teddybears - I'm not sure accidents everywhere can be on a long-term downward trend while also regressing to the mean, that doesn't make any sense.
Also, while individual sites may have a mean accident rate (one of the arguments used to question the efficacy of fixed speed camera sites - that they were placed after an unusually high run of accidents at particular spots that would have reduced naturally, subsequently proven to be wrong) I'm not sure that can really apply to a whole city, as there are simply too many factors involved.