Lecture 10: Media 1 / 19 Media I Many authoritarian regimes around the world block foreign websites to stop uncensored information from reaching citizens. I Even in countries without state censorship there may be other forms of control over the media (Manufacturing consent by Noam Chomsky) I Moreover, even in the presence of a totally free media, people may not necessarily seek the truth as seen in the rise of conspiracy theories in the age of the internet (vaccine hesitancy, beliefs such as moon landings were a hoax, and the earth is flat). I Does a free media really matter? 2 / 19 Censorship in China I Domestic media outlets in China are heavily censored. I This includes self censorship from the fear of severe business and political costs from publishing content that the state deems threatening and objectionable. I Censorship also follows from orders of the Propaganda Department of the Communist Party of China. I Among the most heavily censored topics in 2016 are government corruption, media censorship, civil society activism, ethnic tensions, health, and safety scandals. I Transmission of politically sensitive information on domestic social media such as Weibo and WeChat is also limited due to platform-wide keyword filters and ex-post content deletion. 3 / 19 The great firewall I An important aspect of China’s internet regulation is its effort to block internet users in China from accessing specific foreign websites. I Since 2003 the great firewall serves as the main infrastructure blocking access to potentially unfavourable incoming data from foreign media outlets. I As of 2016 (the time of the experiment) the great firewall blocks entire websites or specific web pages from being accessed by IP addresses located in China. I Some prominent examples are: Google, YouTube, Facebook, Twitter, Instagram, Blogspot, Tumblr, Dropbox, Blogger, Vimeo, Soundcloud, and Flickr. I Examples of blocked news websites include CNN, The New York Times, The Guardian, BBC, Bloomberg, The Wall Street Journal, and Reuters. 4 / 19 Experiment I The authors of the next study conducted an experiment with 1800 university students in Beijing I They gave students access to a virtual proxy network (VPN) to bypass the great firewall with encouragement to do so I The aim of the experiment was to: 1. Does providing access to an uncensored internet lead citizens to acquire politically sensitive information? 2. Does acquisition of such information affect their attitudes and behaviour? 5 / 19 we randomly assign them to either a control condition in which they are subject to censorship as in the status quo, or an access (A) treatment in which they receive Design free access to uncensored internet for 18 months. Among a random subgroup of the students who receive the access treatment, we also assign an encouragement (E ) treatment, where we encourage them to visit foreign news websites blocked by the Great Firewall. No access to uncensored internet Access treatment Encouragement treatment No [C]ontrol (status quo) N = 185 Free access to uncensored internet Yes No Yes [C]ontrol + [E]ncouragement [A]ccess [A]ccess + [E]ncouragement N = 328 N = 313 N = 650 Existing users N = 331 source: Chen and Yang (2019) To address concerns that the encouragement treatment alone generates an experimenter demand effect (e.g., explicit endorsement of specific foreign websites by the researchers) or changes students’ perception of government suppression, we also provide the encouragement treatment to a random subgroup of students 6 / 19 in the control group. These students are presented with the same encouragement Design I I The experiment was conducted between November 2015 to April 2017. After an initial survey of students, existing users of VPNs were filtered out of the experiment. Students were randomly assigned to one of four groups: 1. 2. 3. 4. Control Control + Encouragement. Access Access + Encouragement 7 / 19 Design I I Access: 18 months free access to VPN (valued at 25 USD a month) Encouragement: I I I 4 newsletters introducing students to variety of foreign websites that are blocked highlighting that politically sensitive news is often reported differently in domestic and foreign outlets News quizzes with modest monetary incentives The encouragement treatment is only active between December 2015 to March 2016 8 / 19 Table 1—Summary Statistics, Attrition, Balance Tests Variables Male Birth year Height Han ethnicity Born in coastal province Resided in coastal province Urban hukou prior to college Religious Member of CCP [at baseline] Educational