For reported losses, we establish reserves on a "case" basis within the parameters of coverage provided in the insurance policy or reinsurance agreement. For IBNR losses, we estimate reserves using established actuarial methods. Both our case and IBNR reserve estimates reflect such variables as past loss experience, changes in legislative conditions, changes in judicial interpretation of legal liability and policy coverages, and inflation. We consider not only monetary increases in the cost of what we insure, but also changes in societal factors that influence jury verdicts and case law and, in turn, claim costs.

Subjective judgments as to our ultimate exposure to losses are an integral and necessary component of our loss reserving process because many of the coverages we offer involve claims that may not ultimately be settled for many years after they are incurred. We record our reserves by considering a range of estimates bounded by a high and low point. Within that range, we record our best estimate. We continually review our reserves, using a variety of statistical and actuarial techniques to analyze current claim costs, frequency and severity data, and prevailing economic, social and legal factors. We adjust reserves established in prior years as loss experience develops and new information becomes available. Adjustments to previously estimated reserves are reflected in our financial results in the periods in which they are made.

A reconciliation of our beginning and ending loss and loss adjustment expense reserves for each of the last three years is included in Note 7 on page 60 of this report. That reconciliation shows that we have recorded reductions in the loss provision for claims incurred in prior years totaling $271 million, $627 million and $414 million in 1998, 1997 and 1996, respectively.

The reduction in prior year losses recorded in 1998 was significantly less than the reductions recorded in 1997 and 1996. The 1998 total was affected by the $250 million provision to strengthen loss reserves after the merger with USF&G, of which $189 million was classified as an increase in losses incurred in prior years. In addition, while the favorable prior year development on workers' compensation and medical professional liability coverages continued in 1998, the magnitude of that development was significantly less than that experienced in 1997 and 1996. We also experienced unfavorable prior year loss development in several commercial underwriting operations that have been affected by severe pricing pressures over the last several years, particularly our Middle Market and Construction business centers.

The favorable prior year development on workers' compensation business in our commercial insurance segments over the last several years was largely attributable to the impact of legal and regulatory reforms throughout the country in the early 1990s which caused us to reduce our estimate of ultimate loss costs associated with those coverages. As the benefit of those reform efforts became apparent in the mid-1990s, we reduced our prior year loss provisions accordingly. During 1998, there were no significant additional improvements in our hindsight view of ultimate loss provisions for workers' compensation coverages. As a result, while we still generated a reduction in prior year losses, it was not of the same magnitude as in the preceding several years.

With regard to medical professional liability coverages, the frequency and severity of claims can change suddenly. As a result, our reserving philosophy for this business, which has evolved over many years, requires that we wait for the prior year experience to mature before recognizing any significant favorable adjustments to prior year loss provisions. In contrast, when loss activity indicates an unfavorable change in the frequency or severity of claims our reserving philosophy requires that we respond quickly and cautiously. Such was the case in 1998, when loss activity indicated an increase in the severity of claims incurred in the years 1995 through 1997. Accordingly, we adopted a more cautious view of ultimate loss provisions for those years, resulting in a much smaller reduction in prior year loss provisions than those recorded in recent years.

Favorable development on assumed reinsurance also contributed to the reduction in prior year loss provisions in 1998, 1997 and 1996.