Variations in road accident counts: an explanatory time-series model

Abstract

Monthly traffic accident counts show a yearly recurring pattern. The number of accidents for some months always seems to be higher or lower than average. We investigate a range of factors that can explain this variation in monthly accident counts. Apart from legal, economic and climatologic factors, monthly accident statistics are subject to calendar effects and seasonal variation. The number of accidents may vary with the number of days per month or with the number of times each day of the week occurs. Accident counts are also affected by seasonal variation. Factors like weather and holidays vary strongly with the seasons, leading to a changing level of traffic safety. Although these factors have an influence on the reported accident counts, they can never be changed to increase traffic safety and should therefore be taken into account when modelling traffic accidents. We include their effects on Belgian road accident counts in an explanatory time-series model that can be used to predict traffic safety.