By DAN REYNOLDS, managing editor of Risk & Insurance®
In their quest to take greater ownership of catastrophe property modeling and its sister science, predictive analytics, large commercial insurance carriers are investing more resources and in some cases realizing related benefits, such as a much better communication and risk understanding between modelers, analysts and underwriters.
In doing so, according to experts, the commercial insurers are playing catch-up with the personal lines insurers, who, because of the homogeneity of the risks they insure, have been using modeling and predictive analytics much more widely and for a much longer period of time.
"Certainly, on the personal lines side, it has come to the place that if you are not using predictive analytics in a pretty significant way, in auto, for example, then you are probably not writing personal auto," said Dax Craig, CEO of Denver-based Valen Technologies, a provider of predictive analytics and modeling services.
Motivators in the move to greater model ownership by some commercial insurance carriers include in part recent catastrophes that exceeded the expectations of CAT property models.
"If you go back several years and you look at the performance of various models in actual events we are going to find obviously that there are things that were not contemplated that were not in that version of the model," said Erick Nikodem, head of global property for the excess and surplus lines carrier Lexington.
"I think I am safe in saying that they have underestimated certain of those actual events," Nikodem said.
Insurers have also been motivated to reevaluate how they look at and use models by the emergence this year of a new CAT property model, RMS Risklink v 11, that because of the much higher potential losses it projected, and its lack of focus on tornados and floods, which have been so painful of late, sent carriers looking for alternatives, or hybrid approaches.
"The RMS model changes last year produced heavy skepticism and prompted the market to take more responsibility for establishing their views of risk -- views which are only partly informed by commercial catastrophe models," said Sherry Thomas, Head of Catastrophe Management- Americas, Guy Carpenter & Co.
Willis Re reported in June that insurers in the United Kingdom could see increases of 97 percent in their capital requirements for catastrophe exposures under Solvency II rules when making calculations under the new RMS model version 11.
"The very radical model changes ... got the attention of everybody writing property business in catastrophe-exposed areas," said John DeMartini, who heads the Towers Watson CAT risk management practice.
"People said, 'You know what, this model shows a deeper exposure than we thought we had and we need a deeper understanding of this. We can't be solely reliant on a third party for modeling output, we need to own this ourselves," DeMartini said.
"The difference is that insurance companies now tend to rely less on any single model answer and are more willing to justifiably adjust the output to what they believe more accurately reflects their own view of the risk," Thomas said.
"In the last year, we have witnessed stakeholders seeking a wider range of views and rigorously questioning multiple facets of a catastrophe model, such as frequency assumptions, vulnerability relativities, financial model calculations and so on," Thomas said in an e-mailed response to Risk & Insurance®.
Boston-based Liberty Mutual provides a snapshot of an insurer that writes both personal and commercial lines and is using its experience with data to grow the capability to do more modeling in commercial, including property/catastrophe.
"Over the past five years Liberty has started to build a lot of predictive analytics capabilities internally within the commercial space, jumping off on what we learned in the personal market," said Tracy Ryan, executive vice president and commercial markets chief products officer for Liberty Mutual.
The company has a lot of experience in the area of using data in workers' compensation, and Ryan said the company is now building on that experience and using modeling in other commercial lines.
"When we look at property it is another line that is actually quite rich with data," Ryan said.
"We're using a lot of those internal modeling capabilities in the property space. It is just a clear opportunity," she said.
The tornadoes and flooding that caused so much insured loss in 2011 are also moving insurers to reexamine models, which traditionally have had a large focus on hurricane impacts.
"Given what happened last year, if you look at CAT activity last year the industry had some hurricane events but we also had a lot of hail and tornado and flood which are really not modeled into RMS 11," said Tim Rose, senior vice president and chief underwriting officer, national accounts property, for Liberty Mutual.
"I think that this is a further area where we want to develop and explore our abilities to model these types of things," Rose said.
As for tornados and the horrors they have wrought in the past year or so, the industry knows it has got a lot to learn about that risk.
"It's going to take a while to perfect anything when it comes to tornado, I mean the market itself struggles with that, right?" said Tony Mammolite, the head of the global property division for Ironshore.
"A big challenge the industry has is trying to identify areas across the country that are impacted by tornado and hail and trying to understand the frequency and severity of these events," said Neal Zonfrelli, senior vice president and product manager, property, for Liberty Mutual.
As the bigger carriers expand their modeling work, they are adding more human talent in the effort.
"There has been a significant increase in resources looking at data and analytics across our commercial lines and using that to develop more granular guidance for the underwriter in terms of both risk selection and pricing," Liberty Mutual's Ryan said.
Kelly Weber, head of research and development for Zurich in North America, said her department has expanded from three people in 2006 to where it now numbers more than 20 people.
As the department has grown, she said, so has the risk management communication within the company improved.
"I think it is really valuable to have a predictive analytics team in-house where you can build expertise over many years and they can develop more of a partnership with underwriting and claims," she said.
Elaine Roach, the Chartis U.S. and Canada Region Cat Modeling department co-leader, said Lexington formally started a modeling department in 2004 with a manager and two CAT modelers and now has 15 full-time CAT modelers and five part-time CAT modelers for the Lexington and Chartis energy business and for Lexington property exposures in the US and Canada.
The fact that larger carriers are putting more resources into their modeling teams shouldn't cloud the reality that carriers use models as just one important piece of their risk management approach and would never depend on a model solely in making a decision.
"It is one touch point for us in the underwriting process," said Lexington's Nikodem.
"We manage exposures in the US and Canada not just from a model-centric view but specifically from an aggregation standpoint in addition to a control standpoint," Nikodem said.
For regional, or super-regional carriers that might not have modeling teams, Towers Watson's DeMartini said carriers would tend to rely on a similar, hybrid approach, using the output from models created by reinsurance brokers and/or vendors and leveraging that with a mapping tool that shows them where their exposures fit into the risks outlined by the modelers.
"I am actually someone who endorses this hybrid relationship because I think it is effective," DeMartini said.
"What it does is it allows the company to work with the output of the models rather than spending an enormous amount of time and effort generating the output."
July 17, 2012
Copyright 2012© LRP Publications