When did a reduced speed limit show an effect? Exploratory identification of an intervention time

In a statistical analysis of accident data before and after a speed limit reduction, the time of the countermeasure is, of course, well known. Our understanding of the accident process may, however, be increased if we assume in a thought experiment that this time is unknown. We ask if the data themselves can tell us something about such a possible time. By means of time series of traffic accidents in Zurich before and after a speed limit reduction, different exploratory methods are presented to identify the "unknown" time of this measure. For most of the investigated series, the most likely time was found to lie in the three months before the true introduction. A possible explanation of this result may be that the media already informed the public before the countermeasure was actually introduced. This finding leads to an improved parsimonious intervention model which distinguishes between a possible "preintervention effect" and the usual "intervention effect."