Emerging Estimation Risks

Reforms enacted in the mid-2000s resulted in lower premiums and loss costs and favorable frequency and severity trends. However, workers' comp medical expenses, which are higher than the overall medical Consumer Price Index, have outpaced favorable lower-claims-frequency trends.

According to the National Council on Compensation Insurance, medical costs now contribute a staggering 58 percent of total comp-loss costs, up from 49 percent in 1991. The combined net impact of these trends resulted in deteriorating combined net ratios from 99 percent in 2006 up to 111 percent in 2009.

Since the 2009 global financial crisis, much of the P&C industry has been on the road to recovery, yet workers' comp underwriting results have continued to weaken. This is due to a conflux of factors, including decreasing payrolls; declining premium volume emanating from both competitive rate decreases and prior-year return-premium adjustments; the highest medical-cost trends among all industry segments; and lower long-term investment returns.

Further, the Obama administration's health-care reforms create new and unprecedented uncertainties associated with potential medical-loss-cost shifts between workers' comp insurers and health insurers.

THE Historical Evidence

Loss reserves for workers' comp are essentially forecasts of losses that will be paid over five, 10 and 15-plus years. As a result, they are among the most challenging balance-sheet risks to quantify.

Under standard reserving methods, known as chain-ladder methods, an actuary examines historic data, measures existing patterns and makes forecasts based on the assumption that those patterns will repeat.

Sometimes there is insufficient data to measure the pattern reliably on a given line of business. Yet other times trends change and the patterns from the past either simply do not repeat or are difficult to capture.

The worst cases of adverse reserve development occur when there are material and unpredicted changes in the underlying trends, as we are experiencing in the current climate.

For example, for accident-year 2000, the estimate of ultimate loss increased by around 25 percent from its first estimate at the end of 2000 to its re-estimate nine years later.

The 1997-2002 soft market was underpriced just as workers' comp medical-cost inflation escalated erratically relative to previously assumed assumptions.

How to Minimize Inaccurate Estimates

Over the past 20 years, actuarial pioneers have worked to apply modern statistical theories to loss reserves. In practice, however, these ideas have not been widely adopted. New reserve methods on long-tail casualty lines will need to offer demonstrable advantages that override the natural instinct to keep time-tested ways.

Insurers require a solution with proven accuracy and results they can verify and explain. Programming advanced statistical methods may be too time-consuming for individual companies, so commercial software is needed. Up until recently, there haven't been many practical, cost-efficient solutions available.

Our firm is among those that are now using a statistical approach known as generalized linear modeling (GLM) to identify and display trends in an insurer's workers' comp data that previously were hidden under a chain-ladder method. (See related textbox for technical explanation.)

Under GLM, future trends and related uncertainty are explicitly estimated. The default is not to assume that future trends—in overall inflation or medical-cost inflation or claims frequency, for example—will be the same as in the past and understate loss reserve risk. It is possible under GLM to capture changing trends while stressing various dynamic and inflationary scenarios to current net loss reserve positions.

Currently, companies that attempt to measure loss-reserve risk (the uncertainty inherent in their loss-reserve estimates) use one of two techniques (known as the Mack and Bootstrapping techniques), both of which are based on the chain-ladder method. As we have seen, the chain-ladder method, rooted in the assumption that the future will look like the past, has produced inaccurate point estimates of ultimate losses for decades. It follows that any technique designed to measure uncertainty that relies on that basis premise will be flawed. Future trends will certainly be different from the past, and this uncertainty is being ignored by current methods, significantly understating reserve risk.

Since loss-reserve risk can account for the greatest drain on corporate capital, it is essential that insurers' reserve modeling addresses previous models' limitations while contemplating sound statistical principles. Furthermore, the factors that drive loss-reserve risk also contribute to underwriting risk for long-tail business.

Insurers may initially be drawn to GLM for its ability to measure loss-reserve risk, especially since they increasingly are being required to report these numbers to regulators and rating agencies. For example, A.M. Best has added a new enterprise-risk-management section to its latest Supplemental Rating Questionnaire that inquires about several specific inflationary trends and scenarios—both anticipated and unanticipated—that could impact carriers' net reserve positions.

In a broader sense, GLM will also help insurers navigate the long-tail risks in the dynamic workers' comp environment. The essential premise of this approach— particularly relevant in this environment—is that trends may change. The modeling displays the real trends in an underlying workers' comp insurer's data, whether they have been steady or changing.

This improved understanding of the past has led to important advantages and improvements. Forecasting decisions can now be well informed, transparent and explainable. In addition, reserve risk modeling can better contemplate the risk of a changing future and the risk drivers that dynamic reserve modeling measures can be incorporated into a capital model.

Now, insurers can better manage the risks of long-tail workers' compensation business in the context of their corporate capital requirements.