From a Paid Triangle to Unpaid Liabilities
A paid-loss triangle records how cumulative claim payments for each accident year grow as claims mature. A chain ladder calculation turns this pattern into an estimate of ultimate paid losses. The difference between an accident year's projected ultimate and its latest observed cumulative paid amount is its estimated unpaid liability.
The calculation is straightforward once the triangle is read correctly: select age-to-age factors, chain the factors forward from each accident year's latest maturity, and compare projected ultimate losses with payments already recorded. The central challenge is using the cumulative amounts and the correct development ages at every step.
Model, Factors, and Projection
The paid chain ladder assumes that, at each development age, the relationship between cumulative paid losses at the current age and the next age is sufficiently stable across accident years to use a common factor. In practice, the factor is often selected as a volume-weighted link ratio: total next-age cumulative losses divided by total current-age cumulative losses, using accident years with observations at both ages.
This formula applies to cumulative amounts arranged by development age, not to the incremental payments between ages. Its denominator must be positive, and only accident years observed at both ages contribute. The selected factor may instead be judgmentally adjusted or otherwise selected; once selected, the projection uses that selection consistently.
For an accident year last observed at age \(k\), the cumulative development factor to terminal age \(m\) is the product of the selected age-to-age factors from age \(k\) onward. The projected ultimate is the latest observed cumulative paid amount multiplied by that product. If the terminal age is treated as ultimate, there is no additional tail factor. If a tail factor is selected, it must also be included in the product.
The deterministic chain ladder does not, by itself, specify an interest rate, a mortality model, or a probability distribution for losses. Unless an example explicitly states otherwise, amounts below are nominal paid losses in thousands of dollars: the interest rate is 0% because no discounting is performed; mortality is not applicable; and no stochastic distribution is assumed. The calculation estimates a point value under the stated development pattern, not the uncertainty around that value.
Why the Volume-Weighted Factor Works
The volume-weighted formula makes the selected factor reproduce the aggregate relationship between the observed current-age and next-age cumulative amounts. To see this, suppose the model at one development age is that expected next-age cumulative paid losses are \(f_j C_{i,j}\) for each accident year \(i\). Summing across the eligible accident years gives an expected aggregate next-age amount of \(f_j\sum_i C_{i,j}\). Setting this equal to the observed aggregate next-age amount and solving for the factor gives:
Thus, accident years with larger current cumulative paid amounts carry more weight in the selected factor. This is a feature of the volume-weighted method, not an arithmetic average of the individual link ratios. The stability assumption is crucial: if payment practices, claim mix, or other conditions have changed, one common factor may not describe all accident years adequately.
Complete Paid Triangle Calculation
Worked Example: Projecting Four Accident Years
A hypothetical insurer has the following cumulative paid-loss triangle, in thousands of dollars. Treat age 48 months as the terminal age, with no further development beyond it. Calculate the selected factors, each accident year's ultimate loss and unpaid liability, and the total unpaid liability.
| Accident year | 12 months | 24 months | 36 months | 48 months |
|---|---|---|---|---|
| 2021 | 120.00 | 180.00 | 210.00 | 225.00 |
| 2022 | 135.00 | 202.50 | 236.25 | — |
| 2023 | 150.00 | 225.00 | — | — |
| 2024 | 165.00 | — | — | — |
Model and assumptions. This is a deterministic paid chain ladder using the cumulative amounts shown and volume-weighted age-to-age factors. The interest rate is 0%; no discounting is applied. Mortality is not applicable, and no probability distribution is specified. Treating 48 months as terminal means assuming no additional paid development after that age.
