This paper resents statistical analyses of accident data obtained both before and after a speed limit reduction, in order to ascertain the effect of time on the efficiency of the speed limit. The data is taken from the 1976 to 1985 Zurich Yearbooks (Statisticsches Amt der Stadt Zuerich 1976-1985). The speed limit was reduced in July 1980 from 60 to 50 km/hr, and thus the data cover a time span from five and a half years before to four and a half years after the speedreduction was introduced. The following methods are described: (a) a plot of "seasonal" changes in injury accident with time; (b) a plot of the series of seasonal differences; (c) a cross correlation between a reference signal and the series of the residuals of an ARIMA model; and (d) time pattern of the residual variances of successive intervention models. For most of the series investigated, the most likely time for the speed limit to be effective is three months before its introduction. A possible explanation of this result may be that the media informed the public of the speed limit before it was introduced. This finding has led to the production of an improved parsimonious intervention model which can distinguish between a possible "pre-measure" effect and the usual " post-measure" effect.
Abstract