Catastrophe Model Flaws Leave Workers' Comp Carriers At Risk

Saturday, June 2, 2007 | 0

By Scott Self

A major threat of lowered financial ratings and soaring reinsurance rates is lurking for workers' compensation insurers as a result of flawed catastrophe modeling data and process difficulties. Above all, there is the possibility of ticking time bombs within otherwise profitable portfolios, which might produce surprising loss amounts that could threaten the financial stability of insurance companies.

The 9/11 terrorist attacks concentrated the minds of insurers on the truly gigantic losses that are possible in workers' comp and underscored the need for accurate catastrophe modeling.

However, while much work has been done over the past six years by commercial modeling firms and some insurers to develop the modeling discipline, the accuracy of these models is being hampered by poor data quality, which is a growing concern within the industry.

As models increasingly are used by reinsurers and required by rating companies, ceding companies will be at a competitive disadvantage if they do not gather accurate, consistent data about their workers' comp exposures.

Poor data and imprecise modeling eventually could lead to adverse reinsurance pricing for ceding companies, which also could find their financial ratings under threat.

The current lack of accurate workers' comp modeling could one day pose a problem for underwriters, unless they get a better handle on their loss exposures.

In their development, workers' comp models currently are at a stage similar to that of property models 15 years ago. There are several--and, certainly, very understandable--reasons that workers' comp modeling efforts lag behind property models:

* Modeling is a complicated discipline and takes time to fully develop.

It took the catastrophic loss of Hurricane Andrew in 1992 to provide the impetus for the major developments in property modeling. If we view 9/11 as providing a similar impetus for workers' comp models, real strides have been made in the years since.

* Few people could have imagined that two buildings would collapse in a terrorist attack with such an enormous loss of life and physical injuries.

The 9/11 attacks shattered assumptions about the potential for losses when a catastrophe hits an office building during work hours.

The terrorist attacks on the World Trade Center and Pentagon represented the largest ever workers' comp loss. Indeed, in September 2006, the Insurance Information Institute estimated that workers' comp losses from the event totaled nearly $2 billion.

* There have not been as many resources allocated to modeling workers' comp as there have to modeling property.

One reason for this is that reinsurers have driven much of the effort to develop property models, primarily because of their huge financial stake in the property market.

Nevertheless, reinsurers increasingly are getting behind the effort to create accurate workers' comp modeling techniques in order to protect their own balance sheets.

* Workers' comp is a complex risk and is harder to assess than property risks.

For example, employees move around within locations, making occupancy totals hard to determine in workers' comp modeling.

In addition, a disparate range of injuries--from minor wounds to paralysis requiring lifetime care--can result from the same event.

* Workers' compensation modeling faces two significant obstacles which impede accuracy: poor data quality and high inherent uncertainty.

Alone, either of these two issues can be overcome, but together they function as a major impediment to accurate modeling.

So, what can be done to achieve more effective workers' comp modeling?

The evolution of modeling techniques certainly would be expedited if everyone spoke the same language--the same vocabulary--in order to standardize the data required for computer entry and analysis.

Standards need to be developed because the expectations of the data required for modeling can vary dramatically--for example, between reinsurers and rating agencies. These discrepancies can lead to inadequate and inconsistent data quality.

For the models to be effective, all interested parties need to agree on the data to be captured and how the data will be factored into underwriting decisions and risk management practices. (For a short list of the ideal data required, see infographic.)

An insurer's data often does not reach this level of detail, and instead, for example, is restricted to accumulated employee counts or payroll at a main mailing address. This lack of precise data means that books of business cannot be adequately modeled, as large blocks of exposures are unaccounted for in terms of locations, covered employees, or both.

However, even with accurate and specific data, modeling of workers' comp risk is still complex due to the nature of the perils involved.

Workers' comp offers the added difficulties of accounting for event occurrence times, mobile employees and the fact that most catastrophe perils impacting workers' comp are low in frequency but high in severity.

The timing of an event is a major factor simply because employees do not work all the time. A building is a fixed, static object of constant value for 24 hours each day; employees, however, move around.

For example, an earthquake occurring at 2 a.m. would have a very different impact on a book of workers' comp business than one occurring at 2 p.m., even though the property loss might be the same.

Mobile employees are another complicating factor. Contractors, delivery truck drivers and similar workers cannot be located at specific, fixed locations. Instead, their locations can vary in the course of a work shift. Approximations, therefore, must be made for their loss potential, adding to the underlying uncertainty of an analysis.

Sept. 11, 2001 stands as the seminal moment in workers' comp loss, but severe earthquakes, industrial accidents or new terrorist attacks all threaten to inflict major losses in the future.

This high severity of major events, paired with their low frequency, means that while catastrophe losses in any given year are likely to be zero, the potential for an extreme loss cannot be discounted.

Accurately quantifying this exposure and loss potential is essential to effective risk management. Failing to do so can lead to adverse pricing from reinsurers unable to adequately measure the risk of a book of business.

Likewise, it can engender skepticism from rating agencies about risk exposure and management practices.

Scott Self is senior vice president of Collins, a reinsurance brokerage, and leader of its Workers' Compensation Practice Group in Burlington, N.C. He may be reached at scott.self@collins.com. This column first appeared in National Underwriter magazine. To learn, visit their Web site at:

http://www.propertyandcasualtyinsurancenews.com/cms/nupc/weekly%20issues/channels/Workers%20Compensation

-------------------------------

The views and opinions expressed by the author are not necessarily those of workcompcentral.com, its editors or management.