Listening to the wind of change: Predicting protest dynamics in autocracies using Wikipedia page changes

Mass mobilization has a profound impact on regime stability and democratization. Yet, predicting mass mobilization has received far less attention than predicting armed conflict, despite its clear practical and policy relevance. Predicting protests is still uncommon, particularly at subannual levels with wider geographical coverage due to limited granular data. I propose a new data source to both measure levels of contention and predict protest activity: changes to Wikipedia pages. Topics discussed in society, such as human rights, media freedom, elections, censorship, or food security, are also reflected and debated online on relevant Wikipedia pages. Wikipedia data are available in real time, provide global coverage, and are relatively easy to source and process, which makes them exceedingly attractive for near-real time forecasting and early warning systems. I test my argument by predicting both pro- and anti-government protest incidences and numbers at the country-month level. Empirical evidence suggests that Wikipedia page changes offer a promising new approach. The new measures are compared with more traditional indicators to illustrate the added value and show that Wikipedia data help improve our ability to predict protests both on their own and in combination with other models. The findings contribute to research on repression and resistance as well as ongoing efforts to forecast protests.