What we learned from Katrina - Hurricane Katrina 20th Anniversary Report

What we learned from Katrina

The past decades have shown events like Katrina are not disappearing. If anything, costly catastrophes are an annual occurrence. But along with damage and destruction come valuable lessons — and opportunities to address critical insurance and risk management gaps.

Lessons in risk modeling

Even before Katrina, the insurance industry had embraced the use of catastrophe models in underwriting, having seen their potential for forecasting insured losses during Hurricane Andrew 13 years before. For the most part, though, insurers used these tools in the background.

In the two decades since Katrina, catastrophe models have moved to the forefront. Models are now more sophisticated and allow for deeper analysis, due to advances in computing and the availability of more granular data sets. Catastrophe modeling is now widely used by primary insurers, reinsurers, and even corporate enterprises as a fundamental step in pricing property risk and informing decisions on risk mitigation.

One lesson Katrina imparted to risk modelers was the need to consider consequential events. The wind component of hurricanes had been incorporated into models well before 2005, but flood was not. Even though Katrina was a strong windstorm, it was the failure of New Orleans’ levees and catastrophic flooding — both wind-driven and pluvial — that caused most of the damage.

Other lessons that risk modelers learned at the time of Katrina included the impact of multiple events and “demand surge,” in which local and regional building material shortages increase the cost of property repairs. Both 2004 and 2005 were unusual years for hurricane frequency. In 2004, Hurricanes Charley, Frances, Ivan, and Jeanne struck Florida within a six-week period; in 2005, Katrina was followed by Rita and Wilma.

All were multibillion-dollar loss events. A total of 28 named storms formed in 2005, forcing NOAA to resort to Greek letters to identify them. Similar catastrophe frequency occurred again in 2017, with Hurricanes Harvey, Ian, and Maria. And in 2020, a record 30 named storms formed in the Atlantic.

Models have become faster and more accurate and can now process stochastic outcomes and greater volumes of data sets. And each subsequent catastrophe event offers additional opportunities to revise model assumptions and recalibrate catastrophic loss exposures.

For example, today’s models are beginning to take into consideration topographic changes, an area of growing interest to catastrophe modelers, especially as respects water runoff. If there is nowhere for excessive rainfall to flow, surface flooding will be inevitable.

Models have evolved to allow for more primary and secondary building characteristics, enabling insurers to fine-tune the output of modeled losses and reduce uncertainty. This advancement also is cross-pollinating models for so-called “secondary” perils, such as flood, severe convective storm, hail, and extreme weather.

What is stochastic modeling?

Catastrophe models typically rely on stochastic, or randomly distributed, data that can be statistically analyzed but not precisely predicted. This data is characterized by event intensity, geography, and probability. A stochastic model is probabilistic, meaning it delivers outputs in likelihood of occurrence, somewhat like a weather forecast that indicates a 70% chance of rain.

For example, a stochastic data set for a tropical cyclone might model a Category 5 hurricane making landfall in Miami, and simulate many potential scenarios. Additional components of catastrophe models may include specific hazard, damage, and financial data modules. The results of these models can help guide major decision-making about insurance programs and risk mitigation efforts.

Many risk models widely used across the global insurance industry deploy stochastic data. Moody’s RMS, for example, uses stochastic event data sets for its cyclone, earthquake, and severe convective storm models. These models incorporate landfall rates, shake intensity, and frequency rates specific to regions and basins, and they account for differences in local building codes, antecedent conditions, and findings from historical events.

Verisk’s catastrophe risk models have a similar framework, incorporating exposure data about properties in a given location, replacement values, and physical characteristics. Tropical cyclone models offered by RMS also integrate climate change trends and the impact of inland flooding.

Lessons in risk mitigation

Most property insurance buyers share some of the risk of damage from natural catastrophes through deductibles and self-insured retentions, providing an incentive to help minimize loss. Katrina’s impact was devastating to many businesses because they didn’t anticipate, mitigate, or purchase insurance for flooding. Risk assessment tools developed since Katrina have made risk identification an easier process.

The sophistication of today’s risk models has enabled property owners to see how mitigation steps can reduce their cost of risk. This provides an opportunity to use historical loss data to test returns on investment (ROI) in different types of loss control measures, including floodwalls and defensive landscaping.

Awareness of flood risks and their potential impact following hurricanes grew after Katrina in 2005, Sandy in 2012, and Harvey in 2017. These events demonstrated the value of several important risk mitigation techniques, including:

Elevating critical assets above expected water levels.
Maintaining backup systems in safe locations.
Storing important documents offsite.
Establishing backup communication plans.
Enacting flood protection systems.
Developing and testing business continuity plans.

Retrofitting existing buildings to withstand catastrophic events and extreme weather may be impossible, or prohibitively expensive. Instead, incremental reinforcement of vulnerable spots, such as roofs, windows, entryways, and loading docks may be an easier and more cost-effective way for property owners to mitigate property damage.