background English ability Oversea travel experiences Household characteristics Risk preferences Time preferences Altruism Reciprocity Sample and attrition Baseline Endline Mean Mean t-test (SD) (SD) p-value (1) (2) (3) 0.559 (0.497) 1995.8 (1.278) 170.1 (9.311) 0.912 (0.283) 0.417 (0.493) 0.444 (0.497) 0.784 (0.412) 0.066 (0.248) 0.068 (0.252) 0.000 (1.000) 0.000 (1.000) 0.000 (1.000) 0.000 (1.000) 0.000 (1.000) 0.000 (1.000) 0.000 (1.000) 0.000 Treatment balance Exg. users Mean (4) C Mean (5) CE Mean (6) 0.562 0.843 0.630 0.512 0.579 (0.496) 1995.8 0.696 1995.9 1995.9 1995.8 (1.262) 169.9 0.576 170.6 168.0 170.3 (8.951) 0.914 0.844 0.921 0.898 0.896 (0.280) 0.415 0.918 0.438 0.372 0.398 (0.492) 0.439 0.753 0.474 0.358 0.413 (0.496) 0.771 0.402 0.835 0.752 0.764 (0.420) 0.066 0.958 0.050 0.058 0.093 (0.249) 0.064 0.660 0.058 0.044 0.066 (0.245) −0.028 −0.140 −0.112 0.441 0.086 (0.980) 0.512 0.008 0.045 −0.023 −0.094 (0.986) 0.416 0.208 −0.029 −0.158 −0.048 (0.988) 0.604 0.010 −0.018 −0.067 −0.077 (0.967) 0.639 0.109 −0.017source: −0.011 Chen and Yang (2019) 0.031 (1.008) 0.930 0.057 −0.003 −0.018 −0.044 (0.995) 0.006 0.874 0.011 −0.012 −0.006 (0.995) −0.071 −0.007 0.014 0.702 0.246 A Mean (7) 0.597 AE Mean (8) 0.517 ANOVA test p-value (9) 0.119 1995.7 1995.8 0.252 170.4 169.6 0.063 0.922 0.920 0.585 0.471 0.398 0.168 0.512 0.420 0.017 0.705 0.781 0.150 0.061 0.065 0.430 0.057 0.076 0.549 0.157 −0.109 0.002 −0.089 −0.019 0.383 −0.152 −0.035 0.284 0.068 −0.035 0.284 −0.023 −0.101 0.381 −0.025 0.004 0.939 0.118 −0.041 0.223 0.005 0.006 0.076 9 / 19 Member of CCP [at baseline] Educational background English ability Oversea travel experiences Household Variables characteristics Male Risk preferences Birth Time year preferences Height Altruism Han ethnicity Reciprocity 0.068 0.064 0.660 0.058 0.044 0.066 0.057 (0.252) (0.245) 0.000 0.441 0.086 0.157 −0.028 −0.140 −0.112 1—Summary Statistics, Attrition, Balance Tests (1.000)Table (0.980) 0.000Sample −0.023 0.512 0.008 0.045 −0.094 −0.089 and attrition Treatment balance (1.000) (0.986) Baseline Endline Exg. 0.000 0.416 0.208 −0.029 −0.158 −0.048 −0.152 Mean Mean t-test users C CE A (1.000) (0.988) (SD) (SD) p-value Mean Mean Mean Mean 0.000 0.604 0.010 0.068 −0.018 −0.067 −0.077 (1) (2) (3) (4) (5) (6) (7) (1.000) (0.967) 0.559 0.562 0.843 0.630 0.512 0.579 0.597 0.000 0.639 0.109 0.031 −0.017 −0.011 −0.023 (0.497) (0.496) (1.000) (1.008) 1995.8 1995.8 0.696 0.000 0.930 1995.9 0.057 1995.9 −0.003 −0.018 1995.8 −0.044 1995.7 −0.025 (1.278) (1.262) (1.000) (0.995) 170.1 169.9 0.576 170.6 168.0 170.3 170.4 0.000 0.006 0.874 0.011 0.118 −0.012 −0.006 (9.311) (8.951) (1.000) (0.995) 0.912 0.914 0.844 0.921 0.898 0.896 0.922 0.000 0.014 0.702 0.246 0.005 −0.071 −0.007 (0.283) (0.280) (1.000) (1.007) 0.417 0.415 0.918 0.438 0.372 0.398 0.471 1,807 1,372 — 242 137 259 244 (0.493) (0.492) 0.076 0.549 −0.109 0.002 −0.019 0.383 −0.035 AE Mean −0.035 (8) ANOVA 0.284 test p-value 0.284 (9) 0.517 −0.101 0.119 0.381 1995.8 0.004 0.252 0.939 169.6 −0.041 0.063 0.223 0.920 0.006 0.585 0.076 Born in coastal 0.398 0.168 Number of 490 — province observations Resided in coastal 0.444 0.439 0.753 0.474 0.358 0.413 0.512 0.420 0.017 province (0.497) (0.496) Notes: Mean level of each characteristics are reported in column 1 for all participants who have completed baseline Urban hukou prior 0.784 0.771 0.402 0.835 0.752 0.764 0.705 0.781 0.150 survey in November(0.412) 2015 (and corresponding standard deviation in parentheses), column 2 for participants who to college (0.420) completed endline survey in April 2017 (and corresponding standard deviation in parentheses), column 4 for Religious 0.066 0.066 0.958 0.050 0.058 0.093 0.061 0.065 0.430 endline participants (0.248) who use censorship (0.249) circumvention tools prior to the baseline survey, column 5 for endline participants in the control group (C ), column 6 for endline participants in the control + encouragement group Member of CCP 0.068 0.064 0.660 0.058 0.044 0.066 0.057 0.076 0.549 (CE column 7 for(0.252) endline participants [at ), baseline] (0.245) in the access group (A), and column 8 for endline participants in the access + encouragement group (AE ). All characteristics in the “personal characteristics” (category F.1 in survey) Educational 0.000 0.441 0.086 0.157 0.002 −0.028 −0.140 −0.112 −0.109 arebackground presented. Characteristics in(0.980) “educational background” (category F.2), “English ability and oversea travel (1.000) experiences [at baseline]” (category F.3), “household characteristics” (category