Seen in Insurance Journal:
S& P said “models can be “used or misused,” if they’re not accompanied by “pricing discipline, a sound strategy, and an understanding of the market.” Panelist Stephen L. Way, former CEO of HCC Insurance Holdings Inc. and founding partner of SLW International LLC, noted that “discipline may be more important than all the modeling in the world.” The best model being “your own experience.”
That quote was taken from an IJ article on a recent S&P conference, cautioning on the perils of overreliance on modeling.
I have to admit, I love models. Big, complex algorithms that make sense out of the chaos of data to find new patterns and build better predictive mousetraps are the sorts of things that set might heart a-fluttering. (Of course, I’m strange. But you knew this already.)
However, even I, a model-loving analytic geek, have to admit that there are a lot of models being used in idiotic ways. For example, in a prior life, where my role focused on the use of credit data for insurance purposes, I saw too many competitors that saw credit scores as an easy bolt-on feature to their programs, without considering what adjustments needed to be made to their non-credit variables to avoid “double counting”, or without considering how to implement without alienating consumers.
While the changes in market dynamics largely have insulated insurers from the effects of the double-dipping effects long enough for rating relativities to be recalibrated, it was hard not to notice the consumer and regulatory backlash to one insurer’s idiotically naieve decision to nonrenew the bottom-scoring quintile of their book.
Modeling is a tool providing additional information to consider in rating and underwriting a risk. There’s great potential to using modeling techniques to get your hands around challenging business.
However, modeling is not a license to throw common sense out the window.
(And before someone weary from property insurance rate hikes on the coast in the Southeast misuses the preceding sentence, “common sense” still isn’t a reason to turn a blind eye towards a model’s message when you don’t like what that message says.)