Bornhuetter-Ferguson Method

Bornhuetter-Ferguson starts with an expected ultimate loss and then gives credit only for the part of that loss that should already have emerged. It is less reactive than chain-ladder because it does not let thin early development fully control the reserve.

That is why actuaries often like it for immature years. When observed development is still noisy, an expected-loss anchor can be more stable than projecting everything from sparse emergence alone.

Suppose the expected ultimate loss for an accident year is 1,200 and the selected percentage reported is 60%. If reported loss to date is 700, then the Bornhuetter-Ferguson reserve is 1,200 x (1 - 0.60) = 480 and the resulting ultimate estimate is 700 + 480 = 1,180.

Notice what happened: the method did not project the 700 all the way up through a full development chain. It used the expected-loss view for the not-yet-emerged portion instead.

Exam 5 and ASTAM both include Bornhuetter-Ferguson because it is one of the standard ways to stabilize reserving work when development data is immature or volatile.

In practice, it is often compared directly against chain-ladder. If chain-ladder is the pure development view, Bornhuetter-Ferguson is the blended expected-loss view.

A common mistake is forgetting that the method depends on two separate judgments: the expected ultimate and the percentage reported. If either one is poor, the reserve can still be poor even though the method looks more stable than chain-ladder.

Another mistake is assuming the method is always better for immature years. It is often more stable, but stability is only valuable if the expected-loss assumption is credible.

Bornhuetter-Ferguson is useful to think of as a shrinkage idea in reserving. Instead of trusting sparse observed emergence completely, it pulls the estimate toward an external expected-loss view for the not-yet-developed portion.