Rates of serious crime in the United States dropped greatly throughout the 1990s for virtually all offenses. John Donohue and Steven Levitt have argued that this reduction relates strongly to the 1973 "Roe v. Wade" decision that legalized the abortion of unwanted pregnancies. If such pregnancies result in children with higher lifetime risks of criminality, then the greater ability to terminate these pregnancies after 1973 should reduce crime rates. The purpose of this article is to empirically assess Donohue and Levitt's basic premise that unwanted pregnancies result in children with significantly higher risk for law-violating behavior. This analysis addresses two questions. First, do children born of an unwanted pregnancy become more highly involved in juvenile delinquency during adolescence and criminal behavior during early adulthood? Second, do the consequences of unwanted pregnancies depend upon the social and demographic characteristics of the child and family? The authors address these two questions with panel data gathered from a national sample of children born prior to "Roe v. Wade". (Contains 9 notes.)