Why Select a Factor for Each Development Age?
A loss development triangle records cumulative losses at successive development ages. The change from 12 to 24 months need not resemble the change from 24 to 36 months: early development may include rapid reporting or payment, while later development may consist of slower settlement and adjustment. A single factor applied at every age would ignore that shape.
Chain ladder therefore uses an age-to-age factor for each interval. An actuary selects one factor for each interval after considering the observed link ratios and whether the experience is suitable for projection. The selections together describe a development pattern: the sequence of multipliers from one age to the next, and, when applicable, a tail factor beyond the last modeled age.
The observed ratios are evidence, not automatic selections. They may vary across accident years, and a candidate should not assume that the factor selected for one interval is appropriate at another. Selection is age-specific: examine the evidence for that interval, decide on a factor suitable for the intended projection, and record the choice. The particular averaging rules used to summarize observed ratios are separate methods; the examples here focus on how age-specific selections form and operate as a pattern.
From Selected Factors to a Complete Pattern
A pattern is complete when it specifies the selected factor for every interval from the earliest modeled age through the final observed age, and includes a tail factor if the projection extends beyond that age. For example, factors for 12–24, 24–36, and 36–48 months must be kept in that order. Multiplying them gives the factor from 12 to 48 months.
If a selected tail factor \(f_{\text{tail}}\) extends the pattern beyond the last modeled age, include it in the product for an age-to-ultimate multiplier. The product from a particular age to ultimate is the remaining development factor from that age, not necessarily the product of every factor in the full pattern.
Here, \(m\) is the last modeled development age before the tail, \(F_j\) is the age-to-ultimate multiplier from age \(j\), and \(\widehat{U}_{i}\) is the projected ultimate for an accident year currently at age \(j\). If no tail is used, omit \(f_{\text{tail}}\). The formula assumes that the selected factors are applicable to the losses being projected and that the cumulative loss basis is consistent across development ages.
Why the Factors Are Chained
The multiplication follows directly from the meaning of each selected factor. Under the chain ladder projection, cumulative losses at age \(j+1\) are estimated as the losses at age \(j\) multiplied by \(f_j\). Projecting one more interval multiplies that result by \(f_{j+1}\). Thus the estimate at age \(j+2\) is the original amount multiplied by both factors.
Repeating this step through the final modeled age and, if appropriate, the tail gives the product of all remaining factors. This also explains why a factor for an earlier interval must not be applied to an accident year that has already reached a later age: that development has already occurred.
Worked Example: Building a Pattern from Age-Specific Evidence
Worked Example: Building a Pattern from Age-Specific Evidence
An actuary is reviewing a hypothetical paid-loss triangle. The observed link ratios for three accident years are shown below. After reviewing the experience for each interval, the actuary selects factors of 1.80, 1.30, 1.12, and 1.05 for the four modeled intervals, and a tail factor of 1.02. For a new accident year with cumulative paid losses of $4,200,000 at 36 months, calculate the selected factor from 36 months to ultimate and the projected ultimate. Use the selections as given; amounts are in dollars and round the final dollar amounts to cents.
| Interval | Observed link ratios | Selected factor |
|---|---|---|
| 12–24 months | 1.78, 1.84, 1.79 | 1.80 |
| 24–36 months | 1.28, 1.31, 1.29 | 1.30 |
| 36–48 months | 1.10, 1.13, 1.12 | 1.12 |
| 48–60 months | 1.04, 1.06, 1.05 | 1.05 |
| Beyond 60 months | Tail selection | 1.02 |
Step 1: Identify the remaining factors. At 36 months, the 12–24 and 24–36 intervals are already past. The remaining modeled factors are 36–48 and 48–60 months, followed by the tail.
Step 2: Chain the factors from 36 months.
Step 3: Project ultimate.
The result is checked by applying each factor in sequence: $4,200,000 × 1.12 = $4,704,000; $4,704,000 × 1.05 = $4,939,200; and $4,939,200 × 1.02 = $5,037,984. The product and sequential calculations agree. The common mistake avoided is including the earlier 12–24 and 24–36 factors even though the loss is already at 36 months.
Applying the Pattern at Different Ages
Different accident years in the same triangle may be at different development ages. To project each one, start at its current age and multiply only the factors for intervals that remain. The factor selection is common to the pattern, but the applicable product depends on the accident year’s maturity.
Worked Example: Projecting from 12 Months with a Tail
A hypothetical liability portfolio has selected factors of 1.65 for 12–24 months, 1.25 for 24–36 months, and 1.08 for 36–48 months. The selected 48–60 factor is 1.00, and the tail factor beyond 60 months is 1.03. A recent accident year has $2,400,000 of cumulative paid losses at 12 months. Calculate the complete 12-month-to-ultimate multiplier and projected ultimate. Round the final amount to cents.
Step 1: List every remaining factor in age order. The projection starts at 12 months, so it includes the 12–24, 24–36, 36–48, and 48–60 factors, followed by the tail.
