Introducing Multiple-Membership Survival Analysis: An Application to Turnover

Multilevel paradigms have permeated organizational research, greatly advancing our understanding of organizational behavior and management decisions. Despite advancements in multilevel modeling, taking into account complex hierarchical structures in data remains an intensive venture, especially when dealing with non-linear models. In this study, we build on recent methodological advancements to implement a multilevel survival model taking the multiple-membership structure of the data into account. Our modeling approach also allows to explore non-linearity of the covariate effects using a semi-parametric regression technique. We illustrate the application of the model on a large educational database containing information of 1,642 novice secondary education teachers for a time span of the first five years in the profession in order to predict early career turnover. Additionally, we demonstrate the utility of the model for alternative research questions where time-to-event is of interest.