An investigation was conducted to estimate the effectiveness of air bags as a function of velocity. The study consisted of three parts: a theoretical idealization, an analysis of National Automotive Sampling System/Crashworthiness Data System (NASS/CDS), and a reanalysis of previously published Fatality Analysis Reporting System (FARS) data. The theoretical analysis looked at idealized risk curves as a function of velocity; assuming that the air bag offers a benefit for both belted and unbelted occupants. Analysis of the NASS/CDS data looked at the effectiveness of air bags as a function of velocity for Maximum Abbreviated Injury Scale (MAIS) 3+ injuries. The reanalysis of the previously published FARS data looked at the effectiveness of the air bag as a function of velocity for fatalities. The theoretical analysis indicates that the air bag effectiveness should be greatest at the low velocities. The field data analysis of both NASS/CDS and FARS were consistent with the theoretical analysis, indicating that air bags are most effective at the lower velocities, below 40 kph (25 mph), for both belted and unbelted occupants. Although it was not possible to estimate a different effect for belted and unbelted for fatalities using FARS, it was possible for MAIS 3+ using NASS/CDS. For unbelted occupants the effectiveness goes to zero or becomes negative above 40 kph (25 mph) for MAIS 3+, and for belted occupants the effectiveness stays positive but with significantly lower magnitude for speeds above 40 kph (25 mph).