Conférencier: Juliana Schulz, HEC Montréal

Résumé / Abstract: In this talk, two new classes of bivariate Poisson models will be introduced. The model formulations rely on the notion of comonotonic and counter-monotonic shocks and allow for both positive and negative correlations. The underlying stochastic representation is based on an intuitive construction with an interpretable parametrization. The proposed bivariate models yield a fully flexible dependence structure, with correlations spanning the full spectrum of possible values. Distributional properties and estimation techniques for the proposed models will be discussed and illustrated through data applications. A multivariate extension to the bivariate comonotonic shock Poisson model will also be introduced.