This investigation addresses and evaluates: (a) the frequency of air bag deployments in comparison with belt only protection or no restraint, as a function of calendar year and model year; (b) injury and harm rates as a function of crash severity and restraint use and type; (c) restraint effectiveness in reducing fatalities and injuries as a function of restraint, crash severity, type of impact, and car size; (d) the confounding effects of crash severity; (e) injury patterns by injured body region, injuring contact, and injury severity; and (f) variation of injury patterns as a function of crash severity. It is found that restraints, irrespective of type, appear to be more effective at higher injury severities; the lowest casualty rates, and highest effectiveness values, are associated with the use of an air bag plus safety belt, or safety belt without air bag deployment; the air bag even without the concurrent use of a safety belt appears to offer a certain crash protection. Significant crash severity differences exist among the populations protected by various restraint types. These differences are such as to inflate the effectiveness of belt use only, and depress the effectiveness of the air bag, whether or not with belt use. When controlled for crash severity, the effectiveness of restraints in reducing fatalities, by comparison to no restraint, is found to be in all car crashes:(56.3 +/- 13.2)%, (25.4 +/- 11.8)%, and (48.5 +/- 5.5)%, for the air bag plus belt, air bag only, and belt only, respectively. The corresponding values in frontal crashes are:(76.3 +/- 11.8)%, (19.4 +/- 17.8)%, and (51.6 +/- 7.5)%. It is further found that the air bag plus belt reduces very significantly brain, spinal cord, facial, and abdominal injuries all across the board, at the expense of minor skin and flesh injuries. However the relative importance of a body-region/injuring-contact pair may vary by a large factor, depending on the crash severity range and restraint type under consideration.