Monte Carlo Methods for Non-Life Actuaries

Non-life actuaries have read Glasserman’s “Monte Carlo Methods in Financial Engineering” for two decades and learned variance reduction, low-discrepancy sequences, and the MCSE-based reporting discipline that the book established as the field’s professional standard. They have read Mack’s 1993 ASTIN paper through the Wüthrich–Merz monograph and absorbed the analytic prediction error for the chain ladder. They have read Klugman, Panjer, and Willmot’s “Loss Models” and absorbed the compound-distribution machinery. The three traditions sit on a non-life reserving actuary’s desk, and each is excellent on its own ground.

Each is also incomplete for the question a non-life internal-model team is actually asked at year-end. Glasserman is calibrated against derivatives books, not motor TPL triangles. Mack is the analytic core but is silent on the catastrophe overlay, the copula coupling, the reinsurance treaty layer, and the capital allocation. Klugman covers the compound aggregate but stops short of the Solvency II one-year horizon and the regulator-facing audit defense. The actuary who wants to ship a defensible internal-model submission must currently stitch the three traditions together from scratch, inventing the methodology disclosure and the diagnostic suite anew at every annual recalibration.

This book is that stitch, written explicitly for the non-life setting. It covers the full pipeline from per-claim severity to the firm-level economic capital model, with the Solvency II one-year horizon as the binding regulatory framework throughout. Every numerical result in the chapters is produced by R code printed in Appendix A and reproducible by the reader; every methodology choice is named, defended, and routed against the regulator’s likely audit response. The audience is the practitioner who already knows variance reduction and MCSE — and who wants the non-life specialisation the existing literature does not provide.