Cat Model Limitations

Catastrophe models are indispensable tools. They are not infallible ones — and the 2025 loss experience made that distinction very clear.

Model validation is a continuous process, not a one-time certification. Secondary perils, changing climate patterns, and urban exposure growth are all introducing variability that historical calibrations weren't designed to capture.

The reinsurance market has already adjusted its view of certain perils and geographies based on observed loss patterns. Primary carriers are following suit, reassessing accumulations and coverage terms in segments where model uncertainty is highest.

The honest message for underwriters and executives: use models as the starting point for risk assessment, not the ending point. Apply judgment, track model error, and maintain conservative buffers in lines where uncertainty is genuinely high.