0.681 0.207 0.433 0.298
Public Being
High Junior College Having sector reported CCP Having English Income
Female Age school college or above Minority Religious Single worked worker to member used VPN proficiency category
Outcome Variables (0 or 1) [18, 50] (0 or 1) (0 or 1) (0 or 1) (0 or 1) (0 or 1) (0 or 1) (0 or 1) (0 or 1) (0 or 1) (0 or 1) (0 or 1) [1, 5] [1, 10[
Electronic copy available at: https://ssrn.com/abstract=3579148
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15)
Treatment1: Market 0.050 -0.681 0.070 -0.014 -0.052 -0.001 0.058 -0.009 -0.019 -0.059 -0.024 -0.005 0.021 -0.176 -0.196
(0.064) (1.135) (0.053) (0.041) (0.064) (0.026) (0.049) (0.063) (0.048) (0.051) (0.065) (0.049) (0.063) (0.143) (0.244)
Treatment1: State 0.006 -1.538 -0.030 0.035 0.015 -0.014 0.047 0.040 -0.048 -0.097* -0.092 -0.055 0.041 0.073 -0.019
(0.064) (1.115) (0.048) (0.045) (0.065) (0.022) (0.051) (0.065) (0.051) (0.051) (0.066) (0.047) (0.065) (0.145) (0.253)
Treatment2: Dovish 0.110* 0.826 -0.024 0.013 0.048 -0.010 -0.032 -0.178*** 0.030 -0.008 0.067 -0.031 -0.054 0.075 -0.148
(0.064) (1.141) (0.048) (0.045) (0.064) (0.022) (0.053) (0.062) (0.050) (0.051) (0.065) (0.046) (0.063) (0.143) (0.257)
Treatment2: Hawkish -0.022 -2.048* 0.054 -0.035 0.040 0.017 -0.103** 0.003 0.024 -0.072 0.058 -0.009 -0.008 0.078 -0.291
(0.064) (1.139) (0.054) (0.042) (0.065) (0.025) (0.050) (0.066) (0.051) (0.049) (0.066) (0.049) (0.065) (0.145) (0.251)
A-12
Constant 0.363*** 32.921*** 0.166*** 0.128*** 0.571*** 0.039** 0.193*** 0.446*** 0.829*** 0.250*** 0.465*** 0.183*** 0.402*** 3.116*** 5.165***
(0.058) (1.085) (0.047) (0.040) (0.060) (0.018) (0.047) (0.060) (0.045) (0.051) (0.061) (0.048) (0.059) (0.141) (0.226)
Observations 354 354 354 354 354 354 354 354 354 354 354 354 354 354 354
R-squared 0.016 0.023 0.019 0.008 0.005 0.005 0.017 0.032 0.004 0.016 0.010 0.006 0.004 0.010 0.006
F test p -value 0.232 0.0697 0.185 0.594 0.772 0.721 0.150 0.0164 0.853 0.266 0.468 0.733 0.874 0.442 0.709
Outcome Variable Mean: ∼Full sample 2.165 2.165 2.165 0.000 0.000 0.000
∼Treatment1 = Neutral 2.167 2.167 2.167 -0.035 -0.035 -0.035
∼Treatment2 = Neutral 2.256 2.256 2.256 0.090 0.090 0.090
Control variables No Yes Yes No Yes Yes
Provincial fixed effects No No Yes No No Yes
Observations 830 830 830 823 823 823
R-squared 0.088 0.126 0.160 0.044 0.100 0.125
Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
A-13
Outcome Variable Mean: ∼Full sample 1.578 1.575 1.575 0.000 0.000 0.000
∼Treatment1 = Neutral 1.573 1.573 1.573 -0.013 0.024 0.024
∼Treatment2 = Neutral 1.630 1.630 1.630 -0.013 0.024 0.024
Control variables No Yes Yes No Yes Yes
Provincial fixed effects No No Yes No No Yes
Observations 830 830 830 825 825 825
R-squared 0.074 0.130 0.156 0.035 0.069 0.094
Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
A-14
A-15
Treatment1: Market -0.156* 0.016 0.033 -0.169 -0.007 -0.006 0.013 0.047
(0.087) (0.085) (0.071) (0.166) (0.178) (0.034) (0.031) (0.050)
Treatment1: State 0.044 0.101 -0.050 0.084 0.069 -0.015 0.001 0.031
(0.084) (0.082) (0.070) (0.159) (0.188) (0.034) (0.030) (0.051)
Treatment2: Dove 0.018 0.042 0.129* 0.033 -0.036 -0.025 -0.034 0.076
(0.085) (0.080) (0.070) (0.159) (0.187) (0.034) (0.031) (0.053)
Treatment2: Hawk 0.056 0.134 0.062 0.080 0.119 -0.005 -0.040 0.087*
(0.083) (0.083) (0.070) (0.174) (0.186) (0.033) (0.031) (0.049)
Outcome Variable Mean 0.000 0.000 0.000 8.145 6.731 0.207 0.169 2.218
Control variables Yes Yes Yes Yes Yes Yes Yes Yes
Provincial fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 828 823 822 830 830 830 830 830
R-squared 0.118 0.188 0.404 0.077 0.113 0.147 0.159 0.261
Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. A full set of individual characteristics and provincial fixed
effects are being controlled for in all regressions.
