To assess the joint influence of several risk factors on the DNBP, we estimate the expected DNBP for both high- and low-risk claimants (see Table 6).
We use the simulation approach outlined in Appendix B to estimate the predictive distribution of the total DNBP on the basis of the posterior frailty distribution for each of these claimants.
However, the DNBP is subject to model uncertainty, process uncertainty, and parameter uncertainty (England and Verrall, 2002).
The first panel of Table 7 displays the predictive distribution of the total DNBP during the first year (f = 0, .
The realized DNBP falls just within the 95 percent confidence interval for process and parameter uncertainty.
These two alternative models substantially underestimate the realized DNBP and the underestimation becomes worse for the best estimates further away in time.
The expected DNBP expressed as a percentage of the realized DNBP is displayed in second half of Table 7.
The comparison with the standard Markov model results in the following approach to calculating the expected DNBP.
To incorporate the latest information about the outstanding claim duration in the estimated liabilities, the insurance company can recalculate the DNBP periodically (e.