The paid-versus-incurred gap is where a contingency forecast quietly goes wrong even when the agency isn't trying to inflate it. Paid losses are observable — the claim dollars actually disbursed — but incomplete, because claims keep developing after they're filed. Incurred losses add the reserves for what's expected to be paid, which is the number the carrier uses to test the contingency loss ratio. A forecast on paid data looks cleaner than reality, and the cleaner it looks, the more contingency a buyer is tempted to pay for.
§ 01 · The paid-to-incurred gapWhat "incurred" adds.
| Measure | What it captures |
|---|---|
| Paid losses | Claim dollars actually disbursed — observable but incomplete |
| Incurred losses | Paid + case reserves + a provision for unreported claims |
| Mid-year gap | Incurred runs 20%–40% above paid on active books |
Incurred losses are paid losses plus case reserves plus a provision for claims incurred but not yet reported — the mathematical definition the carrier uses for the contingency loss ratio. The gap between paid and incurred is large mid-year: incurred frequently runs 20%–40% above paid at any point inside a contract year, and the gap closes only as claims mature, typically 12–18 months after year-end. That's why a forecast built on paid data is structurally optimistic — it's measuring an incomplete number and treating it as final. The premium-side half of this same miscalibration (written versus earned) is covered in written vs. earned premium; the two together are the full reconciliation framework — earned premium denominator, incurred loss numerator, plus a development factor.
§ 02 · The threshold-crossing riskFrom 42% to 51%.
A book projecting a 42% loss ratio on paid data at mid-year can settle at 51% on incurred at year-end — meaningful enough to cross a 50% contingency threshold and zero the entire payout under a cliff structure. And even incurred losses develop adversely as claims mature, so apply a conservative 3–5 point development factor to mid-year incurred to forecast what the carrier will actually report. The honest number is the incurred number, developed.
The threshold-crossing risk is the practical danger of forecasting on paid. Because contingency thresholds behave like cliffs, the difference between a 42% paid forecast and a 51% incurred settlement isn't a modeling nicety — it's the difference between earning contingency and earning nothing, when the 50% floor sits between them. The discipline is to add a development factor: even incurred losses tend to develop adversely as claims mature, so applying a 3–5 point conservative factor to mid-year incurred produces a forecast that better matches what the carrier will report at year-end. A buyer modeling a target's contingency on a developed-incurred basis is modeling the number the carrier will actually pay; one modeling on paid is modeling a number that's already obsolete.
§ 03 · The reconciliation testAnd who has the data.
The most efficient diligence test for this is a reconciliation: compare three years of the target's mid-year (months six to nine) contingency forecasts against the actual year-end payouts. Spreads of 20%+ signal paid-loss forecasting; spreads of 5% or less signal incurred-loss forecasting with competent development assumptions. The spread, in other words, grades the forecasting discipline directly. The data-access reality shapes who does this well: larger agencies with strong carrier relationships receive monthly loss runs with both paid and incurred columns, while smaller agencies must specifically request the incurred number during forecast cycles — the data is available, but the discipline of asking for it often isn't. One more reconciliation wrinkle: true-ups and prior-year-paid contingency can inflate a given year's reported contingency on a non-repeatable basis, so a historical contingency-as-percent-of-revenue that looks unusually high against benchmarks deserves a forensic look at that year's specific true-up contributions before it's capitalized.
§ 04 · Seller reset and long-tail exposureThe defensible number.
For a seller, the paid-versus-incurred gap is a pre-listing opportunity rather than a trap: a one-time recalculation of the prior three years on an incurred basis — using carrier-provided reserve data, typically a few hours per contract year — produces a substantially more defensible contingency narrative going into diligence. A seller who reports on incurred from the start preserves the integrity of their reported numbers; one who reports on paid invites buyer-side discounting that may exceed the actual gap, because a buyer who can't trust the basis will haircut to protect themselves. Long-tail lines amplify the whole dynamic — bodily injury, litigation-exposed coverage, and professional liability carry larger paid-to-incurred gaps and longer maturation timelines than books concentrated in personal auto or commercial property, so a book heavy in those lines needs the incurred-basis discipline most. Read on the carrier's basis with a development factor, reconciled against actuals, the contingency number stops being a source of dispute and becomes one a buyer can capitalize. The clawback exposure that sits on top of this loss-ratio uncertainty is in contingency clawback economics.
◆
Terminology on this shelf
- Incurred losses
- Paid plus case reserves plus a provision for unreported claims — the carrier's basis.
- Paid-to-incurred gap
- Incurred runs 20%–40% above paid mid-year, closing 12–18 months after year-end.
- Development factor
- A 3–5 point conservative add to mid-year incurred, because incurred develops adversely too.
- Reconciliation test
- Mid-year forecast vs. actual year-end payout — 20%+ spread signals paid forecasting.
- True-up inflation
- Prior-year true-ups can lift a given year's reported contingency on a non-repeatable basis.
- Long-tail amplification
- Litigation-exposed lines carry larger gaps and longer maturation than short-tail lines.