Insurance Industry Is Rethinking Cat Modeling After Last Year’s Disasters
After Hurricane Harvey swept through Texas last August, it not only left behind nearly 50 inches of rain in some areas and an estimated $125 billion in damage, but questions for the insurance industry in terms of how to approach catastrophe modeling in the future, panelists discussed at the June 2018 S&P Insurance Conference in New York.
“You can run the models all day,” said panelist Steven Kelner, managing director and head of U.S./Canada Analytics at Guy Carpenter & Company LLC. “But when you go down to Houston and look at neighborhoods, there were huge neighborhoods where the whole neighborhood was deemed not flood exposed. And the whole neighborhood but a house or two was flooded, because the maps weren’t accurate; because the data wasn’t up-to-date.”
With this in mind, the industry will need to rethink its approach to cat modeling as it continues learning from recent natural disasters including hurricanes and wildfires, panelists agreed.
“There’s a lot of complexity in getting a cat model to run accurately,” Kelner said. “For example, I can turn [GPS] mapping software on, and it will tell me how long it will take to get from Trenton to Philadelphia at 5 p.m. But the mapping software doesn’t realize that at 5:45 p.m., there’s always a backup. So we need to still have common sense in our day-to-day.”
Panelist John Langione, chief risk officer at QBE North America, agreed, adding that “models are great, but they’re part of the process.”
The challenge for the insurance industry comes in recognizing the importance of using modeling software for pricing, while not becoming overly reliant on the models themselves and looking instead to historical elements to understand exposure and risk, said panelist Bruce Jones, executive vice president and chief risk officer at The Travelers Companies Inc.
“You have this sort of tug-of-war between actual experience and models and trying to make sure that as you’re thinking about it both from a pricing perspective and from a capital perspective, you do have a total view of what’s going on,” he said.
He stated he believes the best way to strike this balance is through collaboration.
“I think you need a collaborative cultural that sort of sits behind the decision-making process to begin with,” he said. “Part of what I think makes a good insurance company work is that you do get different points of view about a risk. And it’s not just the risk, but it’s also the reward. Because we could get concerned about every risk that an insurance company faces and wind up writing no business. Part of the collaboration is balancing the risk side of the equation with the reward side of the equation.”
Although last year was challenging given the number of storms and uncertainty of locations impacted, panelists said these difficulties can prove to be valuable learning experiences in balancing the use of modeling software and careful judgment. Jones pointed to Hurricane Harvey as an example of this learning curve.
“[50 inches of rain] really wasn’t exactly contemplated in any of the real modeling efforts,” he said. “[It] was a bit more than even most of the one-in-100 type things were set at….We do want to go back and make the models a more fair representation of what could happen. I think the actual events offer you some opportunity to begin to learn that.”
Beyond Hurricane Harvey, panelists identified several surprises resulting from last year’s natural disasters, including the unexpected geographical diversity of incidents throughout the country. Langione stated this created unprecedented demand on third parties in some cases.
“We saw third party adjusters and appraisers being paid sometimes twice and three times their normal fees and being told they’re going to the highest bidder,” he said.
As a result, he said QBE North America learned that the management of its third-party vendors in advance rather than post-occurrence was critically important.
“It’s looking at the life cycle of the process of an event in advance that I think reaps rewards later,” he said.
Because last year’s events were so widespread, he added that inevitably, some of QBE North America’s own personnel were impacted as well.
“When we think about the catastrophes, I think it reminds us to think about not only the impact on our cat model and our exposures, but how operationally that affects us,” he said. “When you have wildfires and some of your claim reps have their own houses burnt to the ground, how is that going to impact your response?”
He said this led the company to think more carefully about aligning with its customer base.
“I think it created a sense of great humanity amongst our own people about how we treated customers,” he said.
This meant challenging its claim and underwriting teams to identify critical metrics for preparing and responding to catastrophes when they occur, he stated.
“They had looked at the lifecycle of the claims process and evaluated, ‘What are the key components from a customer perspective that are really going to make a difference, both for us financially and for the customer?'” he said.
In doing this, the claim and underwriting teams were able to determine that a first notice of loss phone call within an eight second time frame and an average time frame of 24 hours for first contact with a policyholder about a claim were critical steps in the response process.
“I think doing that in the beginning and setting the right framework was really important,” Langione said.
Jones agreed, adding that The Travelers Companies works to construct maps based on the footprints of the storms as they’re approaching. This way, it can pinpoint the customers and corresponding agents to speak with and what questions to ask.
“We really consider the involvement of our agents and brokers as a critical part of the claim handling process,” he said.
As a storm approaches, The Travelers Companies has contracts with various third parties that can put tarps on customer’s houses, clear trees or take additional precautionary measures ahead of time, he said.
“That way, when we actually get there, the deals are already in place so that we can quickly go operational with the types of things that we need to do to get the families and businesses back in operation as quickly as possible,” he said.
Jones added that 2017’s Northern California wildfires, about which the California Department of Insurance announced mid-December last year that claims from the fires had already topped $9.4 billion, also taught the industry about the importance of not relying exclusively on just one model for major perils.
“Not all models are created equal,” he said.
Indeed, he explained that the wildfire models are more immature than some of the hurricane, earthquake and terror models. Because of this, the industry needs to understand where particular models sit in terms of maturity and the type of activity that is taking place.
“Then, you step back and say: ‘If you didn’t have the model at all, what would you do?’ he said.
This is when historical experience needs to step in, he stated.
“I think the wildfires were a good example of where they didn’t think embers could fly as far as they did,” he said. “So the scope of what happened to some extent had scientific issues that I think people are going to have to go back and think about. The other thing is, there are five times as many people in that area as there were 20 or 30 years ago. So you do have to look at models not only from the perspective of the science behind the peril itself, but also what the exposures are there that might not have been there 10 or 15 years ago.”
Despite the challenges, one pleasant surprise in last year’s natural disasters from Jones’ perspective was the way the capital markets and the reinsurers responded after the events, he said.
“We had always been wondering if there really was a decent size cat event, were there going to be any anomalies that change our view on reinsurance and how much we should be using the capital markets?” he said. “I think we were very favorably surprised by how calm the markets reacted and how thoughtful the capital was in terms of its availability.”
He said he believes this has given the industry more confidence to use those markets over the long term in order to deliver consistent pricing throughout the process to policy holders.
With all of this in mind, Jones said that he’s seen the insurance industry doing a better job of identifying elements that need to be corrected in modeling along the way, as well as working to better understand how models are built.
“That way, we don’t have to make these major adjustments to models when something comes out of pattern, because we’re adjusting things along the way as we’re seeing different things about every storm,” he said. “That gives us a robust way to keep current with the market place, keep current with whatever is happening with respect to claims and the climate, and also what’s happening to the actual exposures that we’re insuring.”
Kelner stated that overall, he believes the insurance industry’s response to each of the events of last year was positive and will only continue to get better as the industry learns, which he sees as an ongoing process.
“We saw a resilient market in the face of adversity,” he said. “I think the goal of better risk assessment, faster risk assessment and better exposure understanding is a never-ending procedure. We’ve got the benefit of great advances in technology, and so the efforts that companies are making to better understand their risks continue.”