Censored and Hurdle Regression Models in TSCS Data

Electoral Support for Extreme Right Parties in 19 West European Democracies

抄録

The article examines why extreme right parties are successful in some elections while not in others by analyzing the elections held in 19 West European democracies from 1970 to 2000. To that end, the article discusses the methodological limitations that arise from applying the Tobit and the decomposed Tobit model (Cragg's model) to the Time-Series and Cross-Sectional (TSCS) data. The Tobit model in the presence of fixed effects is known to be biased and inconsistent due to the so-called ‘incidental parameter problem’. The article first reviews the possible remedies for the incidental parameter problem, including the conditional frequentist, semi-parametric, and Bayesian approaches, proposed in the econometric literature. While discussing the alternative approaches to cope with the incidental parameter problem, the article pursues an approach to decompose the Tobit model and overviews the Cragg's model and its application to the TSCS data. To demonstrate the Tobit and Cragg's model in the TSCS analyses, the analysis of the vote share of extreme right parties in the extant literature is replicated. The re-analysis of the vote share suggests that the Cragg's model can further improve the Tobit model usually applied in the literature.