The third part of the state-of-the-art focuses on the future of road safety modeling and on conjectures concerning the evolution of national safety indicators. In the absence of econometric developments specific to road safety modeling, the research future must rely on pre-existing statistical procedures of econometrics applied to discrete/count and to aggregate data. In terms of contents, growing interest in the heterogeneity of road accident outcomes by category of victims could lead to treatments of this issue across research streams, say by top-down and bottom-up developments, but this speculation does not rest on extant adequate formulations of the issue of road user class and victim analysis. But understanding the time profile of aggregate national performance indicators is quite another matter.Concerning forecasting, a key question in countries where the absolute maximum of fatalities is still to come is that of its occurrence, but the answer requires a yet missing explanation of “the mystery of 1972–1973”, here hypothesized to result from the passing demographic wave (see Part 1). This ignorance affects the corresponding answer, in countries for which the maximum is long past, as to whether current performance is heading toward a minimum or toward a constant level: such a forecast can hardly be made if the maximum remains unexplained. In addition, it matters whether any envisaged asymptotic limit amounts to a natural rate combined with a random component, or includes more. It is conjectured that a regression component that would include speed, traffic density and vehicle occupancy rates could explain both the peak of 1972–1973 and the current evolution, notably of fatalities.In the absence of a certain explanation of the Meadow/Matterhorn/Cervin peak profile of the past maximum, forecasts can only combine random terms and known explanatory factors in the notion of Conditional Expected Safety Performance, which includes that of (Conditional) Expected Maximum Insecurity (EMI) and seems preferable to Vision Zero or to alternatives based on analogs of the natural rate of unemployment. Conditional expectations do not skirt the issue of the “level of the tide” by assuming the presence of an unexplained trend level and manually changing it by shifts due to well understood specific safety measures.Forecasts of explanatory variables require views on the political market (notably on the identity of the future median voter), on the workings of individual risk compensation, on the role of economic activity and on the chances of decoupling growth from transport demand, a weak prospect where communications appear more as gross complements than substitutes.
Read full abstract