Case Reserve Adequacy Changes in Incurred Chain Ladder

Why Case Reserve Adequacy Matters

Incurred chain ladder projects future development from observed incurred losses. An incurred amount combines paid losses to date with the case reserves currently carried for reported claims. It therefore reflects both claim experience and the insurer’s estimates of how much those claims will ultimately cost. If the adequacy of case reserves changes, incurred amounts can change even when the underlying claims have not.

That matters because chain ladder relies on development patterns being sufficiently stable across accident years. A change in reserving practice can make recent incurred amounts higher or lower relative to historical amounts at the same development age. If the selected factors do not represent the current reserve basis, projected ultimate losses may be biased. This tutorial is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.

Definition: Let \(C_{i,j}\) denote cumulative incurred losses for accident year \(i\), evaluated at development age \(j\). Incurred losses are the sum of cumulative paid losses and case outstanding reserves at that evaluation date. Let \(U_i\) denote the ultimate losses for accident year \(i\). An age-to-age incurred link ratio is \(C_{i,j+1}/C_{i,j}\), when \(C_{i,j}\) is nonzero. A selected factor \(f_j\) estimates development from age \(j\) to age \(j+1\).

How Adequacy Changes Affect Incurred Development

Case reserve adequacy describes how closely the case reserves carried for reported claims reflect the ultimate cost expected for those claims. If an insurer strengthens case reserves, it increases those estimates; if it weakens them, it decreases them. Such changes can affect incurred losses without any corresponding change to paid losses at that evaluation date.

Suppose a case reserve adjustment changes incurred losses at age \(j\) by \(\Delta_j\) and at age \(j+1\) by \(\Delta_{j+1}\). The resulting link ratio is

f_j^{\text{adjusted}}=\frac{C_{i,j+1}+\Delta_{j+1}}{C_{i,j}+\Delta_j}.

This equation shows why it is not enough to say that reserve strengthening always raises or always lowers a link ratio. The effect depends on how much the reserve basis changes at each age. If incurred losses are increased at the earlier age but not at the later age, the denominator rises and the ratio falls. If they are increased at the later age but not the earlier age, the numerator rises and the ratio increases.

There is an important special case. If every incurred amount is multiplied by the same positive proportion \(q\), then the link ratio is unchanged:

\frac{qC_{i,j+1}}{qC_{i,j}}=\frac{C_{i,j+1}}{C_{i,j}}.

A proportional change applied consistently across development ages and accident years cancels in the ratio. By contrast, an additive change, or a proportional change that varies by age or accident year, generally does not cancel.

Why the Chain Ladder Projection Can Be Biased

For a simple chain ladder projection, the estimated amount at the next age is the current incurred amount multiplied by the selected age-to-age factor. A sequence of selected factors can be multiplied to estimate development to a later age or to ultimate. These calculations use a model assumption: the selected development pattern is appropriate for the accident year being projected.

The case reserve adequacy assumption behind an incurred chain ladder pattern is not that reserves must be perfect. Rather, historical and projected incurred amounts must be sufficiently comparable for their observed development to be informative about future development. If recent claims are reserved more adequately at an early age than the historical claims used to select factors, the recent incurred amount may already include more of its eventual cost. Applying a historical factor that assumes a lower early-age case reserve level can overstate the projected ultimate. If recent reserves are less adequate, the early incurred amount may be too low relative to the historical basis, creating a risk of understatement.

The direction is a risk, not an automatic conclusion from the words “strengthening” or “weakening.” The factor can be affected by reserve adjustments at both the beginning and end of the development interval, and the selected factors may include years with different reserving practices. The projection should therefore be interpreted in light of the actual adjustment pattern and the data used to select factors.

Derivation: Under the chain ladder projection at age \(j\), the projected amount at age \(j+1\) is \(C_{i,j}f_j\). If the appropriate factor for the projected cohort is \(f_j^*\), then the projection error is \(C_{i,j}(f_j-f_j^*)\). For \(C_{i,j}>0\), a selected factor larger than the appropriate factor produces an overprojection at the next age; a smaller factor produces an underprojection. The same logic applies to a cumulative product of factors used to project to ultimate.

