Currently showing: Climate/natural disasters > Disaster risk

25 Jul 14 23:00

I was recently asked to take part in a Panel discussion focusing on current topics surrounding "Property Risk" in Kansas. The event was co-sponsored by the Kansas Association of Insurance agents (KAIA), the Wichita Chapter of CPCU, and KIA/1752.

The format was fairly casual with set questions being asked of the panel by a MC with contributions welcomed by the 50+ attendees who included underwriters, managers and agents responsible principally for the Kansas region and the session lasted approximately 2 hours, I always enjoy such meetings as they provide a great opportunity to engage with representatives of companies at a very local level.

As part of the panel consisting of two local agents, a regional company underwriter and a member of the Kansas Insurance Department, we got down to discussing a number of issues that impact them most, including how the advancement of tornado and hail modeling has grown in recent years and is being used within the industry.

Hurricanes rely on a number of measurable data points such as water temperature and upper level winds for formation and sustenance from which conclusions can be made on future expectations on frequency and severity, this is far harder to deduce in the case of Severe Convective Storms. It is very likely that in recent years, some areas, even in the Mid-West are experiencing greater activity than in the measurable historical record and it is reasonable to conclude that many of these areas are more densely populated leading to increases in insured losses and shifts in awareness. In todays interconnected world, graphic reports of damage are made available instantaneously but comparative conclusions on past activity is hard to derive as the historical record on tornado formation is far from complete, 40 -50 years ago we simply did not have the reporting ability that we have now.

The modeling of cat perils has been with us now since the early 90s, but the main focus of the industry's efforts has been on the so called "tail losses", which are the more extreme industry losses that can be expected to originate from hurricane and earthquake events.

Many of the models targeting these principle perils are in their seventh or eighth reiteration and thus have been refined over the past 20+ years. The understanding of the models themselves has also grown alongside these years of development and as an industry; I feel that we have moved from initial sceptiscm, to blind obedience and now to the current status of having a far healthier understanding of the models strengths and weaknesses.

Modeling of Severe Convective Storms (SCS) however, is in its comparative infancy. In recent years nature has provided us with several examples of SCS formations that have produced Industry losses in excess of $6.00 - $7.00 billion dollars. It is this rise in frequency and severity which has produced a drive for greater efforts to be expanded within the modeling fraternity to try and gain a better understanding of the potential damage presented by these events. It is clear that the industry will continue to respond and experience SCS events and the modeling efforts are equally bound to continue and improve in sophistication but tornado/hail modeling is inherently difficult to tie back to a local level.

At a macro level the models may be stress tested to be directionally correct but the damage resulting from SCS events appears so naturally chaotic and random that I feel that the inherent uncertainty contained within the modeling at the local level will always need to be properly understood by the primary insurance industry, re insurers and regulators alike.

Are there any thoughts out there as to how it is felt that tornado/hail modeling itself has changed the reinsurance buying behavior within the industry or is it still felt to be a desire or need driven by local experience?

Category: Climate/natural disasters: Disaster risk

Location: US Mid West - Kansas and beyond

1 Comment

Tom Smith - 31 Jul 2014, 8:14 p.m.

Great article Ken. In my experience, unlike hurricane, tornado/hail model changes have had little impact on reinsurance buying behavior. The industry as a whole appears to have limited confidence in the models and instead it relies on local experience when determining reinsurance needs. I like the descriptive language you used in your second to last paragraph.... where you use words like naturally chaotic, random and inherent uncertainty to describe T/H model output. This is the very reason why we struggle so much with the Cat Agg deals we see that are mostly tornado and hail driven.

If you would like to leave a comment, please, log in.