Interdependent Duration Models in Political Science
27 Pages Posted: 13 Aug 2009 Last revised: 7 Sep 2009
Date Written: 2009
Interdependent duration processes are common in politics and other strategic settings. The time to one type of political event frequently depends on the time to another related event, and the time to an event for one actor often depends on the time to that same event for others. Put in a slightly different way, politics and strategic behavior generate interdependence across durations and duration interdependence across actors. We present a generalized parametric simultaneous equations model that incorporates these two kinds of interdependent duration processes and derive the corresponding full information maximum likelihood (FIML) estimator based on the Weibull distribution. We show with Monte Carlo experiments that our estimator outperforms the alternatives available to those doing applied empirical research. Naive estimators that either ignore the interdependence among duration processes or treat one as exogenous to the others are badly biased when the true relationships are simultaneous ones. Two stage least squares, while consistent, is highly inefficient relative to the FIML. We illustrate these findings in a study of the determinants of government formation duration and survival in European parliamentary democracies and an analysis of the timing of position taking in the US Congress. The interdependence in these durations is substantively important and suggests strategic bargaining over governments in Europe and free-riding behavior among members of Congress.
Keywords: duration, survival, duration analysis, survival analysis, event history, weibull, government formation, government termination, cabinet formation, cabinet termination, position taking, copula, seq, sur, strategic interdependence, interdependent durations
Suggested Citation: Suggested Citation