A Bayesian Analysis of Collective Action and Internet Shutdowns in India

Abstract

Since 2011, internet shutdowns have steadily become an increasingly popular form of digital repression, especially in India - which accounted for more than 50% of global recorded shutdowns from 2016 to 2019. Common shutdown justifications include ensuring public safety' in order to curb the prevalence of collective action in the form of protests and riots. This paper examines the correlation between internet shutdowns and a range of predictors, identifying riots as the main predictor of a shutdown. We focus on shutdowns throughout India between 2016 and 2019 with particular attention to Jammu and Kashmir. Primarily using data from the NGO Access Now and the Integrated Conflict Early Warning System, we apply Bayesian inference via generalised linear modelling implemented using the Stan probabilistic programming language, to estimate correlates of shutdown behaviour. We first examine how the prevalence of collective action may impact the probability of observing a shutdown; and second how the length of a shutdown impacts subsequent collective action. Our main finding is that riots seem to be the key predictor of a shutdown with increased protests and riots increasing the odds of observing a shutdown the same day by 7% with a 95% credible interval of 0.01-0.13 and 15% with a 95% credible interval of 0.03-0.26 respectively. As a predictor, however, the duration of an internet shutdown only has a marginal negative effect on the occurrence of riots at -8% per subsequent shutdown day with a 95% credible interval of -0.16 to -0.002.