F.4), and “fundamental preferences” English ability 0.000 0.512 0.008 0.045 0.383 −0.023 −0.094 −0.089 −0.019 (category F.5) are summarized by(0.986) z-scores. For each characteristic, a t-test is conducted against the null hypothesis (1.000) that students who have completed baseline survey and those who completed endline are not different from each Oversea travel 0.000 0.416 0.208 0.284 −0.029 −0.158 −0.048 −0.152 −0.035 other in term of this characteristic; column 3 reports the corresponding p-value for each test. For each characteristic, experiences (1.000) (0.988) an ANOVA test is conducted against the null hypothesis that students in the control, control + encouragement, Household 0.000 0.604 0.010 0.068 0.284 −0.018 −0.067 −0.077 −0.035 access, and access +(1.000) encouragement groups are not jointly different from each other in term of this characteristic; characteristics (0.967) column 9 reports the corresponding p-value for each test. Risk preferences 0.000 0.639 0.109 0.031 0.381 −0.017 −0.011 −0.023 −0.101 (1.000) (1.008)source: Chen and Yang (2019) Time preferences 0.000 0.930 0.057 0.004 0.939 −0.003 −0.018 −0.044 −0.025 (1.000) (0.995) Altruism 0.000 0.006 0.874 0.011 0.118 0.223 −0.012 −0.006 −0.041 (1.000) (0.995) Reciprocity 0.000 across 0.014 treatment 0.702 0.246the −0.007 0.006 effects 0.076 −0.071 differential attrition groups, and estimated0.005 treatment (1.000) (1.007) 10 / 19 Online behaviour of the access groups 2312 2312 JUNE JUNE 2019 2019 THE THE AMERICAN AMERICAN ECONOMIC ECONOMIC REVIEW REVIEW Table Table 2—Browsing 2—Browsing Activities Activities on on Foreign Foreign Websites Websites Access Access Mean Mean (1) (1) SD SD (2) (2) 39.6% 39.6% 49.0% 49.0% Access + + Access encouragement encouragement Mean SD Mean SD (3) (4) (3) (4) Panel Panel A. A. Extensive Extensive margins margins (( percent percent of of students), students), among among all all students students Activated accounts accounts 54.6% 49.9% 68.2% Activated 54.6% 49.9% 68.2% Active users users Active 45.5% 45.5% 46.6% 46.6% 49.8% 49.8% Panel B. B. Intensive Intensive margins margins (minutes (minutes per per day), day), among among all all students students Panel Total daily daily browsing browsing time time 31.45 64.99 31.87 Total 31.45 64.99 31.87 Google and and related related services services 6.96 13.76 7.09 Google 6.96 13.76 7.09 YouTube 3.64 8.79 4.46 YouTube 3.64 8.79 4.46 Facebook 3.09 7.28 3.27 Facebook 3.09 7.28 3.27 Twitter 2.79 7.29 2.96 Twitter 2.79 7.29 2.96 Top foreign news websites 0.10 0.23 0.59 Top foreign news websites 0.10 0.23 0.59 New York Times 0.07 0.18 0.56 New York Times 0.07 0.18 0.56 Informational websites 2.98 5.84 3.17 Informational websites 2.98 5.84 3.17 Wikipedia 0.05 0.19 0.54 Wikipedia 0.05 0.19 0.54 Entertainment websites 9.07 15.42 8.90 Entertainment websites 9.07 15.42 8.90 Pornographic websites 2.44 8.13 2.52 Pornographic websites 2.44 8.13 2.52 59.14 59.14 13.29 13.29 11.44 11.44 7.83 7.83 7.70 7.70 0.65 0.65 0.61 0.61 5.51 5.51 1.78 1.78 14.53 14.53 8.74 8.74 Chen and Yang (2019) Panel (minutes day), among active users Panel C. C. Intensive Intensive margins marginssource: (minutes per per day), among active users Total daily daily browsing browsing time time 79.17 83.13 69.59 Total 79.17 83.13 69.59 Google and and related related services services 17.50 17.16 15.22 Google 17.50 17.16 15.22 YouTube 9.12 12.06 9.75 YouTube 9.12 12.06 9.75 Facebook 7.74 9.92 7.14 Facebook 7.74 9.92 7.14 Twitter 7.05 10.22 6.51 Twitter 7.05 10.22 6.51 Top foreign foreign news news websites websites 0.25 0.30 1.19 Top 0.25 0.30 1.19 71.20 71.20 16.31 16.31 15.37 15.37 10.36 10.36 10.36 10.36 0.43 0.43 p-value p-value (5) (5) <0.001 <0.001 0.083 0.083 0.922 0.922 0.885 0.885 0.263 0.263 0.732 0.732 0.742 0.742 <0.001 <0.001 <0.001 <0.001 0.612 0.612 <0.001 <0.001 0.871 0.871 0.898 0.898 0.232 0.232 0.198 0.198 0.686 0.686 0.587 0.587 0.625 0.625 <0.001 <0.001 11 / 19 2312 Facebook Twitter 3.09 7.28 THE AMERICAN ECONOMIC REVIEW3.27 2.79 7.29 2.96 7.83 7.70 Online behaviour of the access groups Top foreign news websites 0.10 0.23 0.59 0.65 Foreign Websites New York Times Table 2—Browsing Activities 0.07 on 0.18 0.56 0.61 Informational websites Wikipedia Entertainment websites Pornographic websites 2.98 5.84 0.05Access 0.19 9.07 Mean (1) 2.44 15.42 SD (2) 8.13 3.17 Access +5.51 encouragement 0.54 1.78 8.90 14.53 Mean SD (3) (4) 2.52 8.74 Panel A. margins (minutes ( percent per of students), among all users students Panel C. Extensive Intensive margins day), among active