The purpose of this paper is to show that time series analyses of road safety and risk can be improved by using demographic data. We demonstrate that the distance travelled by drivers or riders of a certain age reflects the fluctuations over the years of the number of people of that age within the population. We further demonstrate that the change over time of per capita distance travelled, i.e. distance travelled per person, is often less subject to stochastic fluctuations, and therefore more smooth than the total distance travelled for drivers of that age. This smoothness is used to obtain forecasts of distance travelled, or to average out year-to-year fluctuations of data of distance travelled. Analysis of such data stratified by age group, gender or both reveals that, for most travel modes, per capita distance travelled is to a large extent constant or slowly changing over time. The consequences for the evaluation of risk, i.e. casualties per distance travelled, with and without the use of population data, are explored. Dutch data are used to illustrate the model concept. It is shown that the analyses and forecasts of distance travelled could gain substantially by incorporating demographic data, as compared to an analysis with data of distance travelled alone. The paper further shows that, for an analysis of risk and therefore for traffic safety forecasts in the absence of any data of distance travelled, stratified analysis of mortality, i.e. casualties per inhabitant, may be a reasonable alternative. © 2012 Elsevier Ltd.