A-16
0.5
−0.5
−1.0
0 25 50 75 100
State Control <− Predisposition on Economy −> Free Market
0.5
Marginal Effects of Treatments
0.0
Treatment
T2: Control
T2: Dove
T2: Hawk
−0.5
0 25 50 75 100
Hawkish <− Predisposition on Sovereignty −> Dovish
Note: The (stacked) histograms at the bottom of each plot show the number of respondents under each
treatment condition given different values of the moderator (in percentile).
A-17
Table A10 presents the estimated treatment effects on the full and recontact samples.
Outcome Variable Mean: ∼Full Sample 2.165 2.274 2.218 0.000 0.040 0.128
∼Treatment1 = Neutral 2.168 2.314 2.339 -0.035 0.023 0.254
∼Treatment2 = Neutral 2.256 2.410 2.341 0.090 0.093 0.199
Control variables Yes Yes Yes Yes Yes Yes
Provincial fixed effects Yes Yes Yes Yes Yes Yes
Observations 830 353 353 823 351 350
R-squared 0.160 0.207 0.201 0.125 0.245 0.194
Panel B Foreign Poicy
Agree to policies that emphasize more on An index of dovish foreign policy
cooperation than sovereignty (1, 2 or 3) (W1: mean = 0; sd = 1)
Two-wave Two-wave Two-wave Two-wave
Full Sample: Subsample: Subsample: Full Sample: Subsample: Subsample:
w1 outcome w1 outcome w2 outcome w1 outcome w1 outcome w2 outcome
(1) (2) (3) (4) (5) (6)
Outcome Variable Mean: ∼Full Sample 1.578 1.541 1.533 0.000 -0.047 0.000
∼Treatment1 = Neutral 1.573 1.568 1.474 -0.013 -0.066 0.001
∼Treatment2 = Neutral 1.630 1.521 1.512 0.024 -0.136 -0.004
Control variables Yes Yes Yes Yes Yes Yes
Provincial fixed effects Yes Yes Yes Yes Yes Yes
Observations 830 353 353 825 352 352
R-squared 0.156 0.236 0.151 0.094 0.137 0.093
Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. A full set of individual characteristics and
provincial fixed effects are being controlled for in all regressions.
A-18
A-19
Income category [1, 10] What is your total disposable income, including salaries, stipends, and allowances? 10 categories, from low to high.
Religious Do you have a religion? 1 = yes; 0 = no.
CCP member What is your political affiliation? 1 = Chinese Communist Party; 0 = otherwise.
Self-reported social class [0, 10] What do you think is your social class on a 0-10 scale? 0 = the bottom of the society; 1 = the top of the society.
Having worked Have you ever had a full-time job? 1 = yes; 0 = no.
Public sector worker What is the nature of your employer? 1 = SOEs or the government; 0 = otherwise.
A-20
Being reported to In your job, are there any people who report to you? 1 = yes; 0 = no.
Pre-disposition on market economy: Private capital should be encouraged to set up private hospitals to provide convenient
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 2 and high-quality services to those willing to pay high prices.
Pre-disposition on market economy: When the market price of grain is low, the government should purchase grain from
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 4 farmers at a price higher than the market price.
1 = The key to the SOE reform is top-level design; state-led SOE reform
can be successful. 2 = The key to the SOE reform is marketization and
SOE media content What is the main content of the video? adhering to the principle of making SOEs competitive and independent
market players. 3 = Many problems exist in the SOE reform; we should be
A-21
Agree policy toward SOE reform is 0 = Completely incorrect; 1 = a little bit incorrect; 2 = in general correct;
Do you think the policy toward the SOE reform described in the video is correct?
correct (0, 1, 2 or 3) 3 = completely correct.
1 = SOEs should be bigger and stronger; the SOE reform should be primarily
led by the government; the key to its success is top-level design. 2 =
Marketization is the key to the SOE reform; we should allow inefficient SOEs
Agree to market reform (1, 2 or 3) Which of the following is closest to your opinion toward the SOE reform?
to be driven out by the market. 3 = SOEs are state-owned assets; as long as
the value of these assets is preserved and increases, any measure is
justifiable.
Index of market-oriented reform: Both SOE and private enterprises are pillars of the Chinese economy; therefore, there
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 1 is no need to give special treatment to SOEs.
