Modelling cat risk: 25 years since Hurricane Andrew

27-11-2017

Modelling cat risk: 25 years since Hurricane Andrew

iStock.com / michaelwarren

The understanding and ownership of scientific content will become a focal point of future catastrophe modelling, says Lixin Zeng, chief executive officer of Validus AlphaCat.

Hurricane Andrew in 1992 marked the beginning of the systematic adoption of cat models by the insurance industry. In the past 25 years have come (i) substantial improvements in their quality and coverage; and (ii) an ever-increasing sophistication in how the models are used by insurers, reinsurers, brokers, and capital market investors. This article highlights the latter.

The golden era of proprietary models

In the mid-1990s, insurance companies started using cat models to generate the data required for reinsurance purchases, typically with assistance from their reinsurance brokers. The availability of such data made it feasible for analytically focused reinsurers to pioneer a dynamic pricing and portfolio management approach driven by cat models.

Moreover, these companies demonstrated that rigorous risk quantification and portfolio optimisation could lead to sustainable and attractive risk-adjusted returns. Consequently, during the following 10 to 15 years, virtually all reinsurers and most large insurers endeavoured to establish this framework, where cat model outputs drove bespoke pricing and portfolio systems developed in-house.

During this period of rapid cat model adoption (1995 to 2010), most users accepted the fact that the cat models themselves remained proprietary in the sense that most of the underlying scientific content was not transparent, and meaningful customisation was difficult or impossible.

The dawn of transparency and diversity

By the end of the first decade of this century, with virtually all competitors equipped with vendor cat models, a growing number of sophisticated re/insurers had recognised that to maintain an edge in risk selection (and portfolio construction) they must truly own their view of risk.

They employ science, actuarial, and engineering professionals to create such in-house views. However, with the exception of event frequency, traditional vendor model designs make it practically impossible for a user to incorporate in-house views (eg, refined hazard assumptions, proprietary vulnerability curves, or advanced simulation methodologies) into the modelling process. As a result, we have observed the following developments:

  • Sensing a business opportunity, new business ventures have been formed to enable open access and model interoperability;
  • The most sophisticated re/insurers have developed internal modelling platforms to execute their in-house views of risk; and
  • With pressures from their customers and competition, incumbent model vendors have embraced a more open framework that allows the incorporation of external risk views while maintaining the clients’ ability to remain autonomous.

The developments serve to establish a healthy foundation for the next golden era of catastrophe modelling, characterised by greater transparency, interoperability and diversity of views. Going forward, it is anticipated that the understanding and ownership of scientific content will become a focal point of catastrophe modelling, consistent with similar trends in other industries, eg, television and media.

At the centre of a changing industry

The insurance industry is undergoing fundamental changes driven by technology advances and capital market encroachment—cat models are integral parts of both trends. Advances in data technology and computing power have enabled new powerful features, such as exposure augmentation and high-resolution modelling.

These nascent functions have substantially increased the business value and credibility of cat models. Meanwhile, this same credibility and transparency is further enhancing the attractiveness of the catastrophe re/insurance business to capital market investors.

This evolution has made cat models an important structural element enabling the ongoing fundamental change in the insurance industry. This was probably not expected by the early pioneers of cat modelling.

The author thanks Simon Biggs, Jeff Gall, and Peter Martin for their valuable inputs and suggestions.

Lixin Zeng is the chief executive officer of Validus AlphaCat.

models, modelling, risk, model, years, catastrophe, scientific, Hurricane, Andrew, reinsurers, Zeng, Lixin, AlphaCat

Bermuda Re