ERROR 404: THESIS NOT FOUND: A Cross-Country Analysis on Internet Censorship

Abstract

In recent decades, the introduction of the internet, along with its subsequent developments, has changed the cost-benefit analysis around censorship. Thus, I pose the question: presently, what determines whether countries censor the internet? Specifically, I analyze two contested determinants of internet censorship: first, polity/regime type (i.e., whether a country is an autocracy, anocracy, or democracy), and second, whether censorship policies and technologies diffuse from early adopters of internet censorship into other countries. I focus on website blocking and filtering, and distinguish between three facets of website censorship: the number of websites censored, the degree to which websites are censored (i.e., are websites censored completely, or are they only partially blocked?), and the types of websites censored (i.e., social content, internet tools, political content, etc.). Using data from Censored Planet, I look at the differences in website blocking/filtering between countries for two types of censorship technology: Domain Name Service (DNS) blocking/filtering and Deep Packet Inspection (DPI). I quantitatively analyze the number of websites censored and the degree that these websites are censored by running several Ordinary Least Squares regressions (OLS), regressing on polity type, proxies for censorship diffusion (involvement in the Belt and Road Initiative and region), and covariates (number of internet users, GDP per capita, and religiosity). I also qualitatively analyze the types of websites censored by categorizing the content of the blocked websites, and comparing these websites between democratic and autocratic countries. I make two major discoveries: first, I did not find evidence supporting the hypothesis that either polity type or censorship diffusion impacts the aggregate number of websites censored. However, I find that polity type does impact the types of websites censored, and it has a partial impact on the degree to which websites are censored. These findings significantly complicate the existing literature, which have noted a linkage between authoritarianism and internet censorship. Second, I find a difference between the implementation of DPI and DNS censorship, especially between democracies and autocracies; in particular, I find that autocracies tend to employ DPI more than democracies, while the difference in the implementation of DNS censorship is insignificant across democracies and autocracies. This finding adds to the scholarly literature by suggesting that different censorship technologies are used at different levels and for distinct purposes.