This paper investigates the effects of political polarisation on cabinet stability in Italy. The research, which extends, looking at a Mediterranean country, the relevant literature that has so far focused on Anglo-Saxon countries, introduces a new estimator of polarisation based on machine learning methods of text analysis. Using the unsupervised algorithm Wordfish (Slapin and Proksch 2008), political actors are scaled along the ideological spectrum, through the analysis of speeches held during parliamentary debates. The relative dis-tances between them are computed, to construct an index of political polarisation meant to capture the level of political conflict between the Government and the Parliament. The paper employs data on the Italian govern-ments that took office between 1994 and 2019, to estimate the correlation between polarisation, as measured by this index, and the duration of a government’s mandate. The paper finds evidence of a negative correlation between polarisation and government survival, statistically significant at the 10% significance level.
