We examine whether abortion removes from the population those infants most at risk of homicide. As part of our identification strategy, we find that abortion reduces the number of unwanted births, estimating that 1 percent increase in the abortion ratio reduces unwanted births by approximately 0.35 percent. Using cross-sectional time-series data for U.S. states between 1970 and 1998, we find that an increase in the abortion ratio (a proxy for unwanted births) reduces the expected number of infant homicides, especially among black infants. Overall, the elasticity of infant homicides with respect to unwanted births is approximately 0.089.