Language of the unheard: police-recorded protests in South Africa, 1997–2013

South Africa remains beset by protest. Notwithstanding an impressive literature, quantifying protests remains problematic; most attempts extrapolate from samples or media-derived data sets. Applying machine learning to the world’s largest publicly available, single-country public-event database – the South African Police Service’s Incident Registration Information System – the article classifies 150,000 events into type and levels of ‘tumult’. The author provides the first holistic picture of all police-reported protest in South Africa over a given period (1997–2013), showing a count increase (partly confirming the ‘rebellion of the poor’ thesis), while more nuanced measures (i.e. protestors per capita) demonstrate a less urban and tumultuous phenomenon than previously theorised.