Pure-jump semimartingales (Q1983627)

This paper introduces an extension of the stochastic integral with respect to integer-valued random measures by using a class of sigma-localized predictable functions. This integral is used to represent sigma-locally finite variation pure-jump processes in a way that parallels the representation of finite variation pure-jump processes as finite variation jump integrals. It is also used to represent semimartingales as the sum of a quasi-left-continuous process and a sigma-locally finite finite variation pure-jump process, similarly to the classical decomposition of local martingales into the sum of a quasi-left-continuous process and a process having predictable jump times. As another application, a hierarchical classification of pure-jump semimartingales is presented.