Step 1: Select the 12-to-24-month factor. Use all three accident years observed at both ages:
Step 2: Select the 24-to-36-month factor. Only 2021 and 2022 have observations at both ages:
Step 3: Select the 36-to-48-month factor. Only 2021 has both observations:
Step 4: Chain the factors and project each accident year. The 2021 amount is already observed at the terminal age, so its cumulative development factor is 1.0000. At 36 months, the factor to terminal age is 1.071429. At 24 months, the factors are multiplied: \(1.166667\times1.071429=1.2500\). At 12 months, all three factors are multiplied: \(1.5000\times1.166667\times1.071429=1.8750\).
| Accident year | Latest observed paid | Latest age | Cumulative factor to 48 | Ultimate paid | Unpaid liability |
|---|---|---|---|---|---|
| 2021 | 225.000 | 48 | 1.0000 | 225.000 | 0.000 |
| 2022 | 236.250 | 36 | 1.071429 | 253.125 | 16.875 |
| 2023 | 225.000 | 24 | 1.2500 | 281.250 | 56.250 |
| 2024 | 165.000 | 12 | 1.8750 | 309.375 | 144.375 |
For example, the 2023 ultimate is \(225.000\times1.2500=281.250\), so its unpaid liability is \(281.250-225.000=56.250\), all in thousands of dollars.
Step 5: Check the portfolio total. Total projected ultimate paid losses are \(225.000+253.125+281.250+309.375=1{,}068.750\). Total cumulative paid losses currently observed are \(225.000+236.250+225.000+165.000=851.250\). Their difference is \(1{,}068.750-851.250=217.500\), which agrees with the sum of the accident-year unpaid liabilities: \(0.000+16.875+56.250+144.375=217.500\).
Common mistake avoided: applying the 12-to-24-month factor to 2023 or 2024's incremental loss would use the wrong base. Chain ladder multiplies the latest cumulative paid amount by the remaining cumulative development factor. The check confirms that the total unpaid amount reconciles.
Using Selected Factors When Ratios Differ
Worked Example: Volume Weighting Unequal Link Ratios
A hypothetical specialty insurer has cumulative paid losses, in thousands, as follows. Use volume-weighted factors and project the 2023 and 2024 ultimate losses, taking 48 months as terminal.
| Accident year | 12 months | 24 months | 36 months | 48 months |
|---|---|---|---|---|
| 2021 | 80 | 120 | 150 | 165 |
| 2022 | 100 | 145 | 174 | 191.4 |
| 2023 | 110 | 160 | — | — |
| 2024 | 125 | — | — | — |
Model and assumptions. Use deterministic volume-weighted paid chain ladder factors calculated from the eligible observations in the triangle. The interest rate is 0%; there is no discounting. Mortality is not applicable, and no probability distribution is specified. Assume no further development after 48 months.
Step 1: Calculate the 12-to-24-month factor. The three eligible pairs give:
Step 2: Calculate the 24-to-36-month factor. Only 2021 and 2022 are eligible:
Step 3: Calculate the 36-to-48-month factor. Both 2021 and 2022 contribute:
The 2023 latest observation is at 24 months. Its cumulative factor to 48 months is \((324/265)\times1.1=1.344906\). Its ultimate is \(160\times1.344906=215.185\) thousand, and its unpaid liability is \(215.185-160=55.185\) thousand.
The 2024 latest observation is at 12 months. Its factor to 48 months is \((425/290)\times(324/265)\times1.1=1.970982\). Its ultimate is \(125\times1.970982=246.373\) thousand, and its unpaid liability is \(246.373-125=121.373\) thousand. Calculations use unrounded factors; displayed results are rounded to the nearest 0.001 thousand dollars.
Method and common mistake avoided: these are volume-weighted selections, so the ratios are calculated from sums of cumulative amounts. Taking a simple arithmetic mean of individual link ratios would give a different method and a different selection. The 2024 projection also uses the 12-to-24 factor once, then each subsequent factor once; it does not add the factors together.
Projecting More Than One Remaining Development Step
Worked Example: Projecting From Different Maturities
A hypothetical claims portfolio has two accident years with latest cumulative paid losses of 350 thousand at 36 months for 2022 and 240 thousand at 12 months for 2023. Selected factors are 1.40 from 12 to 24 months, 1.10 from 24 to 36 months, and 1.05 from 36 months to ultimate. Calculate the ultimate and unpaid liability for each year and in total.
Model and assumptions. The selected factors are treated as fixed deterministic inputs. The interest rate is 0%, with no discounting; mortality is not applicable; and no probability distribution is specified. The 1.05 factor represents the supplied development from 36 months to ultimate.
2022. This accident year is already at 36 months, so only the 36-month-to-ultimate factor applies. Its ultimate is \(350\times1.05=367.50\) thousand. Its unpaid liability is \(367.50-350=17.50\) thousand.
