Multivariate Modelling of Accident Risk on the National Road Network
BERGEL R, INtl.ETS, France
The modelling of road safety indicators according to temporal ranges and applied to French data dates back to the 1960's. Various models have been developed in order to simulate the changes over time in the number of traffic accidents and victims on a dai
The modelling of road safety indicators according to temporal ranges and applied to French data dates back to the 1960's. Various models have been developed in order to simulate the changes over time in the number of traffic accidents and victims on a daily, monthly and yearly basis. These models were designed to be descriptive and explanatory, and they have brought to light various short, medium and long term effects (LASSARRE, 1994).
In this paper, we intentionally do not discuss modelling for long term trends, based on yearly data with variables representative of population and risk exposure, which is presently undergoing international comparisons within the framework of a COST group.
With regard to monthly data, ARIMA type models with exogeneous variables and intervention functions have successively been used in order to quantify the influence of different factors such as risk exposure, weather conditions, calendar configuration, and particularly to evaluate the effects of road security measures concerning speed limitation, and the use of helmets and seat belts. Data relative to practiced speed (LASSARRE et al., 1993) and seat-belt use (LASSARRE, PAGE, 1992) have been constituted over a long period, which permits the study of the relationship between these behavioural variables and the road safety indicators.
The most recent work on road safety monitoring has been the development of a model of daily road safety indicators (BERGEL et al., 1995), used at the Ministry of Transport by the National Intemainisterial Observatory for Road Safety. This tool for daily monitoring is also used on a monthly basis, after agregation of the daily data corrected for local effects due to an unusual meteorology or to the calendar configuration. It allows monthly comparisons to be made after seasonal adjustments, and medium term trends to be studied : for example, the effects due to the introduction of a point system for driver licenses in July 1992 were thus detected. However, this model does not explicitly take into account determining safety variables that are significant on an infra-annual basis, such as economic activity, amount of traffic, or driver behaviour.
We are now developing a more powerful econometric type of operational monitoring tool, for both short and medium term, which will measure the impacts of all of the following: economic variables, traffic level, meteorological variables, behavioural variables (speed, using or not using seat belts), and road safety measures. This tool will be used by decision-makers in order to consider the various determining factors of road safety and to quantify the short and medium term impact of different scenarios for road safety policies.
Association for European Transport