The first discipline is diagnostic, not mechanical. Before selecting development factors and applying any method, the actuary must assess whether the historical data reflects conditions similar enough to the current environment to be predictive. This is harder than it sounds.
Loss triangles can show apparent stability even as the underlying drivers of loss are shifting. Frequency may be declining due to underwriting actions while severity is increasing due to litigation trends. When these effects offset each other in aggregate, the triangle looks well-behaved. The actuary who relies on that appearance without disaggregating the components will underestimate the true uncertainty in the reserve.
Practical diagnostic steps include: reviewing closed-to-open claim ratios by accident year, examining settlement patterns for signs of elongation, and comparing paid-to-incurred ratios across recent years. A sustained increase in the ratio of open claims to written premium is an early warning signal that current development factors may be underweighted for severity.
"A reserve range that cannot survive regulatory scrutiny is not a range. It is a false sense of precision with extra steps."
When development patterns are unstable, relying exclusively on the volume-weighted average development method is a disservice to the client. The appointed actuary's obligation is to consider multiple methods and explain, in writing, why the selected method is most appropriate for the data at hand.
The Bornhuetter-Ferguson method offers a useful complement in volatile environments: by blending the development method with an a priori loss ratio, it reduces the weight placed on recent immature experience. The trade-off is sensitivity to the a priori assumption, which must itself be justified and documented. Using the Cape Cod method to derive the implied a priori from the data is one way to ground that assumption empirically.
For long-tail lines such as medical professional liability, general liability, and workers' compensation, frequency-severity separation deserves more attention than it typically receives. When severity is being driven by factors external to the insurance relationship, such as medical cost inflation, pharmaceutical pricing, or third-party litigation funding, a combined development model conflates two distinct trends that may require different projections.
The requirement to present a range of reasonable estimates has become more meaningful as regulators have grown more sophisticated in evaluating reserve adequacy. A range constructed by applying a uniform percentage above and below the central estimate will not satisfy an informed reviewer. Ranges should reflect the actual drivers of variability in the data.
A structured approach to range development begins with identifying the key assumptions that drive the reserve estimate: the selected loss development factors, the a priori loss ratio, the loss trend assumption, and the weighting between methods. Sensitivity testing each of these assumptions, individually and in combination, produces a range that is derived from the data rather than asserted by convention.
In a regulatory examination or litigation context, the documentation of the reserve analysis will be scrutinized as carefully as the numbers themselves. The narrative must explain not just what was done, but why: why certain years were excluded from factor selections, why one method was given more weight than another, and why the a priori loss ratio reflects the company's current underwriting posture.
We advise clients to treat the written narrative as a communication to a skeptical reader who was not present for the analysis. If a regulatory examiner or opposing expert cannot follow the reasoning from the data through the methodology to the conclusion, the reserve analysis will not hold up when it matters most.
For actuaries and risk managers working in volatile environments, we offer the following principles:
First, invest time in the diagnostic phase before reaching for the development factors. The quality of the estimate depends on the quality of the diagnosis. Second, present multiple methods and explain the weight given to each. A single-method analysis is incomplete, regardless of how well-documented the method is. Third, construct reserve ranges from first principles. The range should reflect the actual uncertainty in the data, not a conventional margin. Fourth, write the narrative as if explaining to someone who will challenge every assumption. Actuarial defensibility is not achieved through hedging language; it is achieved through transparent reasoning.
Reserve analysis in a volatile loss environment requires more judgment, more documentation, and more explicit treatment of uncertainty than stable environments demand. The actuaries who provide the most value in these conditions are those who can communicate that uncertainty clearly, without conveying a lack of confidence in the analysis itself.
