Catastrophe bonds are priced and structured using outputs from a small number of specialist modeling firms. These firms simulate millions of possible disaster scenarios to produce expected loss, probability of first loss, and probability of exhaustion—metrics that drive both sponsor and investor decisions. For anyone allocating to cat bonds, understanding the role of these modelers is essential.
Why Catastrophe Models Matter
Unlike corporate bonds, cat bond payouts depend on physical events (hurricanes, earthquakes, wildfires) whose frequency and severity are estimated statistically. There is no credit model that can derive a “default probability” from financial statements. Instead, (re)insurers and investors rely on catastrophe models that combine:
- Event sets: Catalogs of simulated events (e.g., thousands of possible hurricane tracks and intensities).
- Hazard modules: How strong the hazard is at each location (wind speed, ground motion, flood depth).
- Vulnerability (damage) functions: How buildings and exposures respond to the hazard.
- Exposure data: What the sponsor (or industry) has at risk.
From these, modelers output expected annual loss (EAL), occurrence exceedance probability (OEP), and aggregate exceedance probability (AEP) curves. Those outputs feed into trigger design, attachment/exhaustion levels, and the spreads investors demand.
The Major Players
Three firms dominate the global catastrophe modeling market for property risk: RMS (Risk Management Solutions), AIR Worldwide (Applied Insurance Research, part of Verisk), and CoreLogic (which incorporates the former EQECAT). All three provide models for the perils that dominate the cat bond market—U.S. hurricane, U.S. earthquake, European windstorm, and Japan earthquake—as well as a growing set of secondary perils and non-U.S. regions.
RMS
RMS has been a central player in the ILS and cat bond market since its inception. Its platform is widely used for modeled-loss triggers, where bond payouts are tied to model-derived loss estimates rather than actual reported losses. RMS also supplies industry exposure databases and event sets that support both indemnity and parametric structures. Sponsors and investors often run “dual” or “triple” model comparisons (RMS, AIR, CoreLogic) to gauge model consensus and disagreement.
AIR Worldwide
AIR (a Verisk company) offers a full suite of catastrophe models and is commonly used alongside RMS in deal structuring and due diligence. AIR’s models are used in both indemnity deals (for risk metrics and pricing) and in modeled-loss and industry-loss trigger structures. Many cat bond offering documents disclose expected loss and similar metrics from more than one modeler; AIR is frequently one of them.
CoreLogic
CoreLogic’s catastrophe modeling capabilities include the legacy EQECAT models and exposure data. The firm is particularly known for property-level data and models used in U.S. wind, earthquake, and flood. In the cat bond context, CoreLogic is often used as a third model view or for specific perils and regions where its data or methodology is preferred.
How Models Enter the Cat Bond Workflow
- Structuring: The sponsor and structurer choose attachment and exhaustion points. Model outputs (e.g., expected loss at various layers) inform these choices and the size of the tranche.
- Trigger design: For modeled-loss triggers, the bond’s payout is tied directly to one or more models’ loss estimates for a given event. For indemnity triggers, models are still used to price the bond and to set retention and limits; the trigger itself is actual loss.
- Pricing: Investors and reinsurance underwriters use the same (or comparable) models to derive expected loss and to compare spreads across deals. “Price per unit of expected loss” is a common metric in the market.
- Disclosure: Offering documents typically include expected loss and sometimes OEP/AEP from at least one major modeler. This allows investors to compare deals on a more consistent basis.
Model Risk — What Investors Should Watch
Model disagreement: RMS, AIR, and CoreLogic can produce meaningfully different expected losses for the same deal. That “model spread” is a form of uncertainty. Some investors require two or three model runs and use the average or the most conservative view.
Non-stationarity: Catastrophe models are calibrated using historical data and scientific relationships. Climate change and shifting exposure (e.g., coastal development, wildfire zones) can make the future distribution of events different from the past. Modelers update their models periodically; investors should be aware that today’s expected loss is a snapshot, not a permanent truth.
Trigger basis: For modeled-loss triggers, the bond pays when the model says loss exceeds a threshold. If the model underestimates or overestimates actual loss, investors bear that basis risk. Understanding which model(s) drive the trigger and how they have performed in past events is part of due diligence.
Practical Takeaways
- Expect multiple models: Major cat bond deals are often analyzed with RMS, AIR, and sometimes CoreLogic. Comparing expected loss and structure across modelers improves discipline.
- Read the documents: Offering materials usually state which model(s) were used for expected loss and, for modeled-loss triggers, which model(s) drive the trigger. That clarity is essential for risk and basis assessment.
- Stay updated: Model versions change. Vendors release new editions that can shift expected loss and rankings between deals. Keeping track of model versions and release notes helps avoid stale assumptions.
The cat bond market would not function at its current scale without robust, third-party catastrophe models. RMS, AIR, and CoreLogic provide the common language of risk that allows sponsors to transfer peak peril and investors to price and manage it. A solid grasp of how these firms operate and where model risk sits will make you a more informed allocator.