An increasing number of events across the world have been a warning for democracy stability in established democratic countries. Events such as Hungary’s prime minister Viktor Orbán publicity doubting that liberal democracies could remain globally competitive, and the increasing voting share of anti-establishment parties in European member states are consequences of the political polarization phenomenon which endangers our democracy. To understand if we are becoming more polarized, literature has been focusing on measuring political polarization through surveys and voting data, without consistent evidence for any trend. Although the theoretical definition of political polarization has found stability in the literature, the different forms of measuring it have not. The measurement of political polarization needs to be more robust and extended to mass society besides elite society, enabling a comparison between the two, and within the real life and the digital. This dissertation answers this need, measuring political polarization, using text-as-data and computational social science methods, in an effective way independent of manual tasks, language, survey or pooling, polarization’s actors, and environments. It uses an empirical framework applied to parliamentary discourses and Twitter data to create a Discourse Polarization Index (DPI) which enables the assessment of the evolution of political polarization in discourse, considering its state and process. Portugal is used as use case, showing an increase in political polarization from 2015 to 2021, for the elite and the mass society, with similar behaviour between the two groups. A semantic validation is done, and research future steps are given.

If you use this codebase, or otherwise found my work valuable, please cite:

@article{pereira2022msc,
  title={Discourse polarization index: Analysis of top-down and ground-up political discourses in Portugal},
  author={Pereira, C. T.},
  journal={Repositório Iscte. http://hdl.handle.net/10071/26231},
  year={2022}
}