Activated 54.6% 49.9% 68.2% Total dailyaccounts browsing time 79.17 83.13 69.59 Active users 39.6% 49.0% 45.5% Google and related services 17.50 17.16 15.22 YouTube 9.12 12.06 9.75 Panel B. Intensive margins (minutes per day), among all students Facebook 7.74 9.92 7.14 Total daily browsing time 31.45 64.99 31.87 46.6% 71.20 49.8% 16.31 0.732JUNE 2019 0.742 <0.001 <0.001 0.612 <0.001 0.871 p-value (5) 0.898 <0.001 0.232 0.083 0.198 15.37 0.686 0.587 0.922 0.625 0.885 <0.001 0.263 <0.001 0.732 0.284 0.742 Twitter Google and related services Top foreign news websites YouTube 7.05 6.96 0.25 3.64 10.22 13.76 0.30 8.79 6.51 7.09 1.19 4.46 10.36 59.14 10.36 13.29 0.43 11.44 New York Times Facebook Informational websites Twitter Wikipedia Top foreign news websites Entertainment websites New York Times Pornographic websites Informational websites 0.18 3.09 7.50 2.79 0.14 0.10 22.79 0.07 6.16 2.98 0.25 7.28 7.24 7.29 0.29 0.23 16.97 0.18 12.03 5.84 1.13 3.27 6.72 2.96 1.19 0.59 19.22 0.56 5.49 3.17 0.38 7.83 6.59 7.70 2.48 0.65 16.37 0.61 12.32 5.51 <0.001 <0.001 0.044 <0.001 0.613 0.612 Wikipedia 0.05 0.19 0.54 1.78 <0.001 Notes: Panel A shows the composition among students who received only the access treatment Entertainment 9.07and encouragement 15.42 8.90 14.53(Group-AE 0.871). (Group-A) andwebsites those who received both access treatments They are divided into two nested categories: have Pornographic websites 2.44(i) Activated 8.13 accounts, 2.52 students 8.74 who0.898 activated the censorship circumvention tool provided during the experiment, as of April 10, 2017 last day margins of the experiment); (ii)among Activeactive users,users students who have activated the Panel (the C. Intensive (minutes perand day), tool were actively using the tool (if the79.17 student’s 83.13 account records one browsing Totaland daily browsing time 69.59 at least 71.20 0.232 activity per day for more than 40 days after the encouragement treatment ends). Panel B shows Google and daily relatedbrowsing servicestime in total and17.50 16.31 0.198 the average on various17.16 categories15.22 of websites throughout the experiment after the encouragement treatment ends, among all students (assuming students YouTube 9.12 12.06 9.75 15.37 0.686 without activated accounts spend zero minutes7.74 on these 9.92 websites). 7.14 Panel C replicates panel B, Facebook 10.36 0.587 but among students who actively used the tool. Top foreign news websites, informational, Twitter 10.22 6.51on Alexa 10.36Top Websites 0.625 entertainment, and pornographic websites are7.05 defined primarily based categorization. Column 5 shows p-values of two-sided t-tests extensive0.43 margins<0.001 and the Top foreign news websites 0.25 0.30 on the1.19 intensive margins students. New York Timesbetween the Group-A and Group-AE 0.18 0.25 1.13 0.38 <0.001 source: Chen and (2019)6.72 Informational websites 7.50 Yang7.24 6.59 0.284 Wikipedia 0.14 0.29 1.19 2.48 <0.001 Entertainment websites 22.79 16.97 19.22 16.37 0.044 12 / 19 Trump s Inauguration Chinese New Year holiday week Trump calls President of Taiwan US Presidential Election National Day holiday week 2016 Rio Olympics 50th anniversary of Cultural Revolution Panama Papers Incentivized encouragement period Unincentivized encouragement period 2313 CHEN AND YANG: THE IMPACT OF MEDIA CENSORSHIP 30 8 20 4 10 0 0 20 16 -0 1 20 16 -0 20 2 16 -0 3 20 16 -0 20 4 16 -0 5 20 16 -0 6 20 16 -0 7 20 16 -0 8 20 16 -0 9 20 16 -1 0 20 16 -1 1 20 16 -1 2 20 17 -0 1 20 17 -0 2 20 17 -0 3 20 17 -0 4 12 Percent pol. sensitive articles on NYTimes Total browsing time on NYTimes per week (min) VOL. 109 NO. 6 Access Access + Encour. Percent sensitive articles on NYTimes Figure 1 Notes: Average total browsing time (minutes) on the New York Times per week, among students received only the access treatment (Group-A) and those who received both access and encouragement treatments (Group-AE). New York Times browsing time includes both its English and Chinese websites. Dotted line ( y-axis on the right-hand side) indicates the proportion of articles published on the New York Times that are politically sensitive during that corresponding week. source: Chen and Yang (2019) B. Temporary Encouragement Boosts Immediate Information Acquisition 13 / 19 Effect on willingness to pay 2316 JUNE 2019 THE AMERICAN ECONOMIC REVIEW WTP for uncensored internet access ($/month) 40 35 30 25 20 15 Nov. 2015 Apr. 2016 May 2017 Control Access Access + Encour. Existing users Figure 2 Notes: Average level of willingness to pay for accessing censorship circumvention