Articles

Chain ladder method example: from a reported triangle to IBNR

Weighted age-to-age factors, the age-to-ultimate factors they multiply into, the ultimate of each accident year and the IBNR, on a five-year reported triangle.

The chain ladder projects each accident year to its ultimate cost by assuming that its losses will keep developing the way older years developed at the same ages. On a five-year reported triangle, weighted factors of 1.350, 1.101, 1.029 and 1.011 multiply into age-to-ultimate factors from 1.000 to 1.546, and the latest diagonal grows into $21,139,230 of ultimate losses, of which $2,429,230 is IBNR.

Worked in full in The Loss Reserving Handbook by Julian R. Sterling, with every figure reproduced in a free workbook.Get the DRM-free PDF, $19.99 →

What the method assumes

A claim reported this year is rarely worth, at the end of the year, what it will finally cost. Adjusters raise and lower case reserves as facts come in, late claims arrive, and some files reopen. The chain ladder, also called the development method, does not try to model any of that claim by claim. It looks at the aggregate: how much did the reported losses of each accident year grow between 12 and 24 months, between 24 and 36, and so on, and it applies the same growth to the years that have not reached those ages yet.

That is one strong assumption, and everything else is arithmetic. If the claims department has not changed how it sets case reserves, if settlement speed is stable, and if the mix of claims is steady, the past pattern is a fair guide. When one of those breaks, the method carries the break straight into the estimate, which is why reviewers read the data before they read the factors.

The triangle

A sample book of business, not the book's case. Rows are accident years, the year in which the loss event happened. Columns are development ages, measured in months from the start of the accident year, so 12 months is December 31 of the accident year itself. Each cell is cumulative reported losses, paid plus case reserves, at that age. The evaluation date is December 31, 2024, which is why each row stops one cell earlier than the row above it. All figures are teaching assumptions.

Accident year12 months24 months36 months48 months60 months
2020$2,450,000$3,310,000$3,640,000$3,750,000$3,790,000
2021$2,580,000$3,460,000$3,830,000$3,940,000
2022$2,710,000$3,690,000$4,050,000
2023$2,890,000$3,890,000
2024$3,040,000

The bottom-right half is empty because it lies in the future. The chain ladder fills it in.

Step 1: age-to-age factors

An age-to-age factor, or link ratio, is a later cell divided by the cell before it in the same row. Each column pair gives one link ratio per accident year that has both cells. Analysts then summarize each column pair with an average. This example uses the volume-weighted average over all available years: the sum of the later cells divided by the sum of the earlier cells, for the years that have both. Larger years therefore count for more than small ones, and a single odd year in a small cell carries less weight than it would in a simple average.

DevelopmentYears usedWeighted factor
12 to 24 months41.350
24 to 36 months31.101
36 to 48 months21.029
48 to 60 months11.011

Notice how few observations support the later factors. The 48 to 60 month factor rests on a single year. In a real review that is where judgment and outside information come in, and where an analyst writes down why the selected factor is what it is. Here the weighted averages are taken as selected, rounded to three decimals.

Step 2: tail and age-to-ultimate factors

The triangle ends at 60 months, but claims can still move after that. The tail factor carries development beyond the last age observed. This sample declares a tail of 1.000: it assumes the oldest year is fully developed at 60 months, which is a choice made for the example, not a finding. A long-tailed line would rarely support it.

The age-to-ultimate factor of an age, also called the cumulative development factor, is the product of every selected factor from that age onward and the tail. The product is computed from the rounded selected factors and then rounded to three decimals, so anyone can rebuild it from the printed figures.

AgeAge-to-ultimate factorBuilt from
12 months1.5461.350 x 1.101 x 1.029 x 1.011 x tail
24 months1.1451.101 x 1.029 x 1.011 x tail
36 months1.0401.029 x 1.011 x tail
48 months1.0111.011 x tail
60 months1.000tail

Step 3: ultimates and IBNR

Each accident year's ultimate is its latest reported value multiplied by the age-to-ultimate factor of its current age. IBNR, in the broad sense used here, is ultimate minus reported: it covers both claims not yet reported and further development on claims already known.

Accident yearLatest reportedAgeFactorUltimate
2020$3,790,00060 months1.000$3,790,000
2021$3,940,00048 months1.011$3,983,340
2022$4,050,00036 months1.040$4,212,000
2023$3,890,00024 months1.145$4,454,050
2024$3,040,00012 months1.546$4,699,840
Total$21,139,230

The five years add up to $21,139,230 of ultimate losses. Take away what has already been reported on the latest diagonal and $2,429,230 remains as IBNR. Nearly all of it belongs to the two youngest years, which is typical of the method: the less developed a year, the larger its factor, and the more its estimate depends on a single cell.

