Voter choice in multi-party democracies: A test of competing theories and models

We contrast social-structural theories of voting behavior with spatial theories of voting behavior to explain voter choice in the Netherlands and Great Britain. We hypothesize that voting behavior is best explained by the spatial theory of voting. Markov chain Monte Carlo (MCMC) simulation is used to estimate multinomial probit (MNP) and multinomial logit (MNL) models of voter choice, for which we calculate Bayes factors for the purpose of model comparison. We find that the joint social-structural/spatial model is the best explanatory model in the Netherlands. Our results indicate that the MNP model outperforms the MNL model in our Dutch sample. In Great Britain, on the other hand, a purely spatial model is the best explanatory model, and our MNL model outperforms our MNP model. These results suggest the question of whether to employ MNL or MNP depends crucially on the data at hand.