Polarization : conceptualization and measurement

How do societies polarize? The literature on polarization in the political science subfield of American politics is rich. However, polarization has been measured at different levels of the society instead of the societal level at large. The reasons for this are mainly methodological. Capitalizing on the latest developments in the field of computational discourse analysis, this paper offers a framework for the operationalization and measurement of polarization over time using high-dimensional text data. By measuring polarization in the media, we showcase our framework. We create a (1) corpus of newspaper articles around the federal budget topic, (2) we develop a three-dimensional measure of budget ideology: social, fiscal and foreign, and (3) we establish 550 human annotated gold labels for modelling purposes. To measure polarization in the media over time, we (1) predict the ideology of a news source and (2) calculate issue constraint among dimensions. Our framework is extendable to various societal levels and the comparative study of polarization across countries. The corpus, labels, and code used are made available on GitHub. The ultimate aim is to help researchers offer a comparative understanding of the dynamics of polarization across time and countries