Actuarial Models in Practice · Part 3 of 16
IBNR Reserving: How Insurers Estimate Claims They Haven't Seen Yet
In short: IBNR reserving estimates what an insurer owes on claims that have happened but not yet been notified or fully developed. The Chain Ladder method constructs development factors from a historical triangle of cumulative paid claims and projects the latest diagonal forward to an ultimate settled amount. The difference between the ultimate and the current paid position is the IBNR reserve.
An insurance company’s balance sheet carries a claims reserve — its best estimate of what it will ultimately pay out on all policies currently in force or recently expired. Part of this reserve is straightforward: known claims where the insurer has received notification and set a case reserve. But a significant portion relates to events that have already occurred but whose full liability has not yet emerged. This is the IBNR reserve, and for long-tail lines of business such as liability, professional indemnity, or workers’ compensation, it can dwarf the known claims reserve.
Estimating IBNR is a core actuarial function, but the analytical framework — a development triangle plus a development factor calculation — is accessible to any financial analyst with a grasp of how historical claims evolve over time.
Why Claims Take Time to Develop
Before building the triangle, it is important to understand why claims development exists at all. Three dynamics drive it:
Reporting delay. For some lines of business, the claimant does not know immediately that they have suffered a loss. A worker exposed to asbestos in 1990 may not develop mesothelioma until 2020. A medical negligence victim may not discover the error for years. By definition, the insurer cannot reserve what it does not know about — hence “Incurred But Not Reported.”
Settlement delay. Even reported claims take time to reach final settlement. A large liability claim goes through investigation, legal proceedings, possibly appeal — a process that can take years. During this time, the case reserve set by the claims team may prove inadequate as the full extent of the loss becomes clear.
Subrogation and salvage. Recoveries from third parties or from salvage of insured assets reduce the net claims cost but take time to materialise. A property insurer who pays a claim for storm damage may subsequently recover from the negligent contractor responsible — but that recovery appears months or years after the initial payment.
Short-tail lines (home contents, motor property damage) typically develop to ultimate within 12–24 months. Long-tail lines (employers’ liability, professional indemnity, directors’ and officers’) can take 10–20 years to fully run off.
The Development Triangle
A development triangle arranges historical claims data into a matrix. The rows are accident years — the year in which the insured event occurred. The columns are development periods — how many years have elapsed since the accident year. Each cell contains the cumulative paid or incurred claims for that accident year as observed at that development period.
Example triangle — cumulative paid claims (£000s):
| Accident Year | Dev 1 | Dev 2 | Dev 3 | Dev 4 | Dev 5 |
|---|---|---|---|---|---|
| 2020 | 1,200 | 2,150 | 2,780 | 3,050 | 3,180 |
| 2021 | 1,350 | 2,320 | 3,010 | 3,300 | — |
| 2022 | 1,480 | 2,600 | 3,380 | — | — |
| 2023 | 1,620 | 2,810 | — | — | — |
| 2024 | 1,750 | — | — | — | — |
The latest diagonal (bottom-right stepping cells) shows the most recent data available: Dev 5 for accident year 2020, Dev 4 for 2021, Dev 3 for 2022, Dev 2 for 2023, and Dev 1 for 2024. The blank cells represent future development that has not yet occurred.
Calculating Development Factors
The Chain Ladder method derives a development factor (also called a link ratio or age-to-age factor) for each development period. This factor represents the average ratio by which cumulative claims grow from one development period to the next.
For the transition from Dev 1 to Dev 2:
f(1→2) = Σ (Dev 2 claims across all accident years with both Dev 1 and Dev 2 data)
──────────────────────────────────────────────────────────────────────────
Σ (Dev 1 claims across those same accident years)
= (2,150 + 2,320 + 2,600 + 2,810) / (1,200 + 1,350 + 1,480 + 1,620)
= 9,880 / 5,650
= 1.749
Repeating for each transition:
| Transition | Development factor |
|---|---|
| Dev 1 → 2 | 1.749 |
| Dev 2 → 3 | 1.297 |
| Dev 3 → 4 | 1.097 |
| Dev 4 → 5 | 1.043 |
| Dev 5 → Ult | 1.010 (tail factor) |
The tail factor is applied beyond the last observed development period to capture residual development to ultimate. For short-tail lines it is close to 1.00; for long-tail lines it may be 1.05–1.20 or higher and requires separate actuarial judgement.
