A Simple Test of Abortion and Crime

I replicate Donohue and Levitt’s results for violent and property crime arrest rates and then apply their data and specification to an analysis of age-specific homicide rates and murder arrest rates. The coefficients on the abortion rate have the wrong sign for two of the four measures of crime and none is statistically significant at conventional levels. In the second half of the paper, I present alternative tests of abortion and crime that attempt to mitigate problems of endogeneity and measurement error. I use the legalization of abortion following the 1973 decision in Roe v. Wade in order to exploit two sources of variation: between-state changes in abortion rates pre and post Roe, and cross-cohort differences in exposure to legalized abortion. I find no meaningful association between abortion and age-specific crime rates among cohorts born in the years just before and after abortion became legal.