Ratemaking Data Organization

Study guide: Exam 5

Ratemaking Data Organization is the choice of how premium, loss and exposure transactions are grouped into cohorts for a rate analysis. Werner & Modlin set three objectives for it: accurately match losses and premium for the policy, use the most recent data available, and minimize the cost of data collection and retrieval. The four common methods — calendar, accident, policy and report year — each meet them to a different degree.

\text{Experience Ratio} = \frac{\text{Losses of a cohort, developed and trended}}{\text{Premium of that cohort, on level}}

• Accident Year — losses by occurrence date against premium earned in the same year. The accepted norm for P&C insurers in the U.S. and Canada (Friedland): a better match than calendar year, with premium fixed at year end but losses that still need development.
• Policy Year — every transaction on policies incepting in the year. The best match of premium to loss, and particularly useful when underwriting or pricing changes occur (Friedland's example: a shift to large-deductible policies), but its exposures are not fully earned until 24 months after it starts, so its data takes longer to develop.
• Calendar Year — everything booked in the year. Available quickly and already collected for financial reporting, but its losses include payments and reserve changes on claims from policies issued years ago, a mismatch with the premium earned in the year. Werner & Modlin say it may be most appropriate for lines whose losses are reported and settled relatively quickly, such as homeowners.
• Report Year — claims by report date. Typically used for commercial lines written on a claims-made basis, such as medical malpractice.
• Close Year — claims by settlement date.
• In-Force premium is the full-term premium of every policy in effect on a date — the best estimate of the current mix of business, and often used to measure the impact of a rate change on existing customers.

Two further choices cut across the time basis:

• Gross, ceded or net. Data may be organized before reinsurance, as the reinsurer's share alone, or net of cessions. The indication must be internally consistent — net losses against net premium.
• Granularity. Coarser cohorts are more credible; finer cohorts are more homogeneous and more responsive. Accident quarters are used where volume permits, and the same tension governs the line and class segmentation of the data.

Werner also distinguishes the internal sources (policy and claim databases, accounting records) from external ones — data calls and statistical plans, aggregated industry data, competitors' rate filings and third-party data — needed when pricing a new line and often helpful to supplement internal data.

One Transaction Set, Five Cohorts A 12-month policy is effective 7/1/2024 for \$1{,}200. A loss occurs 11/15/2024, is reported 12/20/2024, is reserved at \$8{,}000, and closes 3/15/2025 for a \$9{,}500 payment.

Assign every element to its cohort under each basis.

Answer

Basis	Premium	Loss
Policy Year	PY 2024: \$1{,}200	PY 2024: \$9{,}500
Accident Year	CY 2024: \$600; CY 2025: \$600	AY 2024: \$9{,}500
Calendar Year	CY 2024: \$600; CY 2025: \$600 earned	CY 2024: \$8{,}000 reported; CY 2025: \$1{,}500
Report Year	—	RY 2024
Close Year	—	Close year 2025

Only the policy year keeps the whole \$1{,}200 and the whole \$9{,}500 together. The calendar year splits both, and splits them differently — the source of the mismatch that makes CY unsuitable for pricing.

Choosing a Basis for a Mid-Year Coverage Change An insurer intends to file a rate change effective 1/1/2026 for a homeowners book. Two facts complicate the choice of experience period: the insurer doubled its minimum deductible on all policies written on or after 7/1/2024, and it needs the filing prepared in early 2025.

What data organization should the analysis use?

Answer The two facts pull in opposite directions.

• The deductible change attaches to the policy, so Policy Year would isolate it cleanly — PY 2025 is all-new-deductible business.
• But in early 2025, PY 2024 is only half-earned and PY 2025 does not exist. Waiting for a credible PY cohort would delay the filing by two years.

The practical answer is accident year with an adjustment: use AY experience for volume and responsiveness, and restate historical losses to the current deductible level using a loss elimination ratio — the same "adjust the data to current conditions" logic that on-levelling applies to premium and trend applies to cost level.