In publications and websites devoted to risk, there is a distinct sub-genre of articles that can be called “Predicting Mega-Disasters.” A mega-disaster is a natural catastrophe that wreaks unprecedented destruction — either by an event of never-before-seen force, or a severe event that impacts an increasingly vulnerable population. Here are just three recent examples of these mega-disaster prediction articles:
What should we make of these predictions?
Models are wrong...
Modeling of all natural catastrophe risk encompasses timelines longer than we can comprehend. Some flood models claim to model return periods of 1,000 years, which is vastly longer than our most ancient river records. Earthquakes and volcanoes operate on a geological time scale, on which modern civilization itself is but a moment. Convective storms, with their tornadoes and hail, are random in their regular annual patterns and unknown by long historical views. As this blog has emphasized many times, all these models are wrong, and it is at the extremes of the model where they are most wrong (i.e., in predicting mega-disasters beyond our collective experience).
... but useful.
Even though the models — and almost all predictions — for mega-disasters are wrong, they serve a definite useful purpose. Unforeseen mega-disasters must be accounted for by insurers, and even more so by reinsurers, because these are the types of events that could bankrupt any one of them if their capacity can’t handle the losses. In their planning, they must always keep a wary eye on the distant horizon, watching for such a catastrophe.
Research is critical.
But if the models are wrong, how can insurers be prepared? Reinsurers have historically invested in the necessary scientific research to understand our dynamic world as much as possible (check out this article). And they continue to do so today, with all global reinsurers maintaining a staff of scientists who study the changing nature of our world: both the natural and man-made aspects. Insurers have also long invested in study and research to understand potential loss-drivers. The understanding extracted from this research finds its way into their business as a small amount of uncertainty to be covered by rates and business rules.
So let the predictions continue.
They raise valuable awareness of potentially devastating events, and they frequently drive constructive efforts to mitigate possible loss of life and property. Just because they’re wrong doesn’t make them valueless. In fact, they are never wrong, as long as they assign a probability to the prediction between 1% and 99%, because you can’t demonstrate a probability is incorrect after the fact.
