Your actuarial report includes a table of selected development factors and a projected ultimate for each accident year. The chain ladder produced most of those projections. If you understand what the chain ladder is doing, you understand the single largest driver of the number you are being asked to accept. If you do not, you are trusting a mechanical process that may not fit your data.
This article explains the chain ladder in plain language: the core idea, the arithmetic, the assumptions it makes, the situations where it works well, and the situations where it produces a number you should not trust. It is written for the risk manager, CFO, or captive board member who receives a reserve estimate and needs to evaluate it, not for the actuary who produces it.
What the chain ladder does
The chain ladder (also called the development technique) is the most frequently used reserving method in practice. It projects future claim payments or reported losses by applying historical development patterns to the current data. The logic is straightforward: if claims at 24 months of maturity have historically grown to 1.45 times their value by 36 months, then today’s 24-month claims should grow by about the same factor.
The method starts with a loss development triangle. The actuary calculates age-to-age factors from historical experience at each maturity interval, selects a representative factor for each interval (using an average, a weighted average, or judgment), chains those selected factors together into a cumulative development factor (CDF, also called an LDF), and multiplies each accident year’s current evaluation by its CDF to project the ultimate.
The chain ladder makes one core bet: the past development pattern is the best predictor of future development. Friedland states the key assumption directly: “claims recorded to date will continue to develop in a similar manner in the future” (Friedland, p. 84). Everything that follows in this article is a consequence of that assumption, including the situations where it holds and the situations where it does not.
The arithmetic, without a spreadsheet
You do not need to build a triangle to understand the mechanics. A single accident year is enough.
Suppose your workers compensation program’s 2023 accident year has $1.2 million in reported losses as of December 2025 (36 months of development). The actuary’s selected development pattern says that, historically, reported losses at 36 months are about 75% of their ultimate value. That means the CDF from 36 months to ultimate is 1.333 (the reciprocal of 0.75).
Projected ultimate = $1.2 million x 1.333 = $1.6 million
The IBNR for this accident year is the gap between the projected ultimate and the current reported losses: $1.6 million minus $1.2 million = $400,000.
Now consider the 2025 accident year, which has only 12 months of development and $300,000 in reported losses. If the CDF at 12 months is 4.00 (meaning 12-month losses are historically about 25% of ultimate), the projected ultimate is $300,000 x 4.00 = $1.2 million, and the IBNR is $900,000.
Two things are worth noticing. First, the more immature the accident year, the more the CDF does the work. For the 2025 year, 75% of the projected ultimate is IBNR; for the 2023 year, it is 25%. Second, the accuracy of the projection depends entirely on whether the historical pattern still applies. If it does, the arithmetic is sound. If it does not, no amount of precision in the calculation helps.
The four conditions the assumption requires
The chain ladder’s core assumption holds only when four conditions are met. When any one of them breaks, the chain ladder still produces a number. It just produces the wrong one.
1. Consistent claim processing. The claims team handles, investigates, and closes claims in a manner similar to the experience period that generated the triangle. If the TPA changed, if staffing levels shifted, or if management introduced a new claims protocol, the speed at which claims move through the system may no longer match the historical pattern.
2. Stable case reserve adequacy. The reported triangle includes the adjuster’s case estimate (also called case outstanding or case reserves). If case reserves became systematically more or less conservative during the experience period (because of a new claims director, a regulatory audit, or a shift in caseload per adjuster), the reported development factors embed that change as if it were real loss development. It is not. It is an artifact of a practice change, and the chain ladder cannot tell the difference.
3. Stable mix of claim types. If the composition of claims changes (more litigated claims, more severe injuries, a shift in geographic exposure), the historical pattern may not match the current book. A triangle blends all claim types together, and a shift in the blend changes the development pattern even if each claim type individually behaves the same.
4. Stable policy limits and retentions. If the self-insured retention increased from $250,000 to $500,000 between policy years, the development triangle now contains two different insurance structures. Claims in the older years are capped at $250,000; claims in the newer years can develop to $500,000. The historical pattern from the lower-retention years cannot reliably predict development at the higher retention.
Paid triangle vs. reported triangle
Actuaries run the chain ladder on both paid and reported triangles, and the two versions fail in different ways.
The paid chain ladder projects from cumulative payments. It is not affected by case reserve adequacy because it ignores case reserves entirely. But it is highly sensitive to changes in claim settlement speed. If claims are closing faster or slower than the historical pattern (because of a TPA change, a litigation management initiative, or a shift in jurisdictional practice), the paid development factors are distorted.
The reported chain ladder projects from cumulative incurred losses (paid plus case outstanding). It reacts faster to emerging severity because case reserves adjust as the adjuster learns more about the claim. But it inherits every change in case reserving philosophy. If the claims team started setting higher initial reserves two years ago, the reported triangle will show lower development factors for those years, not because claims are developing less, but because more of the development was front-loaded into the initial reserve.
A good actuarial analysis runs both, compares them, and explains where they agree and where they diverge. When they diverge, one of the two triangles has been distorted by an operational change, and the actuary should be able to tell you which one and why.
When it works
The chain ladder is the natural default for accident years that have enough maturity for the development pattern to stabilize and enough data for the factors to be credible. For a typical self-insured workers compensation program, that usually means accident years with 24 or more months of development and at least a few dozen claims per year.
For mature accident years (five or more years old in most lines), the chain ladder on the reported triangle often produces the tightest estimate. The remaining development is small, the factors are close to 1.00, and the projection is not heavily leveraged. At this maturity, the paid and reported chain ladders should produce similar indications. If they do not, that is a diagnostic signal worth investigating.
