Content area
Full Text
During the last 15 years, consultants employed by manufacturers in product liability cases and administrative proceedings have begun to base opinions of product defectiveness on comparisons between number of products sold and number of injuries reported to the manufacturer. In such analysis, the number of products sold is multiplied by an estimate of hours of use, which is sometimes factored by expected years of service. This number is then divided by the number of reported injuries to obtain an "accident frequency rate" (AFR) (McCarthy and Finnegan).
This rate is then compared with risks from common activities in order to obtain a value judgment regarding the product's safety (McCarthy and Finnegan--Appendix 1). One article states that any product which exhibits AFR below that reported for all industries and injuries by the National Safety Council (NSC) in 1962 is presumed "reasonably safe" (Barnett 153-173).
SUCH ANALYSIS IS ERROR-PRONE
However, uncertainties exist in all risk estimates and may arise during any phase of the analysis. Errors in the value assigned to number of units in service may lead to incorrect conclusions. In most estimates, certain fact are assumed rather than measured; these assumptions may be erroneous.
Error may also arise from assumptions about injury numbers. For example, to be reasonably accurate, an analyst must assume that the number of injuries reported is, in fact, the actual number that has occurred. Unfortunately, this is not often the case.
Few manufacturers have systems for obtaining data on injuries not resulting in lawsuits. In addition, courts have instituted specific limitations on the admission of negative evidence of injury (See Jones v. Pak-Mor Mfg. Co.). Without a program designed to elicit such injury data, analysts conducting risk estimates cannot assume that reported injury statistics are accurate.
Another avenue for error is the comparison of AFRs. For example, Barnett's analysis failed to show that NSC's all-industry AFR was a statistically valid measure. The assumption of validity was made without any apparent investigation, an oversight that limits the value of the conclusions. This...