tools (US$ per month), elicited using a BDM method, among students in control group (Group-C, pooling C and CE students together), those who received only the access treatment (Group-A), those who received both access and encouragement treatments (Group-AE), and the existing users, across the baseline survey (November 2015), midline survey (May 2016), and endline survey (April 2017). Sample is restricted to 1,372 students who have completed the endline survey. source: Chen and Yang (2019) they pay on average US$21.50 up front for a seasonal subscription (darker bars in Figure 3).22 If we count the students who intend to purchase censorship circumvention 14 / 19 8 8 7 0 0 0 7 .5 .5 .5 5 4 Baseline Midline Endline Midline Endline Baseline Awareness of foreign protests Midline 6 5 4 5 4 Baseline Awareness of protests in Greater China Endline Baseline Midline Endline Self-assessment of knowledge level 1 Endline Endline Baseline Midline Baseline Midline Endline .8 Baseline Midline .6 Endline .4 .4 .15 .4 .4 .1 Awareness of fo .6 -.5 -1 Midline Awareness of fo .8 0 Awareness of protests in Greater China Currently invested in Chinese stock mkt. .2 .4 1 .5 1 Willingness to act -1 -.5 -.5 0 0 .5 .5 1 0 .5 -1 -.5 -.5 0 -1 1 0 .8 Baseline .25 .4 .5 .3 Midline Midline .8 -.5 -1 Endline Location pref.: foreign cities .3 .4 1 Midline Censorship i Censorship i Awareness of protests in Greater China .6 Baseline Baseline Baseline .6 Endline Plan: master degree abroad .4 Endline Trust in non-domestic media outlets Living in democracy is not important .5 1 0 -1 -.5 .8 Midline Endline Endline % quizzes answered correctly: non-sensitive .6 Baseline Satisfaction of govt. performance Midline Midline .6 Endline Participation behaviors Midline -1 Baseline 1 Endline % quizzes answered correctly: non-sensitive .8 0 -.5 -1 Midline .6 Baseline Baseline Baseline .5 Endline Trust in non-domestic media outlets -1 .2 Midline .5 1 1 Baseline 0 .5 1 Endline -1 Trust in foreign govt. .5 2 .5 -1 .8 .8 -.5 .6 Endline Valuation of access to foreign media outlets -1 -.5 -.5 0 Endline Midline Midline % quizzes answered correctly: sensitive Endline Endline Baseline .2 Baseline Midline Endline Baseline Midline Midline Endline Endline .2 .2 Endline Midline .2 .05 .15 Midline Midline Baseline Baseline Midline Baseline Midline 1 Confidence of guesses on Chinese economy Confidence of guesses on Chinese economy Access + Encour. .5 1 1 .5 1 C.C. Economic Economic beliefs beliefs Optimistic belief of Chinese economy Control Access Baseline Baseline Endline 0 Endline Midline Baseline .1 .2 Endline Endline Optimistic belief of Chinese economy .05 Midline Midline Midline .2 .2 Baseline Baseline Endline .1 Baseline -1 Midline Valuation of access to foreign media outlets .2 6 6 1 5 .5 5 4 0 4 -.5 3 Midline Trust in Chinese govt. .4 1 .5 0 -.5 -1 E. Behaviors Baseline Midline 0 Endline Baseline % quizzes answered correctly: sensitive .2 Midline Social interaction in politics Baseline Baseline Baseline .2 1 .5 0 -.5 -1 D. Political attitudes Demand for institutional change Baseline Endline -1 3 Endline Midline Frequently visit foreign websites for info 2 Midline Confidence of guesses on Chinese economy 1 1 .5 0 -.5 -1 C. Economic beliefs Baseline Frequently visit foreign websites for info Baseline 0 Endline -.5 Midline Optimistic belief of Chinese economy .2 -.5 .4 .4 .4 0 .6 .6 .6 .6 .4 .2 Baseline A.A.Media-related Media-related behaviors, beliefs behaviors, beliefs B.B. Knowledge Knowledge and attitudes and attitudes B. Knowledge .5 .8 .8 -.5 6 Endline % quizzes answered correctly: non-sensitive 1 Midline -1 -.5 Baseline .8 Endline % quizzes answered correctly: sensitive .8 Midline -1 -.5 -1 3 2 Baseline .5 Media-related ehaviors, beliefs nd attitudes Effect on attitudes, beliefs, and behaviour Existing users 0 .5 0 .5 Yang (2019)group (Group-C, pooling C and CE students Figure A.9: Average level of outcomes elicited in the panelsource: survey, Chen amongand students in control -1 -.5 -.5 0 -1 -1 -.5 -.5 0 1 1 D.D. P a 1 Midline Endline Trust in Chinese govt. Trust in Chinese govt. Trust in foreign govt. Trust in foreign govt. Satisfaction of go 1 Baseline 1 Endline 1 Midline Demand for institutional change Demand for institutional change 1 -1 Baseline 1 A.52 A.52 together), those who received only the access treatment (Group-A), those who received both access and encouragement treatments (Group-AE) and the existing users, across the baseline survey (November 2015), midline survey (May 2016), and endline survey (April 2017). If there are