By Sam Gutterman
I have often thought about the proper role of judgment in actuarial work. At times it seems inconsistent with the scientific method. But, contrary to some views, science is fundamentally based on observations, rather than universal truth. Nonetheless, why do we as actuaries practice under the banner of actuarial science?
Young actuarial students may be trained to view the world as a predictable place, albeit subject to a bunch of stochastic processes—if you’ve done the math correctly, you’ll get the correct result and a unique recommendation!
I have often preached the importance of using models in assessing risk and uncertainty. But models can be messy, and the results can come with lots of caveats. Models are simply representations of reality, and in some cases only representations of past reality, even when “confirmed” by calibration.
Despite this, complexity, non-linearities, nuance, changes in conditions, and risk aversion can upend modeled results when exposed to the real world. (Alas, it would be great if academic concepts always worked!)
That’s why judgment is involved every step of the way. The identification of the issue/problem studied, the question posed, the factors considered, the type of model applied, and the assumptions/parameters chosen—all involve pivotal decisions that affect the model’s output.
It is crucial to recognize when judgment should be applied. Suppose a single approach or model has been used repeatedly by others—how can it be challenged? Just because it is current “accepted practice” doesn’t mean it is the most appropriate in a given circumstance.
If challenged—and I have been challenged many times by “What’s the basis for this conclusion?”—you have to be ready to support the choices made. If your choices can’t be supported, you may be accused of being unprofessional, subjective, and/or biased. Nevertheless, judgment is applied at every step—each intermediate and final result is influenced by a previous judgment in some way. I personally feel more comfortable using well-thought-out reasoning than unthinking averages. An actuary needs to understand and take responsibility for the judgments applied, with appropriate recognition of and transparency in the basis for conclusions and how they are reached.
Just because an approach has been used for the past 30 years doesn’t mean it remains appropriate today, or will in the future. Past trends are indeed past trends—they need not continue. What about loss development in a P&C reserve analysis when there is new claims management or a surge in what may be a new type of claims—or should all outliers be ignored? I could go on—importantly, there may be possible answers to the question of why. Always remain skeptical, even of your own assumptions.
Even a seemingly relevant and reliable source shouldn’t always be accepted without scrutiny for its applicability to the future. Statistical correlation of one period may not apply to the next due to changes in drivers or lack of causal relations.
What is professional judgment? Of course, an actuary must perform professional services with integrity consistent with ethical standards, objectivity, skill, and care. Judgment can sweep in a lot in this context—lessons learned from prior experience (especially when past estimates or judgments proved wrong). Reasonableness of data or data patterns and intermediate values, as well as obsessing on how conditions have changed or are expected to change; these are all considerations subject to actuarial smell tests—that is, does it feel right?
An example I have been involved with is the mortality improvement assumption that many actuaries use. It is easy to take the average of the past 50 years of experience or even the past five years. But why use such an average? Why factor in the effects of the introduction of antibiotics when there is a concern about antibiotic resistance? Why reflect the effects of adverse smoking experience when the future will see the favorable effects of the reduction in smoking prevalence? How to reflect the opioid and obesity epidemics? We experienced significant improvements in mortality in the first decade of this century and the last half of the last century, while very recently mortality has stalled or been moving in the opposite direction. Some experts have applied linear extrapolation—which to me makes little sense over a long period.
Experience and sophisticated modeling are important, and thoughtful extrapolation of past trends has been our bread-and-butter—but actuarial judgment will always be at the center of our work.
SAM GUTTERMAN, a member of the Academy and a fellow of the Society of Actuaries and the Casualty Actuarial Society, lives in Glencoe, Ill.