Measuring Uncertainty in Increased Limits Factors, A Bayesian Approach

The usual methodology for setting increased limits factors is to extrapolate relationships observed in past data to expected relationships for the future. This is barely adequate for predicting expected values and provides little guidance for measuring the error in that prediction. This paper brings two new elements to the analysis. the first is the availability of data in far greater detail than normally encountered and the second is the use of Bayesian analysis to incorporate the various sources of error. This is done by developing a state-space model for the observations in a particular accident year-development year cell and then projected this model into the future. A demonstration of this technique is provided with an example from products liability insurance. Reinsurance Research - Loss Distributions, Size of