Step 2: Multiply sequentially.
The 12-month-to-ultimate multiplier is 2.294325. The 48–60 factor is 1.00, so it does not change the product, but it remains part of the specified pattern.
Step 3: Apply the multiplier.
The sequential check gives $2,400,000 × 1.65 = $3,960,000; × 1.25 = $4,950,000; × 1.08 = $5,346,000; × 1.00 = $5,346,000; and × 1.03 = $5,506,380. The calculations agree. The common mistake avoided is stopping at 48 months and omitting both the 48–60 interval and the selected tail.
One Pattern, Different Remaining Multipliers
An age-to-ultimate multiplier is determined by the current development age. It is not a single multiplier attached to every accident year. The following example uses one complete pattern to illustrate the consequence.
Worked Example: Matching the Multiplier to Current Age
For a hypothetical paid-loss portfolio, the selected factors are 1.70 for 12–24 months, 1.24 for 24–36 months, 1.09 for 36–48 months, and 1.04 for 48–60 months. The tail factor is 1.01. Cumulative paid losses are $1,600,000 at 12 months, $4,800,000 at 24 months, and $7,500,000 at 36 months for three different accident years. Calculate each projected ultimate. Work in thousands of dollars and round final amounts to the nearest 0.1 thousand.
Accident year at 12 months. All four modeled factors and the tail remain:
The cumulative paid amount is 1,600 in thousands, so the projected ultimate is \(1{,}600 \times 2.413525088 = 3{,}861.6401408\) thousand, or $3,861.6 thousand to the requested precision.
Accident year at 24 months. The 12–24 factor has already been realized and is not applied again:
The projected ultimate is \(4{,}800 \times 1.41972064 = 6{,}814.659072\) thousand, or $6,814.7 thousand.
Accident year at 36 months. Only the 36–48, 48–60, and tail factors remain:
The projected ultimate is \(7{,}500 \times 1.144936 = 8{,}587.02\) thousand, or $8,587.0 thousand.
The age-specific check is that each multiplier starts with the factor for the next interval after the current age: 1.70 at 12 months, 1.24 at 24 months, and 1.09 at 36 months. The common mistake avoided is applying the same full 12-month-to-ultimate multiplier to all three accident years, which would count development already reflected in the older cumulative amounts a second time.
Selection Judgment and Assumptions
A selected factor is an actuarial choice informed by experience, not a guarantee that every future accident year will develop exactly as the pattern predicts. In selecting each interval factor, the actuary should consider whether the observed ratios are based on comparable cumulative losses and whether the experience is relevant to the projected business. The evidence may differ in quantity and reliability by age; an interval with limited observations requires particular care.
Chain ladder use also depends on assumptions. The selected pattern is assumed to be reasonably applicable to the accident years being projected. The loss definition, valuation basis, and development-age definitions should be consistent across the triangle. Material changes in claim handling, coverage, settlement practice, or the composition of the business can make historical development less representative. A factor should not be treated as mechanically valid merely because it can be calculated.
Factors are often greater than 1 when cumulative losses continue to develop upward, but that is not a mathematical requirement. Recoveries, corrections, or other changes to cumulative values can produce a factor below 1. A selection must be interpreted in the context of the loss measure and the data, rather than altered just to fit an assumed direction.
Check Your Understanding
For each problem, use the specified selected factors and identify only the factors that remain from the stated current age.
- A portfolio has selected factors of 1.50 for 12–24 months, 1.20 for 24–36 months, and 1.05 for 36–48 months. A hypothetical accident year has cumulative losses of $800,000 at 12 months. What is its projected 48-month amount? (Answer: $1,512,000.)
- A loss amount is $2,500 thousand at 24 months. The selected factors are 1.16 for 24–36 months, 1.07 for 36–48 months, and a 1.02 tail factor. Calculate the projected ultimate in thousands. (Answer: $3,165.06 thousand.)
- A hypothetical accident year has cumulative losses of $9,000,000 at 24 months. Selected factors are 1.40 for 12–24 months, 1.15 for 24–36 months, 1.06 for 36–48 months, and a 1.03 tail. What is the projected ultimate? (Answer: $11,300,130.)
- A candidate applies factors 1.60, 1.25, 1.08, and a 1.02 tail to a loss already at 24 months, where the factors correspond respectively to 12–24, 24–36, 36–48, and the tail beyond 48 months. For cumulative losses of $3,400,000, identify the incorrect step and calculate the correct ultimate. (Answer: Omit the already completed 12–24 factor; the correct ultimate is $4,681,800.)
- A pattern has selected factors of 1.75 for 12–24 months, 1.28 for 24–36 months, 1.10 for 36–48 months, 1.04 for 48–60 months, and a 1.02 tail. One accident year has $2,000,000 at 12 months and another has $5,000,000 at 36 months. Calculate both projected ultimates and their difference. (Answer: $5,227,622.40 and $5,834,400.00; the difference is $606,777.60.)