Case control studies suggest that cycle helmets offer their wearers protection from injury in the event of an accident. Nevertheless, encouragement and even compulsion of cycle helmet wearing has been controversial. This paper will re-examine another potential source of evidence for the role of cycle helmets. Administrative datasets are attractive because of their availability, but require careful analysis. The results presented here are obtained from analysing such data with an appropriate form of generalised additive model. Whilst helmet wearing surveys in the UK suggest strongly divergent trends in wearing rates between male and female children, there is little evidence from “Hospital Episode Statistics” to indicate similarly divergent trends in terms of head injury. Conversely, it can be confirmed that head injuries are falling faster among cyclists than pedestrians. Although case control studies suggest cycle helmets are not effective in reducing overall injuries, it is worth noting an increase in the proportion of male child cyclists reported by the police as being killed or seriously injured in road collisions. It might be tempting to use these results to suggest that helmets are not effective in reducing head injury at the population level. Whilst the careful analysis of population level data presented here is clearly important, this paper will discuss the reasons why population and individual level analyses of cycle helmets might be different and consider some of the difficulties in assigning cause and effect with imperfect observational data.