Most empirical studies have estimated a positive union-nonunion "injury gap," suggesting that unionized workers are more likely to have a nonfatal occupational injury than their nonunion counterparts. Using individual-level panel data for the first time, I study several explanations for this...