What explains popular support for government monitoring in China?

ABSTRACT

Discussions of China’s recent massive surveillance initiative often present it as evidence of a path to an Orwellian state with omnipresent fear and discontent among its citizens. However, based on a 2018 survey of a nationally representative sample, this paper finds that a large majority of Chinese citizens support various forms of state surveillance. CCTV surveillance receives the highest support (82.2%), followed by e-mail and Internet monitoring (61.1%). Even the most intrusive policy – collecting intelligence on everyone in the country – receives support from more than 53% of citizens. Further, support for surveillance is positively associated with an individual’s preference for social stability, regime satisfaction, and, to a lesser extent, trust in government. Unlike in Western societies, concerns about information exposure and terrorism do not have any significant correlations with citizens’ attitudes toward surveillance in China. These findings might help explain why the Chinese state can expand its surveillance capacity without much open resistance from the public.

KEYWORDS:

• Surveillance
• public opinion
• trust
• security concern
• China

Supplemental data

Supplemental data for this article can be accessed on the publisher’s website.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. “Inside China’s Dystopian Dreams: A.I., Shame and Lots of Cameras,” New York Times, July 8, 2018: https://www.nytimes.com/2018/07/08/business/china-surveillance-technology.html; “Does China’s digital police state have echoes in the West?” Economist, May 31, 2018: https://www.economist.com/leaders/2018/05/31/does-chinas-digital-police-state-have-echoes-in-the-west.

2. We thank one reviewer for this intuitive evidence. Our robustness checks include two measures of a respondent’s survey sensitivity concern to control for self-censorship (Table 3).

3. Elsewhere, Young (2000) proposes three models of the state–civil society relationship: supplementary, complementary, and adversarial.

4. A more recent approach studies civil associations. The third, normative neo-Tocquevillean approach focuses on autonomous associations and their potential contributions to democratization; very few Chinese civil associations fit into this ideal type (Xu, 2004).

5. The level of privacy concern also varies by individual. Unfortunately, our survey did not include a question on privacy concern, preventing us from directly testing its effect at the individual level.

6. Xu, Kostka, and Cao (Forthcoming) collect reports on social credit systems (SCS) from state-controlled media and find that among the 646 pieces collected, only 2.8% are negative; the rest mostly praise the SCS’s trust-building and social-order maintenance functions.

7. Fifty out of 2,855 counties were chosen as primary sampling units (PSUs). Using night-light data, two half-square minutes (HSM) of latitude and longitude were chosen within each PSU. Two spatial square seconds (90 by 90 meters) within each HSM were randomly chosen. Investigators randomly drew 50 dwellings in each HSM and selected respondents using the Kish Grid method.

8. Section C in the online appendix has detailed information on the distribution of respondents by province (Table C1), and the breakdown of answers to the three surveillance policy questions by geographical region (Tables C2–C4) and Hukou type (Tables C5–C7).

9. We have further divided the sample into subgroups to explore whether there are differences in attitudes across different regions and urban and rural residents. Tables C2 to C7 of the online appendix have frequency distributions.

10. We conducted principal component analysis on Trust in administration, Trust in court, and Trust in police and identified a very strong first factor with an eigenvalue of 2.23, explaining 74.4% of the total variation in the three variables. We use this first factor as the Trust in government variable.

11. The factor loadings matrix shows that all seven information sources have strong loadings on the first factor/component.

12. This variable attains a Cronbach’s α of 0.72: Cronbach’s α measure’s the reliability of multiple-question Likert scales in surveys. A value above 0.7 suggests that a set of variables measures well a single, one-dimensional latent aspect of the respondents.

13. A strong first factor was identified, with an eigenvalue of 3.09, and explained 39% of the total variation; the Cronbach’s α is 0.76.

14. The eigenvalue of the first factor is 1.47, 49% variance explained; the Cronbach’s α is 0.47.

15. It has a high one-dimensionality (eigenvalue of 2.51, 63% explained variance) and a Cronbach’s α of 0.80.

16. For a simple interpretation of the result, we coded the original four-point scale into a dichotomous variable: interested in politics (1) or not (0). This coding decision does not affect our results. Those using the four-point scale are identical (see Table B6, online appendix).

17. Questionnaire and coding rules are in Table A1; descriptive and correlation statistics are in Table A2 and Table A3 in the online appendix.

18. We calculated the percentages without using survey weights, which are almost identical to the weighted ones (Figure A1 in the online appendix).

19. Survey weight is calculated based on household size, age, gender, education, and population distribution from the 1% National Population Sample Survey in 2015.

20. The lack of empirical support for social trust might be explained by an alternative causal link between social trust and surveillance support: the more people trust others the less likely they support state surveillance, because they simply trust that other people do not need to be policed.

21. Table B2 of the online appendix reports the effect of a Social Media variable (i.e., Wechat and Weibo) with and without controlling for traditional media (Propaganda exposure) and other categories of new media (i.e., Mobile phones, e-mails, Internet – New media (no social media). Table B3 of the online appendix tests the effect of the New media variable while controlling for traditional media (Propaganda exposure). In all model specifications, neither Social Media nor New media is statistically significant.

22. Also, in our online appendix, we present a table without including the control variables (Table B5).

Additional information

Funding

Shanghai Pujiang Program, Grant/Award Number: [18PJC077]; Ministry of Education of China, Grant/Award Number: [18YJC810010]; National Social Science Fund of China, Grant/Award Number: 19CZZ009.

Notes on contributors

Zheng Su

Zheng Su is an assistant professor in the School of International and Public Affairs, Shanghai Jiao Tong University. His research interests include public opinion, state-society relations, and political economy of development in China. His work has been published in Social Science Research, World Development, and Environmental Politics.

Xu Xu

Xu Xu is an assistant professor in Politics and Public Affairs at Princeton University. He studies the politics of information and political repression in autocracies with a current focus on digital surveillance. He has published in journals including American Journal of Political Science, Journal of Politics, and Journal of Comparative Economics.

Xun Cao

Xun Cao is an associate professor in the Department of Political Science and School of Public Policy, Penn State University. His publications have appeared in journals such as American Journal of Political Science, British Journal of Political Science, Comparative Political Studies, International Organization, International Studies Quarterly, Journal of Politics, and World Politics.