Canadian Underwriter
Feature

Knowing the Unknown


October 1, 2013   by Desmond Carroll Assistant Vice President, Guy Carpenter Canada


Print this page Share

The number of natural hazards that threaten insured risks in Canada is extensive. That said, very few of these perils have an industry-accepted catastrophe model.

This raises a number of questions: What motivates the industry to develop a model for a specific peril? How much does the insurance industry’s current demand impact the likelihood of development of a new catastrophe model? Which non-modelled peril should be the first to become modelled in Canada?

Three recent catastrophic events illustrate these questions: the 1998 Quebec ice storm, the 2003 Okanagan wildfire, and the 2013 Alberta floods. While all are different events, they each share commonalities. Each event was a national tragedy that resulted in previously unseen losses to the Canadian insurance industry.

The existence of these perils was by no means a surprise to the insurance industry. However, the magnitude of loss from these events was not generally contemplated in companies’ disaster scenarios. In each of the aforementioned cases, the natural hazards were largely non-modelled perils at the time of the event.

Despite not being modelled, there is little doubt that the perils of severe winter storm, wildfire and flood were contemplated by insurers before these catastrophic events occurred.

• The Quebec ice storm represented a previously unimaginable extreme for the peril of winter storms, but the hazards of living in a northern climate are well-known to the industry.

• As regards wildfire, across Canada authorities spend $500 million to $1 billion annually combatting this peril. Despite the extraordinary work of the nation’s wildfire suppression professionals, the effect of suppression on fires of the scale witnessed in the Okanagan event is minimal and the hazard to insured property persists.

• In terms of flood, Canada has the world’s most extensive hydrological network comprising more than 2 million lakes and innumerable rivers and streams. There is a robust history of flood events causing significant damage to insured property in the country.

So, while there did exist an awareness of all these perils prior to significant catastrophic events, they were essentially treated as “known unknowns” in the absence of formal models.

In the wake of each of these events, industry demand for the development of a formal model to characterize risk to an insured portfolio was strong, but the actual response seemingly more lackluster. Of the three perils mentioned previously, only winter storm has received specific modelling attention, and even this was not completed until 2008 (by catastrophe modelling firm RMS).

Given the 10-year period between the event and the release of the model, it is difficult to view this response to market demand for a modelling solution as enthusiastic. It is, therefore, a fair conclusion – given the number of natural catastrophes that have occurred in Canada since 1998 and the lack of models that have been released in response to these events – that any single catastrophic occurrence has nominal influence on the catastrophe models that are subsequently developed.

In fairness to the modelling firms, the development of any catastrophe model that provides an informed view of the risk imposed by a given peril is a difficult undertaking that requires significant resources, financial commitment, extensive research, enormous amounts of data and computational power.

For companies looking to develop Canada-specific models, the hurdles of data and computational processing are usually among the most challenging. The lack of available critical geospatial data sets and historical event information can be incredibly frustrating and can cripple otherwise strong efforts at model development.

Coping with the sheer scale of Canada can also never be understated. With a population rivalling California’s, dispersed among an area 24 times larger, it is understandable that some modelling firms simply choose to focus their resources elsewhere. However, industry is starting to see at least some of these obstacles become less daunting for those looking to build models for Canada.

Just a few years ago, the resolution required to accurately model a peril such as flood or wildfire made the task of creating a model for a country like Canada computationally unfeasible.

Today, the computational power exists. However, many of the other challenges linger, making reacting quickly to single events nearly impossible for model developers.

Of the non-modelled natural perils, tsunami, wildfire and flood are commonly cited as having the highest potential to produce large insured and economic losses in Canada. Following the 2011 Tohoku (Japan) and the 2004 Indonesian tsunamis, many insurers and regulatory bodies began focusing on these perils and made attempts to quantify the exposure that exists in Canada. The largest risk to the country from tsunami would be from tectonic activity along the Pacific Rim, which threatens the Pacific Coast, including the west coast of Vancouver Island.

The key mitigating factor to tsunami risk is the low population density in the most exposed areas; for the City of Vancouver, tsunami risk is considered muted because of the shielding effect of Vancouver Island.

Compared to wildfire or flood, the risk posed to insured value is considered to be significantly lower for this peril. As such, one would expect the tsunami threat to be a lower industry priority than the other non-modelled perils.

Wildfire is a peril that has caused two major losses in the last decade and poses a significant risk to insured values, particularly in the interior of British Columbia and Alberta. The Okanagan and Slave Lake wildfires illustrate the difficulty in suppressing the 3% of wildfires that make up 97% of acreage consumed by wildfires.

The Wildland Urban Interface (WUI) is the transition zone between undeveloped land and developed communities – and it is these areas where the risk is greatest. Efforts such as FireSmart, a government program that aims to mitigate the risk by hardening these WUI communities against wildfire, have proven to be a cost-effective way to diminish the danger. However, widespread adoption of these measures has not been achieved in Canada.

Wildfire will continue to pose a meaningful threat to communities that are near the WUI and will continue to prove a difficult peril to effectively mitigate and suppress. As an industry, wildfire is a hazard that is much more on the radar for model development.

At the forefront of many insurance industry professionals’ minds is the peril of flood. In a nation with such a high density of rivers and lakes, it should come as no surprise that it ranks as one of the country’s most common natural catastrophes. The problem is compounded by the fact that our population centres tend to cluster around the areas where the hazard is highest, and is further complicated by the relationship between overland flood and sewer back-up cover in insurance policies.

Should overland flood ever be considered as an insured peril for homeowners policies, a viable hazard model is crucial to assist in underwriting the risk, quantifying exposures and better informing reinsurance purchases.

Unfortunately, acknowledging these challenges with building a natural hazard model for Canada, it is not feasible to address all of the currently non-modelled perils concurrently and with equal immediacy. The choice of which peril should be the focus of model builders’ scarce resources is, therefore, of immense importance to insurers, as well as to reinsurers, governments, infrastructure decision-makers and even policyholders.

With perfect clairvoyance, industry would pick the next non-modelled peril to create a significant industry event in Canada and devote all attention to understanding it, with the hope of mitigating and managing the impact. However, in practice, retrospective tendencies may prevail.

Regardless, priority will be given to those non-modelled perils tha
t deliver the greatest value to the broadest constituency, while also satisfying whatever incentive exists for those building the models.

Despite the natural bias to place greater emphasis on recent catastrophic events, it is important to step back and consider the decision to model a specific peril – not in the context of which event caused the latest record-breaking loss, but which risk is most likely to be both pervasive and catastrophic in the future.


Print this page Share

Have your say:

Your email address will not be published. Required fields are marked *

*