The five components
StandRe cuts non-life insurance risk along two axes — attritional versus individual events, and prior years versus the current year:
| Attritional | Individual | Nat cat | |
|---|---|---|---|
| Prior years (reserve risk) | AER | IE2 | — |
| Current year (premium risk) | AEP | IE1 | NE |
Prior-year risk is AER + IE2, current-year risk is AEP + IE1 + NE, and the AG component aggregates them into the distribution the capital is read off.
What ships
- Parameter Segment — a row of the AER or AEP parameterisation: the estimation inputs, the estimated coefficient of variation and the selected one, and the payment pattern that discounts it.
- Scenario — the FINMA scenario catalogue (sixteen damage events, eight other events) plus your own; together they trace the exceedance frequency curve IE1 and IE2 are calibrated to.
- StandRe Model — the model structure and the run: log-normal moment aggregation for AE, Poisson / generalized-Pareto calibration for IE, Panjer recursion for the aggregates, convolution for AG.
- SST Filing — the reporting face: target capital, the market value margin, the SST ratio and the supervisory band it falls into.
It reuses the model, it does not copy it
Line of business, region, currency, peril and the yield curves are the existing Reinsurance dimensions, referenced by node path. SST adds no dimension of its own — the StandRe LOB and region taxonomies ship as members of the dimensions the rest of the suite already uses, and the FINMA-prescribed curves ship as ordinary Reinsurance/YieldCurve nodes.
Deterministic, therefore testable
There is no Monte Carlo anywhere: the attritional components are closed-form log-normals, the individual components are compound Poissons by Panjer recursion, and the aggregation is a discrete convolution. Every figure reproduces to the last digit, which is what lets the Tests areas assert on them.