Forecasting product liability claims. Epidemiology and modeling in the Manville asbestos case. Foreword by the honorable Jack B. Weinstein. (Q703137)

The Johns-Manville Corporation had been the largest supplier of asbestos products, with an estimated US market share of between 25 and 40 percent. A stream of law suits forced the company into bankruptcy in 1982. As part of the bankruptcy agreement, a two billion dollar trust (The Manville Trust) was established in 1982 and charged to provide compensation for personal injuries. By the end of 1989 the Trust had received 140,000 claims and new claims were arriving at a rate of 17,000 per year. With settling costs averaging US \$ 41,000 per claim, the Trust was quickly running out of money. A panel of independent experts was established in 1991, with the charge to develop statistical models for the reliable projection of the number, timing, and nature of future claims. Accurate projections were needed to assure a fair and equitable distribution of the Trust's limited funds. On the basis of their predictions, the panel recommended in 1994 a distribution schedule that limited the Trust to pay claims at a pro rata rate of \$ 0.10 on the dollar. This book summarizes the statistical models for projecting the number, timing, and nature of future claims, and it discusses how these predictions are used in developing a fair and equitable distribution of insufficient funds. Chapter 2 provides background on asbestos epidemiology and a review of the health risks of asbestos. Chapters 3 through 5 include a detailed assessment of two prior models for projecting the current and future numbers of workers with various asbestos-related diseases. The model by Selikoff combines direct estimates of the size and composition of the asbestos-exposed labor force with risk measures from a variety of case-control and cohort studies of the cancer risks of occupational exposure to asbestos. The model by Walker infers these numbers indirectly from recent incidence data by computing the currently living person-exposure equivalents first exposed at various ages and years in the past. Chapter 6 extends earlier work by Selikoff, Walker and others and proposes a more realistic projection model based on the indirect estimation of past exposure. Chapter 7 provides extensive sensitivity studies and plausible bounds for the uncertainty of the projections. Chapters 8 and 9 incorporate into the model a component that addresses the depletion of the population eligible to file a claim due to prior claim filings. This is the model that was forwarded to the Court in 1994. The last chapter compares the projected Trust experience for the period 1990-1999 with the actual observed claims data. This book provides a fascinating account of how prediction models are used to solve a very-real problem. It makes wonderful reading for statisticians interested in prediction problems, epidemiologists, actuaries, and lawyers involved in product liability suits and having to predict the number of possible litigants.