This report shows how traffic safety data can be analysed to understand the development of traffic safety over time. The report shows a number of principles of time series analysis:
- The current state of the art of structural time series analysis is described. We illustrate the importance of changes in distance travelled, and how this affects the number of accidents or the number of fatalities. Preferably, current methods simultaneously analyse the development of distance travelled, and of the number of accidents or fatalities, so as to allow for robust forecasts.
- Further understanding of the development of traffic safety asks for disaggregation of the safety data into different subgroups. Such disaggregation reveals different trends for different groups. An analysis of the development over time of mortality (fatalities per inhabitant) is given for different European countries. This is compared to motorization rate. The analysis shows how motorization rate and mortality relate, and what different patterns evolve when countries are compared.
- The development of traffic safety is shown to depend strongly on the composition of the population of a country. It is well known that young, unexperienced drivers are at higher risk, especially male drivers. Changes in the demographics are related to safety, and forecasts are sensitive to this composition. We show how to use population data to analyse and forecast the number of fatalities in a country.
- Eventually, models are expected to incorporate other influencing factors or safety performance indicators, such as the quality of roads, fraction of safety belts used etc. At this stage, operational models of that type are a bridge too far, mostly, because most influencing factors are very specific. As an example of a factor that affects all accidents, we chose weather parameters to incorporate in our models. We use two different techniques to analyse the effect of precipitation, frost and temperature on road safety.
Part one of the report gives a bird’s-eye view of the time series analyses that were carried out in the final year of SafetyNet. We did our best to present the main results in such a way that non-technical readers can easily follow the lines of thought. The main lines of reasoning are given and explained, and the highlights of the results are presented.
Part two consists of four separate contributions, each giving a more detailed description of the objectives, methods and results of each contribution. This part is more technical, but still sufficiently accessible to non-technical readers to get the grip of the techniques described.
Part three is a list of appendices, where technical and mathematical details can be found, or large sets of graphical results are presented. This part is especially meant for the very interested or adepted readers.