Actuaries have all the fun. It seems that way sometimes. But that may be because sometimes I am not sure what they are saying.
"Predictive analytics" is a formula used to model claims, in basic and simple terms. This was the subject of a fine presentation at the Casualty Actuarial Society Annual Meeting in Orlando.
The CAS Annual Meeting continues with more fine presentations today, Wednesday, November 16, 2016.
One of the presenters said that his company did not have any negative reaction to report from regulators in his company's modelling of claims. That drove me to think about our own experience in this State when predictions were introduced into rate filings to replace or supplant actual results from the real world, so to speak.
In Florida, we have a State-run computer model to predict catastrophe claims based on the data of actual experience. Our statute was recently amended to allow insurance companies to use their own computer models including models based on predictions and assumptions, but not necessarily based on any real-world results. The Florida experience is pretty well summarized in this April 16, 2012 article, posted here.
"Predictive analytics," like any tool, may be fine in one area such as predicting payouts on claims and predicting which claims are likely to hit bigger than the claims department might have thought at first, but not so accurate in other areas such as predicting premium rates. These trends have been building for years. See, for example, these issues addressed in these articles from 2012, about automobile liability insurance premiums here, and about "evidence-based funding criteria" especially in setting premiums for all lines of insurance here.
The actuarial presenters of CAS in Orlando made a compelling case for learning more about "predictive analytics" in modelling claims.
Please Read The Disclaimer. ©2016 by Dennis J. Wall. All Rights Reserved.
Comments