The debate over the benefits of mandatory seat belt laws and their enforcement status has focused on a controversial empirical enigma: why have these policies, which appear to have increased belt use sharply, had a relatively small impact on traffic fatalities? In this paper, I offer new insights into this question by examining panel data on observed belt use from the National Highway Traffic Safety Administration and self-reported data on belt use from pooled cross-sections of the Centers for Disease Control and Prevention's 1985-1993 Behavioral Risk Factor Surveillance System. By exploiting the panel nature of these data, I demonstrate that prior estimates, which have not conditioned on the unobserved time-varying determinants of belt use, have dramatically overestimated the impact of seat belt laws and their enforcement status on belt use. The true effects are more consistent with the modest impact these policies have had on traffic fatalities without having to appeal to the possibility of risk compensation by drivers. However, I find strong evidence in support of the selective recruitment hypothesis. Belt use among those most likely to be involved in traffic accidents (e.g. males, drinkers of alcohol, the young) has been significantly less responsive to seat belt laws and their enforcement status.