Much ado about nothing: Zeros and the extreme right

The literature surrounding extreme right parties in Europe has developed dramatically over the past two decades. However, the analysis of electoral success for these parties has produced muddled results, and occasionally even conflicting findings. This article argues this confusion is partially due to a reliance on an inappropriate model choice. Through the use of simulations, a goodness-of-fit exercise, and a prediction exercise based on model cross-validation, I show that the traditional Tobit specification—adopted to deal with electoral results of fringe parties—is theoretically untenable, statistically inferior to alternative models, and practically prone to revealing effects that are unsupported by the underlying data. Rather, the results suggest that best practices should see researchers adopt Cragg or Heckman models for two-stage questions, or consider adopting an analysis applying multiple overimputation if the main question is focused on the determinants of electoral success