Until today, standard econometric software packages are of limited help for the empirical analysis of voting behavior in multi party systems. We have profited from the discrete choice-toolkit provided by LIMDEP 7.0 in previous studies when applying nested multinomial logit and multinomial probit models to the decision of voters in multi-party systems. In order to analyze the interdependence structure of attitudes and voting behavior in the framework of a non-linear structural equations model, we cannot rely on standard packages. Instead, we apply simulated maximum likelihood estimation techniques that have first been suggested by Börsch-Supan, Euwals and Eymann (1999) and have been implemented in FORTRAN code by Börsch-Supan, Eymann, and Hajivassiliou.
