The Socio-Political Meaning of Observational Audio Data

While political scientists are increasingly analysing text data, they are not yet paying attention to the spoken voice. In this paper, I fill this lack. First, building on socio-phonetics, I argue that focusing on pitch,k intonation and intensity of voice signals captures a reduced form of what we typically refer as ''accents'. Second, I show how to convert those concepts into quantitative data and how to control speech characteristics by their corresponding text. Third, I build on variation in political units to provide evidence that vocal expression across countries reflects long-standing institutional characteristics such as gender inequality and freedom of speech using data from about 5 million utterances in 75 different languages. The main implication is that institutions affect vocal delivery in a consistent manner across multiple dimensions of speech characteristics.