Title: A copula duration model for multiple-states-multiple-spells
Authors: Shuolin Shi - Copenhagen Business School (Denmark) [presenting]
Ralf Wilke - Copenhagen Business School (Denmark)
Ming Sum Simon Lo - City University of Hong Kong (Hong Kong)
Abstract: Copula graphic estimator for competing risks duration model utilizes the Archimedean copula to model the dependence structure of risk-specific durations. The nested Archimedean copula function permits for different dependence structures between risks, spells, and states, and it does not put any restriction on the cumulative incidence function. The dependence structure between risks is not identified and can only be assumed, yet the dependence structures between spells and states can be identified and estimated. So, the copula graphic estimator for multiple-states-multiple-spells competing risks duration model is flexible and can serve as a tool to assess the relevance of assumptions on the dependence structure of risk-specific durations. The estimation procedure is worked out and tested with both simulations and empirical Danish administrative data. The estimator is still consistent if we ignore the dependence structure, yet a loss of efficiency is expected.