The first methodological condition requires controls for the more (less) risk-averse behavior of helmeted (unhelmeted) riders such as lower (higher) crash speeds and blood alcohol levels. Failure to include crash speed and blood alcohol levels in the analysis will incorrectly assign the injury reducing (increasing) effects of lower (higher) values of these variables to the use (non-use) of a helmet. As a result, the effect of helmets on head injury reduction is overstated. The Compton et al. paper confirms the risk-averse hypothesis by showing that helmeted riders are more likely to be insured, yet the study does not control for crash speed and blood alcohol levels.
The second methodological condition recognized that the engineering limits of helmet effectiveness -- helmets are only capable of absorbing a 13 mph impact before transferring impact forces to the contents of the helmet – must be incorporated in a regression analysis. This requires that the resulting reduction in helmet effectiveness be captured by an additional helmet variable -- a multiplicative interaction between the helmet use variable and the impact speed to the helmet.
Any regression analysis that excludes this interaction variable will overstate the effectiveness of helmets in reducing head injuries by ignoring the critical mechanism via which helmet effectiveness is reduced. This omission will also systematically understate the potential impact that helmets have in causing neck injuries -- past impact speeds of 13 mph , the helmet no longer absorbs/dissipates the impact forces and sets in motion a now heavier helmeted human head transferring those forces to the cervical spine.