The vote of confidence procedure is one of the most basic institutional rules in the life of parliamentary systems: it allows a Prime Minister to make a contested policy in his own parliamentary party an up-or-down proposal by linking it to the survival of government. Current theoretical models try to explain the actual invocation with the following variables: the proposed policy and the electoral costs a member or the prime minister is faced with (Huber 1996), re-election prospects (Diermeier/Feddersen 1998) and majority rules in different countries (Döring/Hönnige 2006). The empirical proof of the models suffers from two major pitfalls: (1) the number of cases is small, so it is difficult to apply standard empirical methods due to over-determination problems (2) the distribution of preferences in parliament and especially within the majority parties are important – however, it is difficult to measure these preferences. We introduce a stochastic, agent-based simulation based on a choice model (Bonabeau 2002) to compensate these problems. (1) We set up utility functions for actors relying on the theoretical models already developed; (2) we especially have a look on the possible distribution of preferences within and between parties and the different majority rules in various countries. Using the simulation, we analyze the sensitivity of changes in the assumed distribution of preferences as well as the conditions for the success or failure of a vote of confidence. Finally, we summarize the implicit and explicit assumptions included in the simulation and draw a comparison to the complex situation it models.
