Something old: In this month edition of the Casualty Actuarial Society’s Actuarial Review, Glenn Meyers introduces a database of company-level Schedule P loss triangles for AY 1987 to 1996.
NAIC made the data available for model-testing. There are triangles for personal auto, commercial auto, medical malpractice, other liability and workers comp. And there are triangles for earned premium, ultimate incurred losses, paid losses, and bulk/IBNR.
Each accident year has 10 years of development data – in other words, the triangle has been squared. The idea is to test a model by using the lower half of the triangle to test the projections based on the upper half of the triangle. I can also see the data driving some college classroom assignments.
Some of the data is at a company level; some is at a group level. The data doesn’t represent the entire industry, though, as the data has been screened for accuracy and consistency. Thus, some companies got dropped.
The data can be downloaded from the CAS web site here.
Something new: AIG filed its group level Annual Statement, and included an extra 28 pages of triangles, SNL reports (sub req).
The giant company insists that it’s not possible to determine whether its reserves are adequate by analyzing the net loss triangles in a typical Schedule P. There are just too many acquisitions and sales, reinsurance treaties and commutations and changes in mix of business. And the rate of change accelerated in 2008 as AIG underwent some structural changes you might have read about. Thus, the extra triangles.
AIG says the additional data don’t make analysis any easier, but do show how convoluted an analysis off Schedule P data is.
The extra triangles appear on this pdf (again, sub. req.), right after Schedule P, Part 7B, Section 6. There are triangles for warranty and excess workers comp (among more company-specific information) if you are looking for those.