In this article a simpte model of Belgian government coalition formation is proposed. Its main advantages over the traditional game theoretical approach are, first, its dynamic nature (explicity linking the outcome of a new 'game' to the characteristics of the previous coalition) and, second, the fact that only one coalition is predicted per 'game '. As a result of the latter characteristic, the integration of this coalition formation submodel in a larger (econometric) politico-economic model should be far less complicated. Moreover, statistical tests indicate that the proposed model works at least as good as the traditional theories, despite its parsimonious nature.