The Risks of Hazard

This week I attended the PCI National Flood Conference in Washington, DC (in Arlington, VA for geographic sticklers like me). Over the weekend, as everyone converged on the city, the sun was shining and the Potomac was sparkling as it flowed gently between its protected banks into Chesapeake Bay. It was a welcoming setting, especially for the delegation coming over from London – a significant number of underwriters and brokers from Lloyd’s.

With the published agenda occupied by 90 percent NFIP material meant to cover assorted minutiae of the program, what might these underwriters be doing there? Private flood is what they were doing there. The remaining 10 percent of the agenda addressed private flood topics, but it was the corridor talk that was predominantly about private flood.

Two weeks ago, we published a post on how analytics aren’t so smart after all. It has been a sensation (only our Cat Modeler’s Guide to the Protection Gap has been read more this year), so it is only natural that we pursue the topic further.

The blog that lit up a little corner of the web was about the limitations of analytics, specifically flood risk analytics. It turns out there is some interesting science and folklore about the topic, so let’s take a look.

Analytics are an attempt to measure an immeasurable phenomenon. Cat risk analytics (flood, quake, hail, etc.) are perfect examples, as they aspire to determine the chance of something happening in a given location during a given timeframe – it is not possible. However, by evaluating a combination of different types of information, they can begin to produce results that are useful for underwriters who need those unknowable answers.

In the past, this blog has asked rhetorically “Why do insurers ignore wildfire?” It’s a peril that causes real losses (about $6B in the past 20 years in the USA alone), yet has data and analytics that can significantly improve residential and commercial underwriting for it. The readership metrics for Risks of Hazard reinforces the question: blogs about wildfire get opened less than any other subject we write about. Today, let’s answer that rhetorical question: Here is why insurers pay less attention to wildfire than they should.

Every summer, wildfire is high profile. The names of some recent fires are evocative: Rim, Sleepy Hollow, Waldo Canyon. This year the headlines are coming from the East: Pennsylvania has the Lehigh Valley Fire, and the late April NFS situation report lists fires in six eastern states. In Canada, the news is harrowing as Fort McMurray (pop. 80,000) is being evacuated as the city burns. Notoriety and media coverage is not the problem.

Last week, Cyrena Arnold, the Director of Product Sciences at Weather Analytics, and I blogged about hail, exploring available analytics with the March storm in the Fort Worth area. The response has been incredible to that blog, so a reprise is in order. Revisiting the topic seven days later might seem like a stretch, but there is plenty to discuss from those seven days.

Texas had a really bad March, with the big carriers posting results impacted by the March storms: Allstate, Travelers, and Progressive led the way. Then, April happened, including the April 12th storm in San Antonio that is shaping up as the costliest in Texas history – expecting over $1.3B in insured losses.

Earlier this month I published a post on LinkedIn about underwriting “challenging” flood in California’s Central Valley. It generated a decent readership and some likes, but, most importantly, it generated some comments. One of the comments posted was a very prescient piece of commentary, and it deserves a blog post to explore the topic it raised: The limitations of analytics.

The comment came from Mr. Tim Pappas (a VP at Gen Re), and I am grateful to him for raising this important topic. Here are the points he raises that this post will address:

• Without understanding the limitations of “superior analytics,” insurance companies can be putting their bottom line in great danger.
• The picture may not be quite as clear as the “high resolution” data provided indicates.
• The complexity of all the factors involved can lead to errors in estimation, making the computations imprecise and sometimes outright faulty.
• If the data is off on just a small percentage of risks, the impact on portfolio profitability can be quite large.