Know Your ‘Basis Risk’ in the Modeled Vulnerabilities of Your Customers' Exposures

SOVs contain crucial data about construction attributes that insurers across the market use as inputs to their catastrophe models. These attributes, such as Construction Class, Year Built, Number of Stories, and Occupancy (not to mention myriad secondary modifiers) map in various models to specific vulnerability curves which then define the damageability of the buildings. Along with hazards, these vulnerability curves are crucial to the modeled losses and drive the AALs and other metrics insurers use to underwrite and price their submissions.

Vulnerability curves, while based on sound empirical data and engineering considerations, are generalizations about the performance of specific types of buildings. While this approach can work well in an aggregate analysis involving a diversity of property types and vulnerability classes, there are some cases where the default use of generalized vulnerability curves do not adequately address the performance of a particular construction style. This ‘basis risk’ can be particularly pronounced for certain types of submissions where an insured focuses on a specific kind of asset class or property type. 

A notable example of this can occur when modeling a submission from certain owners of large industrial warehouses. Industrial warehouses, across the country, are often constructed using precast concrete tilt-up walls. And, when coded as such, they are typically modeled by a singular “primary” vulnerability curve (subject to secondary modifiers) that have been constructed across the US for the last 20 years.

In most models, the seismic vulnerability curves for precast concrete tilt-up buildings tend to produce higher damage ratios than many other types of structures. These parameterizations were heavily influenced by the performance of older tilt-up construction of smaller warehouses where the connections between the roof diaphragm and the exterior tilt-up walls failed resulting in a partial collapse of the building. In these older structures, the lateral load resisting system was limited to exterior concrete (tilt-up) shear walls with the gravity loads carried by the exterior walls and interior steel columns. This failure mode was observed in the 1971 San Fernando EQ and repeated itself, although to a much lesser degree, in the 1994 Northridge EQ. Building codes were updated for precast concrete tilt-up construction in 1997 and have remained essentially the same since then.

However, the trend over the last 20 years has been to develop increasingly larger warehouses with many contemporary warehouses now constructed in the 250K to 1M+ square feet range. A major difference in the construction of these large warehouses is that they are designed with multiple interior bays of braced (or moment resisting) steel frames that resist the lateral loads caused by seismic and wind events. As a structural engineering best practice, it was determined that a flexible steel (or wood) diaphragm should not be required to collect and distribute lateral loads when spans exceed 400-500 feet. Therefore, with these large warehouses, the lateral load resisting design utilizes interior steel braced frames as the primary load resisting system, not the exterior tilt-up walls. I say, ‘primary’, since warehouse roofs are generally flexible diaphragms made of steel decking (or sometimes wood decking) and flexible diaphragms distribute lateral loads based on tributary area, therefore, an interior column line of steel braced frames will typically carry twice the lateral load as compared to an exterior concrete shear wall. For example, if the building has two column lines of steel braced frames, the steel frames carry 66% of the lateral loads, whereas for three column lines, 75% of the lateral loading is carried by the steel frames. 

In terms of structural design of these types of buildings, they can be considered to be a dual system with the primary lateral load resisting system being steel braced (or moment resisting) frames with redundancy provided by the exterior concrete shear walls. This is quite different from a default to a traditional ‘tilt-up’ vulnerability curve, which would overestimate the damage potential of such a property. It’s important, as in this example and in others that I’ll explore in subsequent posts, to understand the basis for the modeling assumptions, and their fit with your customers’ portfolios in order to have a more comprehensive understanding of their risk and how to contextualize the resulting AALs generated by the models.