multiple survey questions elicited within a category or subcategory of outcomes, results are shown using a z-score index aggregating all corresponding questions. The z-score index (weighting by the inverse covariance of the standardized variables) is computed following Anderson (2008). Sample is restricted to 1,372 students who have survey. Endline Baseline Midline Endlinecompleted Baselinethe endline Midline Satisfaction of go 15 / 19 8 8 7 0 0 0 7 .5 .5 .5 5 4 Media-related ehaviors, beliefs nd attitudes Endline Midline Endline 6 5 4 Baseline Awareness of protests in Greater China Midline Endline Baseline Awareness of foreign protests Midline 4 5 -.5 -1 Baseline Endline Baseline Midline Endline Self-assessment of knowledge level Baseline Midline Midline Endline -1 Degree of censorship on foreign news outlets Censorship is .5 8 .5 9 8 .5 9 1 5 Trust in non-domestic media outlets Baseline 1 1 Endline 1 .2 .2 6 1 Midline 0 70 70 .5 0 4 .5 Baseline Endline Midline Baseline -1 -.5 1 .5 1 4 .5 .8 0 Midline Endline Baseline Midline .6 -1 -.5 0 Midline Awareness of fo Endline .4 .15 .1 .05 Endline Midline Endline Endline .2 .2 .1 Midline .05 Baseline Baseline Endline Baseline Baseline Baseline Midline Midline Endline Baseline Midline Endline 1 Confidence of guesses on Chinese economy Access Access + Encour. Existing users .5 Control .5 C. Economic beliefs 1 .15 .2 -.5 -1 Endline Midline Baseline Endline Endline Currently invested in Chinese stock mkt. .3 .2 Endline .1 Midline Midline Midline Optimistic belief of Chinese economy Baseline .25 .40 .4 .5 .4 Baseline Baseline Baseline .8 -.5 -1 Endline Location pref.: foreign cities .3 .4 1 Midline Willingness to act Awareness of protests in Greater China .6 Baseline Plan: master degree abroad 0 .2 Endline Endline 1 0 -1 Endline Midline Midline .4 Midline -.5 1 Baseline Baseline Baseline .2 Endline Living in democracy is not important 0 Midline Participation behaviors Endline Endline Self-assessment of knowledge level .5.6 Baseline Satisfaction of govt. performance Midline Midline % quizzes answered correctly: non-sensitive .8 0 -.5 -1 .8 .8 -.5 -1 Awareness of foreign protests Baseline Baseline .5 4 1 .5 .5 Trust in foreign govt. Endline Endline -1 5 6 -.5 -1 5 Endline 6 -.5 0 Midline Trust in Chinese govt. -1 -1 .2 -.5 0 .5 1 E. Behaviors Midline Endline Degree of censorship on domestic news outlets % quizzes answered correctly: sensitive .6 Endline Baseline Baseline .6 .5 0 -.5 -1 B. Knowledge Midline Social interaction in politics Valuation of access to foreign media outlets Midline -.5 -1 Midline Endline Baseline Baseline 3 2 -.5 1 D. Political attitudes Midline Baseline Demand for institutional change Awareness of protests in Greater China Endline Censorship is unjustified 1 Endline Midline -1 Midline Confidence of guesses on Chinese economy -.5 1 .5 0 -.5 -1 C. Economic beliefs Baseline Midline Frequently visit foreign websites for info Baseline 0 Endline .2 -.5 .4 .4 .4 0 .6 .6 .5 .8 .6 .6 .4 .2 A. Media-related behaviors, beliefs and attitudes Midline Optimistic belief of Chinese economy Baseline Midline % quizzes answered correctly: non-sensitive .8 B. Knowledge Baseline Trust in non-domestic media outlets Baseline -.5 Baseline 1 Endline .8 Midline % quizzes answered correctly: sensitive .8 Baseline -1 2 -1 3 -.5 6 Effect on attitudes, beliefs, and behaviour Yang (2019)group (Group-C, pooling C and CE students Figure A.9: Average level of outcomes elicited in the panelsource: survey, Chen amongand students in control 0 -1 -.5 0 -.5 Trust in foreign govt. Willingness to act Satisfaction of go 1 Living in democracy is not important 1 1 Trust in Chinese govt. 1 Satisfaction of govt. performance 1 -1 Demand for institutional change 1 D. P a Trust in foreign govt. 1 A.52 together), those who received only the access treatment (Group-A), those who received both access and encouragement treatments (Group-AE) and the existing users, across the baseline survey (November 2015), midline survey (May 2016), and endline survey (April 2017). If there are multiple survey questions elicited within a category or subcategory of outcomes, results are shown using a z-score index aggregating all corresponding questions. The z-score index (weighting by the inverse covariance of the standardized variables) is computed following Anderson (2008). Sample is restricted to 1,372 students who have endline survey. Endline Baseline Midline Endline completed Baseline the Midline 16 / 19 .2.2 9 1 Baseline Baseline Midline Midline 8 8 7 7 .5 Midline Endline 6 4 Baseline Awareness of foreign protests Midline Endline Baseline Midline Endline Self-assessment of knowledge level 0 Endline Endline Trustininforeign foreigngovt. govt. Trust Endline Midline Satisfactionof ofgo go Satisfaction 11 Baseline 11 .2 Endline 11 Endline -.5 Midline Midline TrustininChinese Chinesegovt. govt. Trust Baseline Midline -1 Baseline Baseline .4 .4 -1-1 .6 .6 .5 1 .8 -.5 -.5 5 6 5 4 -1 Baseline .5.5 00 .5.5 00 00 .5.5 1 .5.5 .5 00 0 -.5 -.5 Midline Endline .3.3 .25 .25 Baseline Midline Midline Access Endline .15 .1.1 .05 Midline Midline Baseline Endline Endline Control Control Midline Endline Access + Encour. 