2023. This year is at 12 months, so all three factors apply in sequence. The cumulative factor is \(1.40\times1.10\times1.05=1.617\). Its ultimate is \(240\times1.40=336\), then \(336\times1.10=369.60\), then \(369.60\times1.05=388.08\) thousand. Its unpaid liability is \(388.08-240=148.08\) thousand.
Total check. The total ultimate is \(367.50+388.08=755.58\) thousand. Latest cumulative paid losses total \(350+240=590\) thousand. Therefore, total unpaid is \(755.58-590=165.58\) thousand, matching \(17.50+148.08=165.58\) thousand.
Method and common mistake avoided: multiply only the factors for development that remains after the accident year's latest observed age. Applying the 12-to-24-month factor to the 2022 amount would overstate its ultimate because that development is already reflected in its 36-month cumulative paid amount.
Interpretation and Exam Traps
The unpaid liability from a paid chain ladder triangle is the projected remaining paid amount, measured as projected ultimate paid losses less cumulative paid losses to date. It is not automatically an estimate of all unpaid claim obligations under every accounting or reserving definition. The result depends on the triangle's data, the selected development pattern, and the assumed terminal age.
- Mixing cumulative and incremental amounts. The factors above relate one cumulative age to another. An incremental payment is not the appropriate base for the ultimate projection.
- Using observations outside the eligible age pair. A link ratio at one age requires a cumulative amount at both that age and the next. A blank cell is not a zero.
- Averaging ratios without identifying the method. A volume-weighted factor is a ratio of sums. It generally differs from an arithmetic average of individual ratios.
- Multiplying by the wrong factor set. Begin with each year's latest observed age and include only the factors from that age through the selected terminal age.
- Forgetting the terminal assumption. If the final observed age is not ultimate, a suitable further development factor is needed. Treating it as ultimate is an assumption, not a consequence of the triangle format.
- Reporting ultimate as unpaid. The ultimate includes paid losses already recorded. Subtract the latest cumulative paid amount to obtain the unpaid estimate.
A disciplined calculation labels each accident year's latest age, writes the remaining factor product, and then performs the ultimate-minus-paid subtraction. The portfolio reconciliation is a useful final check: total unpaid should equal total projected ultimate less total cumulative paid to date.
Check Your Understanding
For each hypothetical problem, use the stated deterministic chain ladder assumptions. Unless specified otherwise, amounts are in thousands of dollars, no discounting is applied, the interest rate is 0%, mortality is not applicable, and no probability distribution is assumed.
- A paid triangle has a latest 24-month cumulative amount of 126 for one accident year. The selected 24-to-36 and 36-to-ultimate factors are 1.20 and 1.00. Find ultimate paid losses and unpaid liability. (Answer: ultimate 151.20; unpaid 25.20.)
- An accident year has 200 paid at 12 months. Selected factors are 1.25 from 12 to 24 months and 1.08 from 24 months to ultimate. Calculate the cumulative factor, ultimate, and unpaid liability. (Answer: cumulative factor 1.35; ultimate 270; unpaid 70.)
- For the 12-to-24-month age pair, eligible cumulative paid amounts are 100 and 200, with next-age amounts 130 and 250. The latest 24-month cumulative amount for another accident year is 180, and its selected 24-to-36 factor is 1.20; treat 36 months as ultimate. Calculate the volume-weighted 12-to-24 factor and that accident year's ultimate and unpaid liability. (Answer: factor 1.2667; ultimate 216; unpaid 36.)
- An accident year has 90 paid at 12 months. Its selected factors are 1.40 from 12 to 24 months, 1.15 from 24 to 36 months, and 1.05 from 36 months to ultimate. Find the ultimate and unpaid amount, rounding to the nearest 0.001 thousand. (Answer: ultimate 152.145; unpaid 62.145.)
- Three accident years have latest cumulative paid amounts of 110 at 36 months, 160 at 24 months, and 200 at 12 months. Selected factors are 1.10 from 36 months to ultimate, 1.25 from 24 to 36 months, and 1.50 from 12 to 24 months. Find the total unpaid liability. (Answer: 283.5.)