The daily series of the number of accidents and victims on the road in France from 1985 to 1992 have been broken down into three additive components: a short-term meteorological effect, a calendar effect and a residual series corrected for these two effects. The double adjustment for the short-term meteorological effect and the calendar effect represents a local correction, appropriate to the varying context of road safety. The data thus corrected for local effects are then aggregated on a monthly basis, and seasonally corrected by a standard method, applied by means of the Census-X11 program. The final seasonally corrected values result from a local adjustment, within each month, followed by a monthly correction applied over the year. The detailed daily results on the number of fatalities in France apply to the second 4-month period of 1992, which saw the introduction effective from July 1992 of a new point system against drivers at fault. Detailed monthly results cover the whole period modelled from 1985 to 1992. (A)
Abstract