Issue saliences and distance selection in spatial coalition formation models: an empirical investigation

We examine the impact of three factors on the ability to predict political coalitions of spatial coalition formation models: (1) inclusion of issue saliences, (2) choice of distance function and (3) choice of solution method. Issue saliences quantify the relative importance that parties attribute to different policy dimensions. The second factor is the distance function that is used to calculate the difference between two parties’ positions. The classical application employs the most commonly used Euclidean distance in combination with the gravity center as consensus estimate. This is not the case in the consistent distance application, where the selection can be made between three function: Euclidean, squared Euclidean and rectangular. This choice also determines the consensus estimate. These first two factors are often neglected in existing coalition formation models. The three functions in the consistent distance application and the classical application, each either unweighted or weighted with issue saliences, gives us eight possible combinations to apply in four solution methods. These solution methods are different criteria by which to rank potential coalitions and subsequently determine the (set of) optimal coalition(s). An empirical application including 28 democracies illustrates the impact these factors have on the predictive power of the spatial coalition formation models. The factor with the most important impact on the predictive power is the chosen solution method. Overall, the inclusion of issue saliences or the choice of distance function do not appear to have a significant influence on the predictive power. However, these factors do have significant interaction effects with the solution methods. For some methods, although not all, the issue saliences and distance function can have a significant influence on the model’s performance.