Reserving Methods9 min read

Choosing a Stochastic Reserving Method: A Decision Guide

By Jonas Mohamed Osman Abdelghafour

Published 16 September 2026

A structured comparison of bootstrap, Mack, Bayesian and individual-claim approaches, with the data conditions each one requires.

Executive answer

There is no universally best stochastic reserving method. Selection should follow from the data available, the tail length of the class, the materiality of the reserve, the decision the number supports, and the firm's ability to validate and maintain the method.

Matching method to condition

Stable, high-volume triangles suit the bootstrap and Mack. Sparse or immature classes benefit from Bayesian methods that formally incorporate prior expectation and credibility. Classes dominated by a small number of large claims are better handled by frequency-severity or individual-claim models. Classes undergoing structural change need explicit adjustment before any stochastic method is applied.

The cost of complexity

Every additional layer of sophistication adds implementation risk, validation burden and key-person dependency. A simple method that the reserving committee understands and challenges usually outperforms a sophisticated method that nobody can interrogate. Complexity should be justified by a demonstrable improvement in decision quality.

Governance considerations

Method selection should be documented as a decision with reasons, reviewed periodically, and revisited when the portfolio changes materially. An inherited method that has never been re-justified is a standing model-risk finding.

Conclusion

Method selection is a governance decision supported by statistics, not a statistical decision made in isolation. Jonas Mohamed Osman Abdelghafour approaches it as part of the firm's overall model risk framework.

Primary sources

stochastic reservingmethod selectionmodel risk