Why a claim is not paid the moment it happens
In the life-insurance rungs you mostly knew, on the day of a claim, exactly what was owed: a policy promises a fixed sum, someone dies, the sum is paid. Non-life insurance is not so tidy. When a car crashes on a rainy Tuesday, almost nothing is known yet. Will the other driver sue? Is the neck injury a sprain or a year of physiotherapy? Will a court rule in two years' time? The loss event has happened, but the *amount* is a slow-developing fog — and the cheque cannot be written until the fog clears.
Two separate lags stretch the gap between an accident and its payment. First the reporting lag: time passes before the insurer even hears that a loss occurred — a leak found months later, an injury that worsens, a liability claim filed near a legal deadline. Then the settlement lag: once reported, a claim is investigated, negotiated, sometimes litigated, and paid out in instalments over years. A simple fender-bender might close in weeks; a serious bodily-injury or asbestos claim can stay open for a decade. The slower the line of business, the longer money sits owed but unpaid.
The biggest number on the balance sheet
Add up everything an insurer still owes for accidents that have already occurred — every reported-but-open claim, plus every claim it does not yet even know about — and you get the unpaid-claim liabilities, also called loss reserves. For most non-life insurers this is the single largest item on the liability side of the balance sheet, often dwarfing every other figure. Get it a few percent too low and the company looks healthier and more profitable than it is; a few percent too high and it is needlessly starving its capital. The whole rung exists to estimate this number honestly.
Bust the misconception now, because it runs deep: a loss reserve is not a pile of cash sitting in a vault labelled 'claims.' It is an accounting *estimate* — a number on the liability side of the balance sheet recording money the insurer believes it will have to pay out later. The cash backing it is invested in bonds and other assets, earning a return while the claims slowly settle. A reserve is a measured promise about the future, not a locked box of money waiting to be handed over.
Because the money is owed *later*, the time value of money quietly enters. A claim you will pay in five years costs you less today than the same amount paid now — discounting at the present value you learned earlier would lower the reserve. Yet much non-life reserving is held *undiscounted*, on purpose: the conservatism cushions against the very uncertainty this guide is about, and short-tailed lines settle too fast for discounting to matter much. Whether to discount, and at what rate, is a real and contested choice — not an oversight.
Case reserves vs bulk reserves: two ways to estimate
There are two fundamentally different ways to put a number on what is owed, and the unpaid-claim total is built from both. The first is the case reserve: a claims adjuster looks at one specific, *reported* claim and estimates, file by file, what it will ultimately cost — 8,000 for this whiplash, 250,000 for that liability suit. It is bottom-up, hands-on, and tied to a real claimant. Case reserves are the careful human judgement of people who have read the file.
Case reserves are indispensable but they cannot be the whole story, for two reasons that map directly onto the next section. They only exist for claims that have been *reported* — they are blind to accidents the insurer has not heard about yet. And adjusters, being human, tend to set the estimate at first notice and revise it slowly, often understating how claims grow as facts emerge. So on top of the file-by-file estimates, the actuary adds a bulk reserve: a single statistical top-up for the whole portfolio, estimated from how losses have *developed* historically rather than from any one file. The case-versus-bulk split is the spine of the entire reserve.
IBNR and IBNER: the two halves of what case reserves miss
The bulk reserve has a famous two-part name, and the two parts answer the two weaknesses of case reserves precisely. The first is IBNR — *Incurred But Not Reported*. These are accidents that have already happened but that the insurer has not heard about as of the valuation date. The leak in the wall, the injury that has not yet been claimed, the lawsuit not yet filed. By definition there is no file and no case reserve for an IBNR claim — you are reserving for losses you cannot point to one by one. This is the IBNR reserve, and it is largest in long-tailed lines where reporting drags on for years.
The second part is IBNER — *Incurred But Not Enough Reported*, sometimes called case-reserve development. Here the claim *is* known and *does* have a case reserve, but that estimate is going to move — usually upward — as the facts harden. The whiplash reserved at 8,000 becomes 30,000 once surgery is needed; the liability suit reserved at 250,000 settles for 600,000 after the verdict. IBNER is the expected future development on claims already in the file. One letter separates the two — IBN-R is about claims not yet *Reported*, IBN-E-R is about claims not *Enough* reported — but they are genuinely different gaps, and a good reserve estimates each.
Accident year 2025, valued at 31 Dec 2025
Paid to date (cheques cut) ........ 1,200,000
Case reserves (adjuster files) ........ 900,000
-------------------------------------------------------
Reported / incurred so far ........ 2,100,000
+ IBNER (case reserves will grow) ........ 300,000
+ IBNR (unreported accidents) ........ 600,000
-------------------------------------------------------
Estimated ULTIMATE loss ........ 3,000,000
Unpaid-claim liability = Ultimate - Paid
= 3,000,000 - 1,200,000
= 1,800,000 (the reserve held)One honest footnote: the reserve must also cover the cost of *handling* the claims, not just the indemnity paid to claimants — the lawyers, adjusters, and experts. That is the loss-adjustment expense, and a complete unpaid-claim liability includes it. Different countries and standards slice IBNR and IBNER along slightly different lines, and some lump 'pure IBNR' with 'IBNER' under one heading. The concepts, though, are universal: there are claims you have not heard of, and claims whose true cost you have under-recorded. Both must be reserved for.
Organising the chaos: accident years and the triangle ahead
To estimate IBNR and IBNER you need a tidy way to watch claims grow up over time, and the first move is to file every claim under the year its *accident* happened, not the year it was reported or paid. This is accident-year organisation: all losses from accidents in 2023 belong to the 2023 cohort forever, no matter when the cheques eventually clear. Group claims this way and a pattern emerges — each accident year's losses build up over successive calendar years in a remarkably stable rhythm.
Lay those cohorts in rows and the calendar of their growth in columns, and you draw the loss-development triangle — the central, almost iconic tool of this rung. Older accident years have run on longer, so their rows stretch further to the right; the most recent year has barely started, so its row is short. The data fills only the upper-left, leaving an empty lower-right corner. Estimating the unpaid claims is, quite literally, the art of filling in that blank corner: projecting each immature row out to its ultimate cost.
That is exactly where the rest of this rung goes. Next you will learn to read the triangle and compute development factors; then the chain ladder, which projects the future from the past pattern alone; then Bornhuetter-Ferguson, which blends that pattern with a prior expectation of losses for the green, immature years where the chain ladder is dangerously unstable. And running underneath all of them is the honest theme you already met with VaR: these are *estimates* with real uncertainty, and a single best-estimate number quietly hides a wide range of possible outcomes — which is why stochastic reserving exists to measure that range.