Abortion Legalization and the Crime Rate

Not surprisingly, Steven Levitt and Stephen Dubner's new book "Freakonomics" ignores their academic critics, but Steve Landsburg's review disappointingly does so too (Leisure & Arts, April 13). Take just the book's first claim: Unwanted children are more likely to grow up to be criminals and that abortion can therefore reduce crime, a plausible idea that has been around since the beginning of the abortion debate. Yet, despite Messrs. Levitt and Dubner's claims, legalization doesn't explain 75% of the drop in murder rates during the 1990s, and if anything the reverse is true.

Their data had a serious error. The Planned Parenthood affiliated organization that supplied them with the data incorrectly claimed that when abortion was legalized during the late 1960s and early 1970s, states went from a complete ban to complete legalization, but abortions had been allowed before complete legalization when the life or health of the mother was endangered. The Centers for Disease Control data show that before Roe v. Wade many states that had allowed abortions only when the life or health of the mother was endangered actually had higher abortion rates than states where it was completely "legal."

If Messrs. Levitt and Dubner were correct, crime rates should have first started falling among younger people who were first born after legalization. Only as they aged would you start seeing crime fall among older criminals. But in fact the precise opposite is true. Murder rates during the 1990s first started falling for the oldest criminals and very last for the youngest.

John R. Lott Jr.
Resident Scholar
American Enterprise Institute
Washington

Levitt has since responded that three of his regressions in the 2004 paper with Donohue use the Centers for Disease Control abortion numbers. The problem with this as Levitt presumably knows is that these are not the disaggreated regressions that directly link abortions for when a particular age group was born with the murder rate when those people were older. This disaggregated type of data was what Whitley and I were studying, and it is the paper with Whitly that I was referring to. Even those regressions have problems since he uses the arrest data from the Uniform Crime Reports and not the Supplemental Homicide Data that more directly links the murders with the age and other characteristics of the offender, but the three regressions that Donohue and Levitt look at only use his poorly constructed and unable to be duplicated "effective abortion rate" data that is an average of abortions across all ages using weights from 1985 (unexplained why only one year is used to construct these weights, though my paper with Whitley shows how sensitive these weights are to the year used).