Loss Reserves Challenge
By Stephen W. Philbrick

In an earlier Brainstorms column (August 2004), I discussed Mary Frances Miller's challenge to the profession to improve the actuarial science behind loss reserves. I supplemented her challenge with one of my own - how better to align incentives with goals. Loss reserves are highly uncertain, and reasonable people can reach plausible conclusions that have profoundly different financial implications. It is important to not only have the tools to derive the best possible estimates, but a process to encourage the booking of the best possible estimates, as well.

Arthur Schwartz followed up with a suggestion on how to improve the process:
There is actually a fairly basic approach that would remove much of the second-guessing about what reserves should be. It's based on the model adopted by life insurance actuaries decades ago. Their method is that an entire block of business is reserved for using a standard mortality or disability table. If there's an exception—an outbreak of some unusual disease for example—the exception is easily noted and highlighted.

The property casualty industry can do likewise. Consider an industry task force that set a reasonable set of ceilings and floors expressed as a ratio to premium. The ceilings and floors would vary by line of insurance, separately for each year shown on Schedule P. Insurers could exceed the caps (reserve higher or lower) but management (not the actuary) would have to explain why…

As a disclaimer, the above represents my personal opinion only.

I'd like to expand on his idea.

Arthur's idea essentially creates two ranges—one could be called a "safe haven"—a range established by an independent group, such that any loss reserve estimates within the range would be established in the ordinary way. If the company wishes to book something outside this range the company management must provide a reason. On one hand, I would like to ensure that a reserve outside the range is not viewed as a black mark, but a result requiring additional documentation. On the other hand, I want the burden of falling outside the range to be meaningful, not simply a letter supplied by management. One option is to outline the types of analysis necessary to support a value outside the range-items which might go beyond that done in a normal reserve review. For example, if a company asserts that its payment pattern is speeding up as an explanation for a relatively low reserve, it might be required to support that with a detailed claim audit.

It is easy to identify downsides, but one can usually solve problems if the benefits are sufficient.

Arthur suggested that the "safe haven" range would have both ceilings and floors. I would want to ensure that the additional burden of reserves outside the range is asymmetrical—it should require more of a burden to deviate down than to deviate up.

One obvious issue is whether the ranges, presumably by line of business, should be the same for every company. My initial reaction is that companies with an established track record of results better (or worse) than the industry averages should have their range modified to reflect this reality. Startups could be problematic but a study of past startups could shed light, possibly indicating an upward adjustment to the range.

A fellow actuary suggested that companies should have to file the results of standard loss reserving approaches. For example, they might have to calculate an ordinary chain ladder for each line, as well as a Bornhuetter-Ferguson (where the initial expected loss ratios come from an industry source). The company would still file their best estimates using whichever methods they feel are most appropriate, but deviations from the standard approaches might require some level of explanation. Perhaps the nature of the addition documentation would increase as the filed results differ from the standard method results.

Each of these ideas does contribute some measure of incentive. In a case where the actuary has a wide range of plausible alternatives, it may be helpful that selecting estimates that are out of line with industry or one's own experience carries an additional burden in terms of explanations and management actions. Of course, this approach is no panacea. If the actuary's best estimates are above the industry ranges, there may even be a perverse incentive to consider more optimistic selections. Finally, the details of this industry task force haven't been addressed, and it is likely to be a formidable project. It is easy to identify downsides, but one can usually solve problems if the benefits are sufficient. Are they?