Changing Conditions

Study guides: Exam 5, Exam GI-101

Changing Conditions are changes in an insurer's internal operations or in its external environment during the experience period that make historical claims data a biased guide to future development. Every development-based estimate of unpaid claims assumes the past pattern will repeat; a changing condition is whatever breaks that assumption.

\widehat{\text{Ult}}_i = C_{i,k} \times \text{CDF}_k^{\,\text{hist}}

• C_{i,k} is accident year i's reported (or paid) claims at age k; the CDF comes from older years. With reported claims the development technique assumes case adequacy has not changed; with paid claims, that settlement speed has not.
• Internal (operating) changes — the subjects of Friedland's management interview:
  • claims coding (claimant vs. occurrence counts, reopened claims) and claims processing (staffing, systems, case reserving guidelines, fast-track limits, settlement speed);
  • underwriting and policy provisions (eligibility, deductibles, limits, forms);
  • marketing — new programs, target markets or distribution channels, which change who is insured;
  • reinsurance retentions and limits, which change what a net triangle contains;
  • the treatment of recoveries — whether data are net or gross of deductibles and salvage and subrogation.
• External environment: inflation (a calendar-year effect on every open claim) and the legal environment (tort reform, court decisions, litigation). Changes in rate level and mix of business come from both sides; they move claim ratios and development patterns respectively.
• Which estimate each change biases (Friedland's worked scenarios):
  • Rising claim ratios: development techniques still work; the expected claims technique fails unless its ratio is updated; BF and Cape Cod understate, partially.
  • Case strengthening: reported development overstates, reported BF and Cape Cod overstate less; paid methods and expected claims are unaffected.
  • Faster settlement: paid development overstates; reported methods are largely unaffected.
  • Growth in a slower-developing segment: every technique on the combined data understates.
• Responses: detect the change (interview plus diagnostics — average case outstanding, closed-to-reported counts, paid-to-reported ratios); lean on methods the change does not touch; restate the data (Berquist-Sherman); adjust a priori ratios for rate level and trend; segment for mix; restate history to current retentions or benefit levels.

Case Strengthening Through Four Methods A stable auto liability book develops each accident year to \$10{,}000{,}000. Historically 50\% of ultimate is reported and 30\% paid at 12 months (reported CDF 2.000, paid CDF 3.333). In 2025 the claims department strengthened case reserves by 40\%: AY 2025 at 12 months shows paid \$3{,}000{,}000 and case \$2{,}800{,}000 (reported \$5{,}800{,}000). The a priori expected ultimate is \$10{,}000{,}000.

Estimate AY 2025 ultimate by reported and paid development and by reported and paid BF.

Answer

\begin{align*}
\text{Reported dev.} &= \$5{,}800{,}000 \times 2.000 \\
&= \$11{,}600{,}000 \\
\text{Paid dev.} &= \$3{,}000{,}000 / 0.30 \\
&= \$10{,}000{,}000 \\
\text{Reported BF} &= \$5{,}800{,}000 + \$10{,}000{,}000\left(1 - \tfrac{1}{2.000}\right) \\
&= \$10{,}800{,}000 \\
\text{Paid BF} &= \$3{,}000{,}000 + \$10{,}000{,}000\,(1 - 0.30) \\
&= \$10{,}000{,}000
\end{align*}

True IBNR is \$4{,}200{,}000. Reported development books \$5{,}800{,}000 — \$1{,}600{,}000 too much, because the extra \$800{,}000 of case is multiplied by a CDF built on weaker reserves. Reported BF books \$5{,}000{,}000 — too much by the strengthening itself, once: its IBNR comes from the a priori and the unreported proportion, so it takes no credit for the claims now booked in case. The paid methods are untouched. The fix is not to discard the reported data but to restate the historical case reserves to today's adequacy before selecting factors.

Reading a Management Interview A reserving actuary learns: (a) a new claims system from 1/1/2024 opens a file per claimant rather than per accident; (b) the per-occurrence reinsurance retention rose from \$500{,}000 to \$1{,}000{,}000 for AY 2024; (c) a marketing push into commercial auto is growing that segment 30\% a year inside a mostly personal auto book; (d) tort reform caps non-economic damages for accidents from 7/1/2024.

Classify each change and say what it distorts.

Answer

• (a) Internal — claims coding. Reported counts jump on the 2024 diagonal with no change in cost: frequency rises, severity falls, and count-based frequency-severity projections break. Restate counts on one definition or use post-change data only.
• (b) Internal — reinsurance. Net data before 2024 are capped at \$500{,}000, AY 2024 at \$1{,}000{,}000, so historical net factors understate AY 2024's net development. Restate history at the new retention from claim-level data, or project gross and ceded separately.
• (c) Internal — marketing, producing a mix shift. Commercial auto develops more slowly, so combined factors — weighted toward the older mix — understate the recent years. Segment the triangles.
• (d) External — legal. Accidents after the reform should settle lower and possibly faster; the old pattern and claim ratios overstate them. Split accident years at the effective date and adjust expected claim ratios with a law-change estimate.

None of these is visible as a cause in the triangle — only as a symptom. That is why the interview comes before the method selection.