Auto Insurers and the Airbag: Reply

Abstract

Peterson and Hoffer (1996) provide interesting evidence of offsetting behavior and other useful data in their comment on our article (Kneuper and Yandle, 1994). We welcome their effort test hypotheses related our research on airbags. Their comment provides an opportunity for us clarify a key argument in our piece and note the linkages between our argument and those of Peterson and Hoffer. Our article describes airbag technology as a means for auto insurers to change the nature of the automobile hazard, effectively alter the nature of losses and convert uncertainty of future claims into (p. 108). We discuss the possibility that the dollar value of property claims might rise if airbags shift personal injury losses property losses, but note that predictability is what matters, not the magnitude of claims. It is in this context that we call attention Peltzman's work on offsetting behavior, noting that, under our theory, insurers prefer predictable risk uncertainty even if improved auto safety induces riskier behavior (footnote 7, p. 113). The ability set accurate premiums matters most. Given a population of seatbelt-equipped cars, insurers cannot ex ante predict accurately the behavior of drivers. Therefore, ex post variance or risk associated with claims will be higher. Risk management requires risk estimation. Mandatory airbags move the population of autos in the direction of homogeneous exposure units and convert uncertainty, which cannot be estimated, risk, which can. Following our discussion, Peterson and Hoffer see the possibility that conversion airbags will change the frequency of claims as between personal injury and property losses. They test the hypothesis that the ratio of claims for property loss personal injury claims will be higher for a population of airbag-equipped cars. They find that the frequency for both categories of claims rises and present suggestive evidence that property loss claims may rise faster than personal injury claims. The tentative evidence could be seen as being consistent with our argument about loss substitution.