Non-life loss reserving

On December 31, the actuary of a property and casualty company has to put a number on what it will still pay for claims that have already occurred. Some of them have not been reported, and the rest are reserved by claims handlers who do not all reserve the same way. Non-life reserving stacks layers. Case reserves, set by the claims department on each open file. IBNER, which corrects the shortfall of those reserves on known claims. Pure IBNR, for claims incurred and not yet reported. Future claims handling expenses, ALAE and ULAE. The shared tool is the loss development triangle, paid and incurred, by accident year. Chain ladder projects it as it stands. Bornhuetter-Ferguson and Cape Cod replace the projection with a loss ratio on the young years, where the triangle says almost nothing yet. Mack and the ODP bootstrap put a distribution around the figure. No practitioner keeps a single method: several are run, the gaps are explained and one ultimate is selected per accident year, line by line. Since January 1, 2016, Solvency II also requires a best estimate that is discounted and probability weighted. Article 48 of the directive gives the actuarial function the opinion on its reliability. The next closing is the judge: prior year development, in the published accounts, says whether the estimate was right. The work most often breaks on a silent change in the data, which the methods read as a change in the risk. A new claims system is enough, or an instruction to strengthen case reserves, or faster settlement. Not to be confused with the claims reserve, which is the amount carried on the balance sheet, nor with pricing, which projects the claims of policies yet to be written.