Key points are not available for this paper at this time.
Monthly data on road casualties in Great Britain are analyzed in order to assess the effect on casualty rates of the seat belt law introduced on January 31, 1983. Such analysis is known technically as intervention analysis. The form of intervention analysis that is used in this paper is based upon structural time series modelling and differs in significant respects from standard intervention analysis based upon ARIMA modelling. The relative merits of the two approaches are compared. Structural modelling intervention techniques are used to estimate the changes in casualty rates for various categories of road users following the introduction of the seat belt law. We first note the high rate of compliance with the seat belt law. By February 1983, the wearing rate had jumped to 90 percent and the rate has remained at approximately 95 percent from March 1983 onwards. There can be no doubt of the success of the law as regards to compliance. In considering the casualty figures we distinguish between those directly affected by the law, namely car drivers and front seat passengers, and those not directly affected by the law, that is car rear seat passengers, pedestrians and cyclists. Taking first numbers killed and seriously injured (KSI), we found a reduction of 23 percent for car drivers and 30 percent for front seat passengers. Thus, for those directly affected by the law, there have been substantial reductions. For rear seat passengers KSI we found a rise of 3 percent, for pedestrians a fall of one-half percent, and for cyclists an increase of 5 percent -- all three of these of these values being statistically insignificant. We conclude that there is no significant evidence of a change in numbers of KSI of those not directly affected. For numbers killed, we found for those directly affected a reduction of 18 percent for car drivers and 25 percent for front seat passengers. However, for those indirectly affected by the law, our model gave an increase of 27 percent for rear seat passengers, 8 percent for pedestrians, and 13 percent for cyclists. The value for rear seat passengers is highly significant and the other two values are on the borderline of significance. We conclude that there was an increase in fatalities of those not directly affected. We are unable to provide a completely satisfactory explanation of the difference between the figures for KSI and killed for rear seat passengers, pedestrians and cyclists. The article contains 13 pages of comments by discussants and 3 pages of authors' response. Language: en
Harvey et al. (Wed,) studied this question.