Regression based models have two major drawbacks, if they are used for the explanation of social/political behavior: they are often more deterministic than the fuzzy reality and they generally neglect the intentionality of the goal-oriented behavior of the modeled actors. As a possible solution, the author proposes to consider evolutionary models based on repeated games with a variable, situation dependent payoff-matrix. Since the standard assumption of mathematical game theory is the maximization of the respective payoffs, intentionality as a major explanatory factor of social/political behavior is given due consideration. Moreover, the outcome of many games such as the repeated chicken game or the prisoners‘ dilemma is indeterminate, especially for actors with bounded rationality. Hence, causal analysis is in this case not directly targeted at the explanation of the indeterminate behavior of the concerned actors but rather focuses on the structure of the payoff-matrix, which may be influenced by the power distribution of interest groups, institutional settings, etc. Of special interest within this frame of analysis are those „tipping-points“ in the payoff-structure, where an indeterminate game turns into a game dominated by one of the two players such that the situation becomes unstable and subsequently disappears. Hence the described game theoretic models are also very useful for the causal explanation of the non-presence of political situations. As an illustrative case, this paper analyzes the size of the welfare state in different European countries. For this purpose, it studies the relationship between the power of different antagonistic interest groups and the size of the welfare state. Moreover it demonstrates, how lacunas in the data structure can be used in order to estimate by iteratively re-weighted regression the „tipping points“ in the payoff-matrix, where the system becomes unstable and further growth or decline of the welfare state is not possible.
