Loss Development

Study guides: Exam 5, Exam GI-301

Loss Development is the change in the reported or paid value of a cohort of claims as it matures — the difference between losses valued at an early age and their eventual ultimate settled value. Because a ratemaking or reserving analysis must compare cohorts at a common (ultimate) basis, immature Accident Years must be developed before they can be used.

$\text{Ultimate Losses} = \text{Losses at age } n \times \text{CDF}_{n \to \text{ult}}$

$\text{CDF}_{n \to \text{ult}} = \prod_{k \ge n} f_{k \to k+1} \times \text{Tail}$

Development has two drivers: pure IBNR (claims that occurred but have not yet been reported) and IBNER (incurred but not enough reported — development on claims already known, as case reserves are re-estimated toward settlement). Reported triangles capture both; paid triangles also capture the lag between a case reserve being set and the cheque clearing.
Reported (incurred) development is normally faster than paid development, because claims are reported sooner than they are paid, so cumulative reported CDFs are usually smaller than paid ones at the same age. Comparing the two is a standard diagnostic: a change in the paid-to-reported ratio signals a change in settlement patterns or in case adequacy — a rising ratio is consistent with faster settlement or weaker case reserves, a falling one with slower settlement or stronger case reserves — and the actuary must investigate which (see Case Adequacy, Settlement Rate).
Development can be downward. Lines dominated by salvage/subrogation recoveries or by conservative initial case reserving produce age-to-age factors below $1.000$ ; capping factors at $1.000$ because "losses only go up" is a classic error.
Long-tail lines (liability, workers compensation) have large CDFs at early maturities and need a Tail Factor beyond the observed triangle; short-tail lines (property, auto physical damage) are close to ultimate within a year or two.
The mechanics live in the Development Triangle → Age to Age Factor → Cumulative Development Factor chain; the Chain Ladder Method applies them directly, while the Bornhuetter-Ferguson Method uses the complement $(1 - 1/\text{CDF})$ to weight an a priori expectation.
In ratemaking, developing losses to ultimate is a distinct step from trending: development moves an immature year to its own ultimate value, while trend restates it to the cost level of the future policy period. Applying one without the other under-states the indication; applying both is not a double count — Werner & Modlin call that belief the overlap fallacy.

Developing an Immature Accident Year Reported losses for AY 2024 at $12$ months are $\$4{,}200{,}000$ . Selected age-to-age factors are $12\text{–}24: 1.650$ , $24\text{–}36: 1.220$ , $36\text{–}48: 1.080$ , $48\text{–}60: 1.030$ , with a tail factor of $1.015$ beyond $60$ months.

Estimate ultimate losses and the implied IBNR.

Answer Chain the factors to get the cumulative development factor from age $12$ to ultimate:

$\text{CDF}_{12 \to \text{ult}} = 1.650 \times 1.220 \times 1.080 \times 1.030 \times 1.015$

$= 2.2729$

Apply it to reported losses:

$\text{Ultimate} = \$4{,}200{,}000 \times 2.2729 = \$9{,}546{,}180$

IBNR is the portion of ultimate not yet reported:

$\text{IBNR} = \$9{,}546{,}180 - \$4{,}200{,}000 = \$5{,}346{,}180$

Leverage at early maturities A CDF of $2.27$ means more than half the ultimate estimate is unreported. A $5\%$ error in the $12\text{–}24$ factor moves ultimate by roughly $\$480{,}000$ — which is why actuaries lean on the Bornhuetter-Ferguson Method or Expected Loss Method at the immature end of the triangle rather than on the Chain Ladder Method alone.

Reported vs. Paid Development as a Diagnostic For AY 2022 at $36$ months, paid losses are $\$6{,}300{,}000$ and reported losses are $\$9{,}000{,}000$ . At $36$ months the historical average paid-to-reported ratio for this line is $0.58$ .

What does the comparison suggest?

Answer $\text{Paid-to-reported ratio} = \frac{\$6{,}300{,}000}{\$9{,}000{,}000} = 0.70$

At $0.70$ against a $0.58$ benchmark, AY 2022 has paid out a materially larger share of its reported losses than history would suggest. Two readings are consistent with this:

Faster settlement — claims are closing sooner, so historical paid development factors (built on slower settlement) will over-develop this year.

Weaker case reserving — case reserves are lower relative to eventual payments, so historical reported factors will under-develop this year.

Either way the historical factors are biased for this cohort. The standard responses are the Berquist-Sherman Method (restate the triangle for the changed settlement rate or case adequacy) or a shift toward the method less sensitive to the change.