What is IBNR in reinsurance?

IBNR is incurred but not reported: an estimate of losses that have happened, on the contract's basis, but are not yet in the claims file as reported claims. It is a reserve, not a fact on a bordereau row. IBNER — reported but not enough reserved — is a different bucket. Mixing them in one cell called "IBNR" is how two conversations become one number nobody can defend.

The estimate needs an as-at date, a triangle or a method note, and claims bordereaux that can actually be used. A model sentence that reserves look adequate is not IBNR. Cedent reporting and reinsurer operations both consume the same dirty input: undated loss runs.

As-at, then method

Fictional walkthrough: ACME Construction Ltd, acme.example. The Q2 2026 claims bordereau, as-at 30 June 2026, shows ACME paid USD 1,100,000 and case outstanding USD 400,000. Reported incurred is USD 1,500,000. An actuarial note, as-at the same date, holds IBNR of USD 250,000 on the class that includes ACME, not necessarily all on ACME as a named insured. Ultimate on the class slice you assigned is USD 1,750,000 only if you have a cited method for that assignment. If the loss run has no as-at date, you cannot put ACME's USD 1,500,000 on a triangle. That is a gap. Do not chart it as current.

The bordereaux automation guide is the claims-file side: event date, report date, paid, outstanding, policy reference, year of account. Garbage in, fluent IBNR out, is still garbage. A facultative ACME loss run that is stale, or that omits the warehouse that created the TIV conflict (slip USD 42,000,000 versus SOV USD 47,100,000), will understate development if that location later claims.

Treaty IBNR is usually at class and year of account, not at named insured, except for large-loss lists. Do not invent an ACME-specific IBNR of USD 250,000 because the class number looked handy. If the large-loss listing should include ACME at USD 8,200,000 ground-up and it does not, the triangle is missing a tail event, which is a file problem before it is an actuarial problem.

Claims-made versus occurrence

On occurrence cover, IBNR includes late-reported accidents. On claims-made, the unreported bucket is a different shape: claims not yet made, plus tail if any. Applying an occurrence development pattern to a claims-made ACME professional book is a method conflict. Extract the basis from the wording before you pick a pattern.

Reinsurance IBNR also depends on reporting lag from cedent to reinsurer. A bordereau received 45 days late is not extra IBNR by itself; it is a lag you should name. Booking extra IBNR because the file was late, without a method, is a plug.

What IBNR is not

It is not the gap between slip TIV and SOV TIV. That is a schedule conflict. It is not premium deficiency. It is not a reinstatement. It is not parametric basis risk. It is not a loss ratio, though loss ratio uses incurred that may include IBNR — and you must say whether it does.

If two as-at dates appear, show two estimates. Do not blend 30 June and 30 September into one ACME figure. If the method note is missing, IBNR is a gap in the accounts pack. Dual-control on booking a round IBNR "like last year" is still dual-control of a guess.