Assessing the Relative Predictive Power of Different Models of Populist Vote Choice

Are we currently living in a populist Zeitgeist? Recent political events suggest that populist and anti-elitists sentiments play a substantial and major role in citizens’ political behaviour. Following a wave of new individual-level measures of populism, out-group perceptions and others, ‘alternative’ explanations of vote choice for such parties beyond spatial and valence based models, have recently received substantial scholarly attention. While these approaches mark important steps forward, both theoretically and empirically, we know very little about the relative predictive power of these new measures relative to classical measures, such as the left-right spectrum, when it comes to (populist) vote choice. Utilising the Advanced Release of the Comparative Study of Electoral Systems (CSES) Module 5 data and additional election studies from other Western liberal democracies and a popular machine learning approach, Random Forests, our findings suggest while populism indeed matters for voter choice, other explanations, particularly political ideology, are more potent predictor. At the same time, populism, nationalism and out-group perceptions add to our ability to predict how citizens vote.