"Recall that Donohue and Levitt assume that the abortion ratio is zero in all 45 nonrepeal states in 1972. As noted above this assumption is extreme. Nevertheless, the absolute change in resident abortion rates between 1972 and 1973 in nonrepeal states is correlated with the level of the abortion rate in 1973. The weighted mean abortion rate in 1973 for states below the median is 6.6 abortions per 1,000 women 15 to 44 as compared to 16.0 in states above the median. If there is a “dose-response” effect of abortion on crime, then the effect of Roe v. Wade on arrest and homicide rates should be greater in absolute value for the states with greater abortion rates in the year immediately following Roe. I find no evidence of such an effect."