This is the central diagnostic: determine whether the factor selected from historical experience is appropriate for the current cohort’s case reserve basis. A factor may be calculated correctly from the data and still be unsuitable for a cohort whose reserve adequacy has changed.

Worked Example: Strengthening Early Case Reserves

A hypothetical insurer observes the following incurred development for an older accident year: \(C_{12}=500\) at 12 months and \(C_{24}=800\) at 24 months. The selected 12-to-24-month factor is 1.60. For this illustration, the selected 24-month-to-ultimate factor is 1.25. A newer accident year has 12-month cumulative paid losses of 200 and case reserves of 500. Its eventual 24-month incurred amount is 800, and its ultimate is 1,000. Calculate the chain ladder projection using the older pattern and assess the impact of strengthened early case reserves.

Method. Calculate the older 12-to-24-month factor, apply it to the newer year’s incurred amount, and then apply the 24-month-to-ultimate factor.

First, the older year’s 12-to-24-month factor is

f_{12,24}=\frac{800}{500}=1.60.

The newer year’s 12-month incurred amount is paid losses plus case reserves:

C_{12}=200+500=700.

Applying the historical factor projects 24-month incurred losses of

$$ 700(1.60)=1{,}120. $$

The projected ultimate is therefore

$$ 1{,}120(1.25)=1{,}400. $$

The stated ultimate is 1,000, so this projection is 400 higher. The newer year’s actual 12-to-24-month factor is \(800/700=1.1429\), below the historical factor of 1.60. The higher case reserves at 12 months mean the newer incurred amount already includes more estimated claim cost at that age. Common mistake: applying 1.60 and treating the result as a reasonable forecast without checking whether the early-age incurred basis is comparable. Here it projects 1,120 at 24 months even though the eventual 24-month incurred amount is 800.

Worked Example: Weakening Case Reserves

Two historical accident years have incurred amounts of \(500\) and \(600\) at 12 months, and \(800\) and \(900\) at 24 months, respectively. Use the volume-weighted 12-to-24-month factor. A newer year has 12-month paid losses of 300 and case reserves of 100. Its eventual ultimate is 1,000. For this simplified example, the historical pattern’s 24-month-to-ultimate factor is 1.25. Calculate the projected ultimate and its error.

Method. The volume-weighted link ratio is total 24-month incurred divided by total 12-month incurred. Apply it to the newer year, then apply the stated 24-month-to-ultimate factor.

The selected factor is

f_{12,24}=\frac{800+900}{500+600} =\frac{1{,}700}{1{,}100} =1.5455.

The newer year’s 12-month incurred amount is

C_{12}=300+100=400.

The projected 24-month amount is

$$ 400(1.5455)=618.20. $$

Applying the 24-month-to-ultimate factor gives

$$ 618.20(1.25)=772.75. $$

The estimate is \(1{,}000-772.75=227.25\) below the stated ultimate. The newer year’s low case reserves make its early incurred amount low relative to the historical basis, so a factor selected from the older years does not develop it far enough. Common mistake: assuming that a lower starting incurred amount necessarily means lower ultimate losses. In this example the ultimate is 1,000; it is the changed reserve basis, not a lower ultimate, that creates the mismatch.

Worked Example: A Reserve Change at Both Ages

In a hypothetical portfolio, a representative cohort would have incurred losses of 600 at 12 months and 900 at 24 months under its original case reserve basis. A policy change adds 120 to case reserves at 12 months and 30 at 24 months. A new cohort with the same underlying development is subject to the same policy change. Calculate the adjusted link ratio and show whether applying the unadjusted factor would be appropriate for this new cohort.

Method. Add the reserve adjustments separately at each age, then calculate the adjusted and unadjusted link ratios.

The unadjusted factor is

f_{12,24}=\frac{900}{600}=1.50.

The adjusted 12-month amount is \(600+120=720\), and the adjusted 24-month amount is \(900+30=930\). Thus, the factor under the new reserve basis is

f_{12,24}^{\text{adjusted}}=\frac{930}{720}=1.2917.