Projecting to Ultimate and Calculating IBNR
Apply the development factors to the latest diagonal of the triangle to project each accident year to ultimate:
| Accident Year | Latest paid (£000s) | Factors applied | Ultimate (£000s) | IBNR (£000s) |
|---|---|---|---|---|
| 2020 | 3,180 (Dev 5) | × 1.010 | 3,212 | 32 |
| 2021 | 3,300 (Dev 4) | × 1.043 × 1.010 | 3,476 | 176 |
| 2022 | 3,380 (Dev 3) | × 1.097 × 1.043 × 1.010 | 3,906 | 526 |
| 2023 | 2,810 (Dev 2) | × 1.297 × 1.097 × 1.043 × 1.010 | 4,212 | 1,402 |
| 2024 | 1,750 (Dev 1) | × 1.749 × 1.297 × … | 4,588 | 2,838 |
| Total | 19,394 | 4,974 |
The total IBNR reserve of £4,974k is the insurer’s best estimate of claims it will pay on accident years 2020–2024 that have not yet been paid. This sits on the balance sheet as a liability alongside the case reserves on known reported claims.
Alternative Methods and Actuarial Judgement
The Chain Ladder method is the most widely used but not the only approach:
Bornhuetter–Ferguson (BF) method. Blends the Chain Ladder projection with an a priori expected loss ratio. Useful when the early development periods have little data — for example, the most recent accident year where only one development period has elapsed. Pure Chain Ladder gives an unstable result with limited data; BF anchors the projection to a credible prior.
Average cost per claim (frequency × severity). Decomposes ultimate claims into claim count (frequency) and average cost per claim (severity). Useful for lines where the number of open claims is tracked carefully, such as motor bodily injury.
Paid vs incurred triangles. Running the same analysis on incurred claims (paid plus case reserves) as well as purely paid claims provides a cross-check. If paid and incurred triangles give materially different answers, the case reserving may be inadequate or excessive.
Actuaries apply judgement to weight these methods depending on the line of business, data volume, and known anomalies. A large catastrophe event in accident year 2022, for example, would distort the development factors if not excluded or adjusted — the actuary must identify this and either cap the factor or exclude the year from the factor calculation.
Key Takeaways
- IBNR reserves estimate the insurer’s liability on events that have occurred but where claims have not yet been fully reported or settled.
- The development triangle arranges cumulative claims by accident year and development period. The latest diagonal is the current position; blank cells are future development.
- Chain Ladder derives development factors from historical ratios and projects the latest diagonal to an ultimate settled amount. IBNR = Ultimate − Current paid.
- Short-tail lines develop within 1–2 years; long-tail liability lines can take 10–20 years, requiring careful tail factor selection.
- Chain Ladder is the base method but must be supplemented with Bornhuetter-Ferguson for thin data years and cross-checked against incurred triangles and frequency/severity methods.
Practice
Build a 5×5 development triangle in Excel using the paid claims data above. Calculate development factors for each transition using the volume-weighted average. Apply a tail factor of 1.012 beyond development period 5. Project each accident year’s paid claims to ultimate and calculate the IBNR reserve for each year and in total. Then replace 2024’s development factor from Dev 1 to Dev 2 with an alternative of 2.10 (reflecting a large late-emerging event year) — quantify the change in the total IBNR reserve and comment on the sensitivity.
Frequently asked questions
What does IBNR mean in insurance?
What is the Chain Ladder method?
What is a claims development triangle?
What is the difference between IBNR and IBNER?
Topics
Related posts
Learn how general insurers measure underwriting performance using loss ratios, expense ratios, and the combined ratio — with worked examples and a guide to interpreting results across different lines of business.
Actuarial & Insurance Models Experience Analysis: How Actuaries Compare Actual Results to AssumptionsLearn how actuaries conduct experience analysis — comparing actual mortality, lapse, and claims experience to model assumptions using A/E ratios, credibility weighting, and structured investigation frameworks.
Actuarial & Insurance Models Solvency II: What the Capital Framework Means for Insurance AnalystsA practical introduction to Solvency II — the EU and UK regulatory capital framework for insurers. Covers the three-pillar structure, SCR components, Own Funds tiers, and how the SCR ratio is used by investors and management.