Modeling Parameter Uncertainty in Cash Flow Projections

In order to be complete dynamic financial analysis (DFA) models should deal with both the amount and timing of future loss and loss adjustment expense payments. Even more than asset cash flows, these future payments are very uncertain. However, even with this uncertainty, one would expect to see payments that are somewhat stable from year to year. This paper presents an approach that can deal with this seeming contradiction. By separating total uncertainty in future cash flows into its parameter and process components we present a method to model future liability cash flows that maintains the desired total uncertainty characteristics. However, it will also result in specific payment flow “paths” having less variation from year to year than would a completely random sample from the expected total payout would indicate. There is also a companion of this paper, titled “Estimating Uncertainty in Cash Flow Projections” that considers the problem of estimating the distributions, including separate consideration of process and parameter uncertainty.