Risky behaviours: Preferable to crashes for evaluating road safety mass media campaigns?

Decades of research have failed to establish whether or not mass media advertising can reduce road crashes. The probable reason is that the random variability in crash numbers is too great (and, campaigns being very cheap per person reached, even low effectiveness may be enough to be worthwhile). Three alternatives to before-after comparison of crashes as the method of determining effectiveness of an intervention are discussed. These are real-world experiments of high methodological quality, laboratory experiments of the social psychological type, and the measurement of safety-related behaviors. The third of these, before-after comparison of behaviors or variables that can be objectively observed and are closely linked to safety, is suggested as the most promising. However, the behaviors that might plausibly be used as proxies for crashes are quite few in number, and there is an urgent research need to find more of them, together with theory implying that a change in the behavior does indeed mean a change in safety.