Will the real populists please stand up? A machine learning index of party populism

Abstract

The existing literature on populism has seen numerous attempts to empirically quantify this oftentimes ambiguous concept. Despite notable advances, continuous populism measures with a clear theoretical background and a considerable coverage are still hard to come by. This paper proposes a novel approach to measure party populism by combining several different expert-surveys via supervised machine learning techniques. Employing the random forest regression algorithm, we create two novel populism indicators, which are based on the discursive and the ideational approach, respectively. The resulting measures capture party-level populism on a continous 0-10 scale, covering 1920 parties in 163 countries from 1970 to 2019. The paper further provides several validity tests, highlighting the benefits of our indicators for empirical social science research on populism.

Keywords: Populism, Populism indicator, Machine Learning, Parties