Chain Ladder

The Chain Ladder (or Loss Development) method is a standard actuarial technique used to project historical claims experience to ultimate levels.

---

Key Assumptions

1. Stable Development Patterns: The relative change in cumulative claims from one evaluation age to the next is stable and predictable.
2. Historical Continuity: The historical patterns observed in the development triangle will continue into the future.
3. Independence of Accident Years: The development of one accident year is independent of the development of other accident years.

---

Mathematical Formulation

Let $C_{i,j}$ denote the cumulative claims for accident year $i$ at development age $j$.

1. Age-to-Age Factors (Link Ratios)

Age-to-age factors (LDFs) measure the growth rate of cumulative claims from age $j$ to age $j+1$:

$$f_j = \frac{\sum_i C_{i, j+1}}{\sum_i C_{i, j}}$$

Actuaries select these factors using weighted averages (e.g., volume-weighted, simple average, or medial average of historical ratios).

2. Cumulative Development Factors (CDF)

The CDF projects cumulative losses from age $j$ to ultimate maturity ($u$):

$$\text{CDF}_j = \prod_{k=j}^{u-1} f_k$$

If there is outstanding development beyond the oldest age in the triangle, a tail factor ($f_{\text{tail}}$) is applied: $\text{CDF}_j = \left( \prod f_k \right) \times f_{\text{tail}}$.

3. Ultimate Losses and Unpaid Claim Estimates

The ultimate losses for accident year $i$ evaluated at current age $t$ are:

$$\text{Ultimate Losses}_i = C_{i, t} \times \text{CDF}_t$$

The Total Unpaid Claims (including case reserves and IBNR) are:

$$\text{Unpaid Claims}_i = \text{Ultimate Losses}_i - \text{Paid Losses}_i$$ $$\text{IBNR}_i = \text{Ultimate Losses}_i - \text{Reported Losses}_i$$

---

Limitations and Sensitivities

• Immature Years: Projections for the most recent (immature) accident years are highly sensitive to small changes in early LDFs, leading to high volatility.
• Operational Shifts: The method is severely distorted by changes in case reserving practices (causing reported LDF shifts) or claim settlement speeds (causing paid LDF shifts). For such shifts, adjustments like the Berquist-Sherman technique must be utilized.
• Atypical Events: A single large loss in a cell can distort the selected averages.