Stochastic Cash Flows and Backward Semi-Markov Reward Processes

Raimondo Manca
University La Sapienza, Rome

A full treatment of continuous time homogeneous and non-homogeneous backward semi-Markov reward processes will be presented, as far as the authors know, for the first time in the continuous time continuous state non-homogeneous case. In the continuous time semi-Markov process environment, the distribution function that rules the transitions between the states of the studied system can be of any type and not only exponential. This fact is an important generalization as regards the Markov environment. The introduction of backward time makes it possible to consider the instant in which the system entered a state, even if it entered before the time under consideration.

Rewards permit the introduction of a financial environment into the model. Considering all these properties any stochastic cash flow can, in the authors' opinion, be naturally modeled by means of semi-Markov reward processes. Furthermore, the backward case offers the possibility of considering the duration of an event that began before the time in which the system is observed, and this fact can be very useful in the evaluation of some insurance contracts.