Index of market-oriented reform: SOEs bears a lot of policy burden; therefore, it is totally reasonable for the state to
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 2 support them with (special) policies.
Index of market-oriented reform: The goal is to have the market to play a decisive role in resource allocation; hence,
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 3 to maintain a high or low proportion of SOEs in the economy is not the goal.
Index of market-oriented reform: Private enterprises contribute 80% of the employment in China; therefore, they need
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 4 more support than the SOEs.
Index of market-oriented reform: Sectors related to national security and important to the national economy and
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 5 people's livelihoods must be controlled by state-owned enterprises.
Index of market-oriented reform: The government should not use fiscal resources to give subsidies to companies that
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 6 are on the verge of bankruptcy but hire many employees.
Index of market-oriented reform: The government should use various measures to help SOEs go out and have them
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 7 hold advantageous positions in international competition.
Index of market-oriented reform: For government procurement projects, SOEs should receive more preferable
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 8 treatment than private enterprises.
Index of market-oriented reform: The government should introduce market competition to sectors in which SOEs are
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 9 monopolies in order to strengthen the dynamism of SOEs.
Pass FP manipulation check The previous video concerns the relationship between China and what country? 1 = the Philippines; 0 = Vietnam, the United State; Indonesia; do not know.
order to maintain peace and stability of the region, China should table
sovereign disputes with the Philippines and the two countries should jointly
extract natural resources.
Index of dovish foreign policy: On the issue of South China Sea, sovereignty is the premise; only under this premise,
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 1 cooperation on economic and trade issues can be discussed.
Index of dovish foreign policy: Economic interest is an important component of the national interest. (We) should
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 2 table the disputes on sovereignty and maximize economic interests.
Index of dovish foreign policy: To maintain a long-term stable cooperative relationship with the ASEAN countries
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 3 including the Philippines is more important than de facto control of a few islands.
Index of dovish foreign policy: Economy is a short-term interest and territory is a long-term interest. (We) should
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 4 seize any opportunity to strengthen the control of islands in the South China Sea.
Index of dovish foreign policy: The government should safeguard national unity and territorial integrity through
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 5 diplomatic and economic measures as much as possible to avoid military conflicts.
Index of dovish foreign policy: If it is in the nation's interest, China can unilaterally impose economic or trade
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 6 sanctions on other countries.
Index of dovish foreign policy: The government should attach as much importance to the development military
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 7 power as it does to the development of the economy.
Index of dovish foreign policy:
When faced with military conflicts, China should not use preemptive military tactics. 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 8
Index of dovish foreign policy: China should forgo the diplomatic principle of "to hide one's capacities and bide one's
5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
component 9 time" and defend its rights more resolutely on international platforms.
Media source What do you think is the source of the two videos? 1 = CCTV; 0 = local TV news; Internet news; others.
Other indicators
Patriotism: component 1 I am happy that I am a Chinese. 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
Patriotism: component 2 In general, China is better than most other countries. 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
Patriotism: component 3 When other people criticize China, I feel that they are criticizing me. 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
Even if I can choose to become a citizen of another country, I prefer to be a Chinese
Patriotism: component 4 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
citizen.
Patriotism: component 5 I feel proud and lucky to be Chinese. 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
Nationalism: component 1 Unless China achieves modernization, foreign countries will manage to exploit China. 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
Electronic copy available at: https://ssrn.com/abstract=3579148
The modern diplomatic history of China is a history of China being victimized and
Nationalism: component 2 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
beaten by foreign powers.
"The century of humuiliation" describes not only China's past, but also how foreigners
Nationalism: component 3 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
treat China today.
Foreign criticisms of China always have ulterior motives and we as Chinese should
Nationalism: component 4 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
refute them resolutely.
China should be the leader of East Asia because of its long history and splendid
Nationalism: component 5 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
civilization.
Trust in the central government Do you trust the central government? 10-point Likert scale. 1 = not trust at all; 10 = trust completely.
Trust in the local government Do you trust the local government? 10-point Likert scale. 1 = not trust at all; 10 = trust completely.
Efficacy: component 1 People like me do not have the ability to evaluate government performance. 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
Efficacy: component 2 Government officials do not care about opinions of people like me. 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
A-23
Efficacy: component 3 The government can timely respond to opinions and suggestions of ordinary citizens. 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
Efficacy: component 4 Politics is too complicated for people like me to understand. 5-point Likert scale. 1 = strongly disagree; 5 = strongly agree.
Having used VPN Have you used a VPN during the past year? 1 = yes; 0 = otherwise.
Have you provided suggestions to the government using various channels during the
Complaint-making 1 = yes; 0 = no.
past year?
Voting in local NPC elections Have you voted in a local NPC election during the past year? 1 = yes; 0 = no.
Discuss government policies Do you discuss topics concerning government policies? 1 = almost never; 2 = rarely; 3 = often; 4 = almost every day.