0 .1 Baseline Baseline .1.1 .15 Endline Endline Baseline Endline .1 .2.2 .25 Control Endline .05 Midline Midline Midline -1-1 .2 Baseline .1 Baseline Baseline -1-1 1 .5 -.5 -1 Baseline Currently invested in Chinese stock mkt. .2 .3 -.5 -.5 .5 -.5 -.5 Endline Midline Midline Locationpref.: pref. Location .2.2 Endline Baseline Baseline .15 .15 00 1 Midline Location pref.: foreign cities Endline Endline 0 0 -.5 Baseline Willingness to act Plan:master masterdegree degreeabroad abroad Plan: -1 -1 Endline Plan: master degree abroad .4 00 Midline Midline Midline Baseline Baseline Baseline Midline Midline Midline Endline Endline .05 .05 Baseline .3.3 Endline .3 Midline Participation behaviors Baseline Baseline .4.4 0 -.5 11 Baseline .5.5 -.5 -1 -1 .5.5 -.5 11 Living in democracy is not important .5 .5 Endline Endline Participationbehaviors behaviors Participation 0 Socialinteraction interactionininpolitics politics Social 0 Midline Midline -1-1 Satisfaction of govt. performance Baseline Baseline .5 Endline Endline 1 Trust in foreign govt. Midline Midline 1 -1-1 Endline 1 1 -.5 -.5 -.5 -.5 -.5 Midline Trust in Chinese govt. -.5 -1-1 Midline Endline Endline Degree of censorship on foreign news outlets 0 -.5 .5.5 00 Endline Midline -.5 -.5 -1 Baseline Baseline Baseline -1 -.5 Baseline Midline Midline 0 1 .5 0 Midline Endline Endline Demandfor forinstitutional institutionalchange change Demand Midline Endline Baseline .5 1 -1 -.5 0 .5 Endline Social interaction in politics -1 E. Behaviors E. E. Behaviors Behaviors D. Political attitudes Midline Baseline Baseline Degree of censorship on domestic news outlets Confidence of guesses on Chinese economy Endline Demand for institutional change Baseline Endline Endline Censorship is unjustified Awareness of protests in Greater China .2 .2 Baseline -1-1 1 .5 0 -.5 -1 C. Economic beliefs Midline Baseline .8 -.5 -.5.8 -1-1 Midline Midline .4 Baseline Baseline Endline Optimistic belief of Chinese economy Baseline Endline 11 D. D. Political Political attitudes attitudes Midline 11 .5 0 -.5 Midline % quizzes answered correctly: non-sensitive .6 .8 .4 .2 A.52 A.52 B. Knowledge Baseline Midline Midline Confidenceof ofguesses guesseson onChinese Chineseeconomy economy Confidence -1 Baseline % quizzes answered correctly: sensitive .6 1 1 11 0 -.5 .5.5 -1 Endline 00 Midline Baseline Baseline Trust in non-domestic media outlets Optimisticbelief beliefof ofChinese Chineseeconomy economy Optimistic 9 Endline Endline .2.2 .2.2 Midline Midline Valuation of access to foreign media outlets .5 6 5 4 3 2 Baseline C. C. Economic Economic beliefs beliefs A. Media-related behaviors, beliefs and attitudes Baseline Baseline Frequently visit foreign websites for info .2.2 .4.4 .4.4 .4.4 .4.4 wledge wledge Effect on attitudes, beliefs, and behaviour Baseline Baseline Midline Midline Endline Existing users Access Access Yang (2019)group (Group-C, pooling C and CE students Figure A.9: Average level of outcomes elicited in the panelsource: survey, Chen amongand students in control together), those who received only the access treatment (Group-A), those who received both access and encouragement treatments (Group-AE) and the existing users, across the baseline survey (November 2015), midline survey (May 2016), and endline survey (April 2017). If there are multiple survey questions elicited within a category or subcategory of outcomes, results are shown using a z-score index aggregating Access Access Figure A.9: A.9: Average Average level level of of outcomes outcomes elicited elicited in in the the panel panel survey, survey, among among students students in in co co Figure 17 /both 19 together), those who received only the access treatment (Group-A), those who received 7 7 .5 6 6 0 0 0 -.5.5 4 C. Economic beliefs 0 Endline 5 4 Baseline Midline Endline Baseline Midline Endline Baseline Midline Endline Self-assessment of knowledge level .5 1 .8 Awareness of foreign protests 4 5 -.5 -1 -.5 -1 .8 -.5.8 .6 Endline Midline Endline Satisfaction of g 1 Midline 1 Baseline 1 Endline 1 1 Trust in foreign govt. Endline Willingness to act -1 .2 Trust in Chinese govt. Baseline Midline Living in democracy is not important Endline 0 -.5 1 -1 Endline Baseline Midline -1 .25 -.5 .3 0 Location pref Baseline Endline Baseline Midline Endline .2 Midline .15 .2 .1 .1 .2 -.5 .25 .15 .3 Currently invested in Chinese stock mkt. Midline Control Endline Baseline Midline Access Access + Encour. Midline Midline Baseline Endline Endline Midline Endline Endline .05 .05 .1 Baseline Baseline .1 .15 Endline Endline Baseline Midline Endline Baseline Midline 0 Midline Midline Midline .05 Baseline Baseline Endline 0 Baseline -1 Endline .1 .05 Baseline -1 -1 -.5 .2 .1 0 .05 .5 .5 0 Endline Location pref.: foreign cities 0 .4 Midline Willingness to act Endline .4 -.5 Baseline Plan: master degree abroad .3 Midline Midline Plan: master degree abroad .3 Endline Baseline -1 -.5 -1 Midline Baseline -11 .5 0 .5 1 Living in democracy is not important Endline Endline 1 .15 .5 Baseline .1 Endline -1 -.5 -.5 0 0 .5 0 0 -.5 -1 1 .5 0 Midline Participation behaviors Midline Midline Participation behaviors Currently invested in Chinese stock mkt. -.5 Baseline Satisfaction of govt. performance Baseline Baseline -1 .5 -1 .25 Endline Endline Social interaction in politics Location pref.: foreign cities 10 .2 -.5 -1 Trust in foreign govt. Midline .5 10 Endline Midline -.5 .3 Midline Trust in Chinese govt. Baseline -.5.5 .15 .5 .5 1 .5 .5 .5 1 -1 1 .5 0 -.5 -1 1 Endline 0 .5 Baseline Baseline -.5 -1 E. Behaviors E. Behaviors D. Political attitudes Midline Social interaction in politics Midline 00 0 Endline Endline Baseline Demand for institutional change Midline Endline Baseline Satisfaction of govt. performance -1-.5 -.5 -.5 Midline .2 -.5 Baseline .4 0 -1 Midline .4 Baseline Confidence of guesses on Chinese economy Demand for institutional change Plan: master degree abroad Access Midline Awareness of protests in Greater China Baseline -1 1 .5 0 -.5 -1 C. Economic beliefs Baseline Midline Baseline -1 .4 .2 1 1 Endline Optimistic belief of Chinese economy seline Endline .6 .8 .6 .4 Midline D. Political attitudes Trust in foreign govt. Midline Midline % quizzes answered correctly: non-sensitive .2 A.52 B. Knowledge Baseline seline Baseline -.5 Endline % quizzes answered correctly: sensitive .6 Midline 0 -1 3 2 Baseline .5 ia-related aviors, beliefs attitudes Effect on attitudes, beliefs, and behaviour Baseline Midline Endline Existing users Control Access + Encour. Access Existing users Yang (2019)group (Group-C, pooling C and CE students Figure A.9: Average level of outcomes elicited in the panelsource: survey, Chen amongand students in control together), those who received only the access treatment (Group-A), those who received both access and encouragement treatments (Group-AE) and the existing users, across the baseline survey (November 2015), midline survey (May 2016), and endline survey (April 2017). If there are multiple survey questions elicited within a category or subcategory of outcomes, results are shown using a z-score index aggregating all corresponding questions. The z-score index (weighting by the inverse covariance of the standardized variables) is computed following Anderson (2008). Sample is restricted to 1,372 students who have completed the endline survey. Acces Figure A.9: Average level of outcomes elicited in the panel survey, among students in co el survey, among students in control group (Group-C, pooling C and CE students together), those who received only the access treatment (Group-A), those who received both oup-A), those who received both access and encouragement treatments (Group-AE) and the existing users, across the baseline survey (November 2015), midline survey (Ma mber 2015), midline survey (May 2016), and endline survey (April 2017). If there are multiple survey questions elicited within a category or subcategory of outcomes, res or subcategory of outcomes, results are shown using a z-score index aggregating all corresponding questions. The z-score index (weighting by the inverse covariance of t g by the inverse covariance of the standardized variables) is computed following 18 / 19su Anderson (2008). Sample is restricted to 1,372 students who have completed the endline Recommended reading I Chen and Yang “The Impact of Media Censorship: or Brave New World?” AER 2019 I To understand how the attitudes, beliefs, and behaviour variables are defined consult Online Appendix D (p A.8 – A.22) 1984 19 / 19