Developing a driving Safety Index using a Delphi stated preference experiment.

Whilst empirical evidence is available concerning the effect of some aspects of driving behaviour on safety (e.g. speed choice), there is scant knowledge about safety thresholds, i.e. the point at which behaviour can be considered unsafe. Furthermore, it is almost impossible to ascertain the interaction between various aspects of driving behaviour. For example, how might drivers' lateral control of a vehicle be mediated by their speed choice-are the effects additive or do they cancel each other out. Complex experimental or observational studies would need to be undertaken to establish the nature of such effects. As an alternative, a Delphi study was undertaken to use expert judgement as a way of deriving a first approximation of these threshold and combinatory effects. Using a stated preference technique, road safety professionals make judgements about drivers' safe or unsafe behaviour. The aim was to understand the relative weightings that are assigned to a number of driver behaviours and thereby to construct a Safety Index. As expected, experts were able to establish thresholds, above (or below) which changes to the behavioural parameters had minimal impact on safety. This provided us with a Safety Index, based on a model that had face validity and a convincing range of values. However, the experts found the task of combining these driver behaviours more difficult, reflecting the elusive nature of safety estimates. Suggestions for future validation of our Safety Index are provided.