Reading the result

The leverage is easy to underestimate. The 2024 year is worth $4,699,840 only because its one cell is multiplied by 1.546. If that cell had been overstated by a large case reserve set early, the overstatement would be multiplied too. The opposite happens when a slow year reports less than usual at 12 months: the method projects too little. This is the main reason analysts rarely take the chain ladder alone for the youngest years, and why the Bornhuetter-Ferguson method exists.

The mature years behave the other way. The 2021 year moves by about one percent, and 2020 not at all under the declared tail. For years like these, the chain ladder is usually the most direct reading of the company's own experience.

Paid or reported?

The same steps run on paid losses. A paid chain ladder ignores case reserves entirely, so it is not moved by a change in how adjusters reserve, but its factors are larger and its estimates for young years are more volatile, and it is moved by any change in settlement speed. A reported chain ladder uses the information inside case reserves, which makes it more responsive early, and is moved by any change in reserving practice. Running both and explaining the gap between them is one of the first diagnostics of a review.

The logic is not limited to insurers. An employer health plan runs the same idea on monthly lag triangles and completion factors; IBNR in a self-funded health plan shows the first year end of such a plan.

When the chain ladder misleads

Points a reviewer looks at

Mistakes that recur

The same steps in the book

The book's case is fictional: Ashmerrow Mutual Insurance Company, a regional mutual writing commercial auto liability in its home state, reviewed at December 31, 2026 over accident years 2017 to 2026, gross of reinsurance, with ALAE kept out of the loss triangles. Every amount, premium, factor and ratio is a teaching assumption, not a market or industry figure.

Chapters 4 to 6 run these steps on ten accident years and ten ages. The incurred age-to-ultimate factor at 12 months is 1.883, so the 2026 diagonal of $9,118,000 becomes $17,169,194; at 120 months only the declared tail of 1.005 remains, and the 2017 diagonal of $9,925,000 becomes $9,974,625. The ten incurred chain ladder ultimates add up to $126,861,941. The paid chain ladder reads $129,483,944, $2,622,003 higher, and the book explains that gap before choosing. In the end the incurred chain ladder is selected for the six oldest years only; the four youngest take Bornhuetter-Ferguson.

Sources and limits: the standards named on this page are summarized as read in November 2026: Actuarial Standard of Practice No. 43, Property/Casualty Unpaid Claim Estimates (Actuarial Standards Board), and a Casualty Actuarial Society study text on basic techniques for estimating unpaid claims (Version 3, 2010). They are paraphrased in principle and they can change. This is not actuarial advice and not a statement of actuarial opinion: every figure is a sample, and only a qualified actuary working on the company's own data can say which method and which factors fit it.

Questions people ask

How do you calculate the chain ladder method?

Build a cumulative triangle, compute age-to-age factors and average them by column, multiply the selected factors and the tail into age-to-ultimate factors, then multiply each accident year's latest value by the factor of its age. IBNR is ultimate minus reported.

What is the difference between age-to-age and age-to-ultimate factors?

An age-to-age factor measures growth from one age to the next. An age-to-ultimate factor is the product of all the later age-to-age factors and the tail, and takes a year from its current age to ultimate.

Should the chain ladder use paid or reported losses?

Both are common and they fail in different ways: paid is moved by settlement speed, reported by case reserving practice. As read in November 2026, ASOP No. 43 asks the actuary to consider methods, generally more than one, and requires no specific method.

Why does the chain ladder overstate or understate young accident years?

Because the youngest year rests on one cell multiplied by the largest factor. Any distortion in that cell is multiplied with it.

Is the chain ladder enough for a reserve?

Rarely on its own. Analysts compare it with other methods, read diagnostics first and select by accident year. A qualified actuary decides what fits a given company's data.

Read the whole case

This article is one step of The Loss Reserving Handbook. The book follows one regional insurer through one year-end review: a claims registry turned into paid, reported and count triangles, diagnostics read before any method, factors chosen with their reasons, two declared tails, the chain ladder, the expected loss ratio method, Bornhuetter-Ferguson and Cape Cod side by side, an estimate selected year by year with its range, ALAE and ULAE added once, the next year compared with what was expected, and a captive with a thin triangle.

Paperback PDF edition, $19.99 Book page and free files

Free: the loss reserving calculator in Excel.

Also on this site

Case figures computed by the book's model; the companies are fictional and every term is a teaching assumption. Figures checked on 21 November 2026. Standards described as read on the dates given; not actuarial advice.