For the new cohort, applying the adjusted factor projects \(720(1.2917)\), which is approximately 930. That matches the amount generated by the stated underlying development and reserve adjustments. Applying the old factor instead gives \(720(1.50)=1{,}080\), which is 150 too high at 24 months. Common mistake: using the original factor simply because the underlying claims are assumed to have the same development. The case reserve adjustment is not the same at the two ages, so the incurred link ratio changes.

Practical Diagnosis and Common Traps

A change in case reserve adequacy is a potential structural change in incurred data, not merely an unusual individual link ratio. When investigating it, first establish which accident years and development ages were affected. Then assess whether reserve adjustments changed the numerator, denominator, or both for each link ratio. Finally, consider whether the selected factors and the projected cohort reflect comparable reserve practices.

Separate incurred components.
For the relevant development ages, identify cumulative paid losses and case reserves. This helps distinguish changes in payment emergence from changes in claim estimates.

Locate the adequacy change.
Determine which accident years and evaluation ages experienced strengthening or weakening. A change affecting only one age can influence a link ratio differently from a change applied consistently at all ages.

Reassess factor comparability.
Compare the reserve basis in the experience used for factor selection with the basis in the accident year being projected. A mechanically calculated historical average does not by itself establish comparability.

State the direction and its condition.
Explain whether the current incurred amount is relatively high or low, which factor is likely to be too high or too low, and why. Avoid asserting a bias direction without identifying the affected ages.

A frequent exam trap is to say that “strengthening increases incurred losses, so it increases the factor.” That conclusion is not generally valid. A reserve increase at the later age tends to raise the numerator and the link ratio; the same increase at the earlier age tends to raise the denominator and lower the ratio. If both ages change, calculate the ratio using both adjusted amounts.

A second trap is to confuse a change in the estimate of reported-claim costs with a change in the ultimate losses themselves. A reserve strengthening can raise current incurred amounts without creating new claims or changing their eventual cost. The chain ladder bias arises when the resulting observed pattern is applied as if it were representative of another reserve basis.

A third trap is to treat every strengthening as an additive change. If a single multiplier applies to every observation at both ages, it cancels in the link ratio. But a percentage increase that differs by age or accident year does not generally cancel. The details of how the policy changed matter.

Key takeaway: Incurred chain ladder projects a development pattern in which case reserves are part of the observed amounts. A change in case reserve adequacy can therefore distort age-to-age factors and bias ultimate estimates when the historical and projected cohorts are not on a comparable reserve basis. Diagnose the affected ages and calculate the resulting ratios before stating the likely direction of bias.

Check Your Understanding

For each problem, identify the relevant reserve-basis effect as well as the requested calculation.

A historical cohort has incurred losses of 400 at 12 months and 640 at 24 months. A newer cohort has 12-month paid losses of 180 and case reserves of 320. Using the historical factor, what is the projected 24-month incurred amount? (Answer: 800)
A historical incurred link ratio is 1.40. For a new cohort, the incurred amount at 12 months is 500, and its reserve basis makes the appropriate 12-to-24-month factor 1.20. What does the historical factor project at 24 months, and by how much is it higher than a projection using the appropriate factor? (Answer: 700; it is 100 higher than the appropriate projection of 600)
A cohort’s original incurred amounts are 800 at 12 months and 1,200 at 24 months. A reserve change adds 160 at 12 months and 60 at 24 months. Calculate the original and adjusted link ratios. (Answer: Original factor = 1.5000; adjusted factor = 1,260/960 = 1.3125)
Historical cohorts have incurred amounts of 600 at 12 months and 900 at 24 months. A newer cohort has 12-month incurred losses of 720 because its early case reserves are stronger. Its appropriate 12-to-24-month factor is 1.25. If the historical factor is applied, what is the projected 24-month amount, and what is the difference from the appropriate projection? (Answer: Historical projection = 1,080; appropriate projection = 900; overprojection = 180)
An insurer multiplies incurred amounts at both 12 and 24 months by 1.10 for every accident year. An original 12-to-24-month factor is 1.50. What is the factor after this proportional change, and why? (Answer: 1.50; the common multiplier cancels, since \((1.10C_{24})/(1.10C_{12})=C_{24}/C_{12}\))

What You'll Learn