Predicting Cabinet Type Using Banzhaf Power Scores

We offer a new model to predict which type of government will form (minority, minimal winning, surplus) as a function of the structure of the party system. We compare the predictive power of this model, based on Banzhaf power scores using an index first proposed by Caulier and Dumont with the Laver and Benoit classification scheme using party seat shares. Looking at situations where no single party wins a majority of seats in parliament, we propose a three-fold power score classification that generates straightforward cabinet type predictions. Further improvements in predictive accuracy come from a classification which, like Laver and Benoit’s, counts five categories. We first show that the latter allows us to identify more party system formats where we would expect minimal winning coalitions than what alternatives using party shares do. Second, we show that our Banzhaf power scores-based schemes provide more robust results across different real-world cabinet formation contexts. Third, the rate of predictive accuracy of these schemes is shown to be about 10 percentage points higher than Laver and Benoit’s when analysing post-electoral cabinet formations in Europe.