Monte Carlo Stochastic Reserving

[Vincent-EDSA]

The current Bootstrapped stochastic reserving method is sufficient for reserve confidence interval estimation however, an additional stochastic model such as Monte Carlo Simulation might be a good way to compare/ validate the process errors as calculated with the Mack and Bootstrapped methods.

[jbogaardt]

Hi @Vincent-EDSA , sorry for the late reply. Can you elaborate on the idea? Is there a public paper/example that we can reference?

[phuang1226]

Hi @jbogaardt , there is a CAS monograph by Glenn Meyers: https://www.casact.org/sites/default/files/2021-02/01-Meyers.PDF

Thanks.

[jbogaardt]

I see, I believe we have an open issue (#30) open for this monograph. My experience with MCMC is limited and I think we might need an extra dependency on pymc to accomplish this.

[Vincent-EDSA]

Hi @jbogaardt, the attached paper might shed some light on the MC-Chain Ladder. At a high level, here we rather sample the age-age development factors - volume-weighted to ensure the correct distribution results.

MonteCarloBasicChainLaddrTheory.pdf