For the broader method landscape and how the chain ladder fits alongside Bornhuetter-Ferguson, Cape Cod, and frequency-severity, see the hub article on the five core methods.
When it breaks
The failure modes are specific and predictable. Each one maps to one of the four conditions above.
Case reserve strengthening or weakening. This is the most common source of distortion in a reported chain ladder for self-insured programs. It happens after TPA changes, new claims leadership, adequacy audits, or regulatory pressure. The signal shows up as inflated or deflated age-to-age factors in recent calendar periods relative to earlier years. If the reported development factors are systematically different in the last two or three diagonals, the first question is whether case adequacy shifted. For a deeper look at how to diagnose this, see What’s Actually Driving Your IBNR Higher?.
Settlement speed change. This is the corresponding failure for the paid chain ladder. A faster closure rate makes it look like paid losses are developing more quickly, which depresses the CDF and understates the ultimate. A slower closure rate does the opposite. The signal is a shift in the ratio of closed to open claims at a given maturity.
Change in mix, retention, or limits. These changes affect both triangles. A retention increase means newer accident years have a longer development tail than older ones. A shift toward more litigated claims means the development pattern is longer and steeper. A limit decrease caps development for newer years relative to older ones. In each case, the historical pattern is not representative of the current exposure.
Unusually large claims in immature years. For a self-insured program with a $500,000 retention, a single claim that hits the retention limit early in an accident year can dominate the reported losses at 12 months. The chain ladder treats that large reported amount as the base and multiplies it by the CDF, producing an inflated ultimate. Conversely, if the large claim has not yet been reported, the base is artificially low and the projected ultimate is understated.
The self-insured leverage problem
This is the single most important wrinkle for self-insured buyers.
Self-insured programs generate far fewer claims than a carrier’s book. That means the loss triangles are thinner, the development factors are noisier, and the most recent accident year’s CDF is highly leveraged. A paid bodily injury program might have a 12-month CDF of 90.00, meaning the chain ladder projects ultimate losses at ninety times the current paid amount (see Friedland, p. 134, for this example). At that leverage, a single unexpected payment of $50,000 on the current diagonal moves the projected ultimate by $4.5 million.
This is not a theoretical risk. It is the practical reason most actuaries do not rely on the chain ladder alone for the most recent accident years of a self-insured program. They use Bornhuetter-Ferguson or the expected claims method to anchor the projection to something more stable than a thin, leveraged base. BF is specifically designed to handle this: it blends the chain ladder signal with an a priori expected loss, weighting toward the expectation when the data is immature and toward the data when it is mature.
If your actuary is relying on a chain ladder projection for your most recent accident year and the CDF is above 5.00, you should ask why BF or expected claims was not preferred. The answer may be defensible, but it should be explicit.
What a buyer should ask their actuary
These questions test whether the chain ladder projection in your report is supported by the data and the method’s assumptions.
1. Which triangle drove the selection for the most recent accident year: paid or reported? The answer tells you which set of assumptions the actuary is betting on. If reported, the implicit bet is that case reserve adequacy has been stable. If paid, the bet is that settlement speed has been stable. Ask for the evidence.
2. Can you show that case adequacy (or settlement speed) has been stable over the experience period? This is the assumption that makes the chain ladder work. If the actuary cannot demonstrate stability, the chain ladder should not be the primary method for the affected accident years.
3. What CDF is applied to the latest accident year, and how sensitive is the ultimate to that factor? Ask the actuary to show what the projected ultimate would be if the CDF moved up or down by 10% or 20%. If a modest change in the factor swings the projected ultimate by a material amount, you are relying on a volatile projection and should understand the range of outcomes, not just the point estimate.
4. Were any age-to-age factors excluded from the selected averages, and if so, why? Actuaries commonly exclude high or low factors before averaging. Exclusions are not inherently wrong, but they should be documented and explained. An exclusion without explanation is a hidden assumption.
5. How does the chain ladder indication compare to BF for each accident year? Where the two methods agree, the chain ladder is well-supported. Where they diverge, the actuary should explain which method’s assumptions are more defensible for that year. Divergence between these two methods is one of the most useful diagnostic signals in a reserve review.
What to require in documentation
For any accident year where the chain ladder is the selected method, your report should include:
- The triangle (paid and reported) used in the analysis, with enough history to evaluate pattern stability.
- The selected age-to-age factors, with the averaging method identified (all-year, five-year, volume-weighted, excluding outliers) for each development interval.
- The resulting CDFs by accident year.
- A comparison of the chain ladder indication against at least one other method (typically BF) for each accident year.
- An explicit statement about whether case reserve adequacy and settlement speed have been stable over the experience period, and if not, what adjustments were made.
- A sensitivity exhibit for the most recent accident year showing the effect of reasonable CDF variation on the projected ultimate.
If the report does not contain these items, the chain ladder number is unsupported. It may still be correct, but you have no way to evaluate it, and that is the same problem as having no number at all. For why a single number without context is the least useful output, see Point Estimate vs. Range.
Further reading
For the foundational concept of IBNR that the chain ladder estimates, see IBNR, Explained Without the Jargon. For a walkthrough of the triangle that feeds the chain ladder, see How to Read a Loss Development Triangle. For the broader method landscape and how the chain ladder fits alongside BF, Cape Cod, and frequency-severity, see How Actuaries Estimate Your Unpaid Claims. For diagnostic signals that the assumptions behind a chain ladder have broken, see What’s Actually Driving Your IBNR Higher? and Five Leading Indicators of Adverse Reserve Development.