Replication Data for: Genetic Imprints, Party Life Cycles and Organizational Mortality: An Application of State-Space Duration Models

It is a classical argument that how parties are born affects the way they die. However, few studies have theorized and rigorously estimated the impact of central formative features on the risk of organizational death over the course of parties’ entire lifespan. Using a life cycle perspective, we theorize how and when party mortality is shaped by four formative features constituting parties’ heritage: insider status, societal rootedness, ideological novelty, and roots in pre-existing parties. We fit a state-space competing risks model to a new dataset covering 204 party trajectories in 22 consolidated democracies to assess the dynamic influence of these formative features on two distinct types of death: dissolution and merger. Our flexible approach to modelling time dependency outperforms conventional methods and generates novel insights about the time-varying relationship between party heritage and mortality fundamental to whether party renewal is likely to enhance democracies’ representative capacity.

Given the complexity of the models estsimated in the paper, the Markov chain Monte Carlo (MCMC) algorithms were coded in C++ and called from R through Rcpp (Eddelbuettel, 2013), and executed via the University of Exeter High Performance Computing (HPC) cluster (see Section A.3 of the accompanying Online Appendix). While all the models can be estimated using a laptop or desktop computer, the execution time is considerably (around 5 times) shorter using a HPC cluster. For the code to work properly, you have to install an SSH Client (such as Putty or smilar) to connect to an HPC cluster, as well as the following R packages: "Rcpp", "RcppArmadillo", "coda", "ggplot2", "grid", "gridExtra", "gtble", "plyr", "psych", "survival", and "VGAM". When using a HPC cluster, you should load the R (we used version 3.3.1-foss-2016b) and GSL (we used version 2.1-foss-2016b) modules.