Loss Reserving Methods: Chain Ladder and Bornhuetter-Ferguson

Chain Ladder Method

The chain ladder (development) method uses historical loss development patterns to project ultimate losses. From a loss triangle, development factors are calculated at each maturity: f_j = sum of losses at age j+1 divided by sum of losses at age j. Cumulative development factors (CDFs) project immature losses to ultimate. The reserve for each accident year equals the current reported loss times (CDF minus 1). This method assumes past development patterns will continue and works best with stable, high-volume portfolios.

The loss triangle is the foundational data structure. Rows represent accident years (when the loss-causing event occurred), and columns represent development periods (how long after the accident year the losses are evaluated). Each cell contains cumulative reported losses for that accident year at that development age. The triangle shape arises because more recent accident years have fewer development periods observed.

Computing Development Factors

Age-to-age development factors (also called link ratios) measure how losses grow from one development period to the next. The weighted average method computes f_j = (sum of all losses at age j+1) / (sum of all losses at age j), weighting each accident year by its volume. Alternative selection methods include the simple average, the medial average (excluding high and low), and the geometric average.

For example, suppose at development ages 12 to 24 months, three accident years show losses developing from (1000, 1200, 900) to (1300, 1500, 1170). The weighted average factor is (1300+1500+1170)/(1000+1200+900) = 3970/3100 = 1.2806. The cumulative development factor (CDF) from age j to ultimate is the product of all subsequent age-to-age factors: CDF_j = f_j * f_{j+1} * ... * f_{ultimate}. The tail factor beyond the oldest observed development period requires judgment or curve fitting.

Applying Chain Ladder Reserves

Once CDFs are determined, projected ultimate losses for each accident year equal the current reported losses multiplied by the CDF from the current age to ultimate. The reserve (IBNR, or Incurred But Not Reported) for each accident year is the projected ultimate minus the current reported losses, which equals reported losses times (CDF - 1).

Consider accident year 2023, currently at 24 months of development with reported losses of 5,000. If the CDF from 24 months to ultimate is 1.45, the projected ultimate is 5,000 * 1.45 = 7,250 and the reserve is 7,250 - 5,000 = 2,250. The total reserve across all accident years is the sum of individual reserves.

Bornhuetter-Ferguson Method

The Bornhuetter-Ferguson (BF) method blends the chain ladder projection with an a priori expected loss estimate. The BF reserve equals the a priori expected ultimate loss times the proportion of losses still unreported (1 minus 1/CDF). This method is more stable for immature accident years because it does not rely entirely on reported-to-date experience. The BF ultimate equals reported losses plus the BF reserve.

The a priori expected ultimate losses are typically based on the expected loss ratio times earned premium. For instance, if earned premium for accident year 2023 is 10,000 and the expected loss ratio is 65%, the a priori expected ultimate is 6,500. If the CDF from 24 months to ultimate is 1.45, the proportion unreported is 1 - 1/1.45 = 0.3103. The BF reserve is 6,500 * 0.3103 = 2,017, and the BF ultimate is 5,000 + 2,017 = 7,017.

Comparing the Two Methods

The chain ladder and BF methods differ most for immature accident years. The chain ladder method projects entirely from reported experience, so a single large claim in a young accident year can dramatically inflate the ultimate estimate. The BF method tempers this by using the a priori loss ratio for the unreported portion, producing more stable estimates when limited data is available.

For mature accident years (where most losses are already reported and the CDF is close to 1), both methods produce similar results. The proportion unreported is small, so the BF adjustment is minimal and both methods essentially equal the reported losses.

A useful way to understand the relationship is: Chain Ladder Ultimate = Reported * CDF, while BF Ultimate = Reported + Expected Ultimate * (1 - 1/CDF). The chain ladder gives 100% credibility to observed data; the BF gives zero credibility to observed data for the unreported portion, using only the prior expectation.

Common Pitfalls

Students often confuse the CDF with the age-to-age factor. The CDF is the product of all remaining age-to-age factors to ultimate; a single age-to-age factor covers only one development period. Another common error is applying the wrong CDF to an accident year. The CDF applied should correspond to the current development age of that accident year, not to its calendar age.

In BF calculations, a frequent mistake is computing the unreported proportion as (CDF - 1) instead of (1 - 1/CDF). While these are related, they are not the same. The proportion unreported (1 - 1/CDF) is the fraction of expected ultimate losses not yet reported, while (CDF - 1) divided by CDF gives the same result from a different algebraic path. Be consistent in your formula usage.

Study Tips

Exam STAM frequently presents loss triangles and asks you to compare chain ladder and BF reserve estimates. Practice extracting development factors from loss triangles quickly and accurately. Set up your work in a structured table with columns for accident year, current reported losses, development age, CDF, chain ladder ultimate, and BF ultimate. This organization prevents errors and makes it easy to verify your work. Know both methods well enough to compute reserves from scratch given a loss triangle and an expected loss ratio.