I present evidence that social capital reduces traffic accidents and related death and injury, using data from a 10-year panel of 48 U.S. states. The econometric challenge is to distinguish the causal effects of social capital from bias resulting from its correlation with unobservable characteristics by state that influence road risks. I accomplish this by employing snow depth as an instrument, and by restricting attention to summertime accidents. My results show that social capital has a statistically significant and sizable negative effect on crashes, traffic fatalities, serious traffic injuries, and pedestrian fatalities that holds up across a range of specifications. (JEL R41, I18, Z13)