An insurer books fast-closing property claims and slow-dragging liability claims into one reserve triangle. Property settles inside a year; liability runs for a decade. Blended, the triangle describes neither, and the unpaid estimate inherits the error.
Reserving methods assume the claims in a triangle share one behavior. Chain ladder applies a single 12-to-24-month factor to every accident year in the group. That factor is only valid if the claims actually develop alike.
Group claims that report and settle at different speeds, and no single factor is right. A fast peril is nearly paid at 24 months. A slow casualty coverage has barely begun. Averaging their development produces a factor too high for the fast claims and too low for the slow ones.
KEY: Actuaries judge whether claims belong in one group by several characteristics: consistency of the coverage triggered, the volume of claim counts, the reporting pattern, the ability to set a reliable case outstanding estimate over the claim's life, the settlement pattern, the likelihood the claim reopens once settled, and average severity.
Common mistakes
- Blending fast and slow claims into one triangle. A property-plus-liability factor of 1.50 fits neither the 1.20 nor the 1.80 pattern and misprices any shifting mix.
- Treating homogeneity as the enemy of credibility. Homogeneity raises credibility; the real cost of splitting is lost volume, so name the trade correctly on an exam.
- Assuming a stable combined factor means no bias. A steady blended number can still drift as the long-tail share climbs; only the split triangles reveal it.
Bottom line
- A homogeneous segment holds claims with similar development, frequency, and severity behavior, so one set of factors fits them all.
- Credibility is the predictive value you assign to a body of data; both more volume and greater homogeneity raise it.
- The tension is real but narrow: splitting for homogeneity keeps behavior alike while draining the volume each cell needs to stay stable.
- Combining heterogeneous claims lets a shifting mix distort aggregate development factors and mask real trends.
Exam shortcut
When a blended triangle looks stable but reserves run deficient, suspect a shifting claim mix and split into homogeneous segments before trusting any factor. Read granularity as a trade between homogeneity and volume, not homogeneity and credibility: split for uniformity, stop before volume runs out, and when forced past that line, credibility-weight with Z = n/(n+k) against a broader complement.
The full lesson (about 1,894 words, 13 min read) adds 2 worked examples, all 7 common mistakes, a self-check, free in the app.
Learning objectives
- B3
Browse all free Exam 5 lessons or jump into free Exam 5 practice questions.