Slapdash Steve Levitt Rides Again: Abortion and Infanticide

Slapdash Steve Levitt Rides Again: I'm starting to realize that you can't trust anything economist Steven D. Levitt says in his bestseller Freakonomics without checking it out on Google for yourself. A reader called my attention today to Levitt's statement on p. 139:

To be sure, the legalization of abortion had myriad consequences. Infanticide fell dramatically.

When I originally read this a few months ago, I thought to myself: "How could anyone possibly doubt that?"

Well, I should have realized that if Levitt tells you the sun is coming up in the East, you'd better get outside and check for yourself. Here is what Child Trends Databank had to say about homicides of infants (below age 1):

The infant homicide rate increased from 4.3 per 100,000 in 1970 to 9.2 per 100,000 in 2000, before falling to 7.8 per 100,000 in 2003 (preliminary estimate). In 2003, 318 infants died due to homicide.

Their graph shows that infanticide increased from a 4.3 rate in 1970 (when there were only about 200,000 legal abortions) to 5.9 in 1980 (when there were about eight times more abortions)!

Moreover, when infanticide is looked at by ethnic group for year 2002, there's a positive correlation between the abortion rate and the infanticide rate.

The FBI statistics for homicides of children under 5 only go back to 1976, so nothing too definitive can be seen here, but you'd expect to see, according to Levitt's statement, a decrease in the white numbers over the first half decade as the white abortion rate continued to rise. Instead, the small child homicide numbers did not fall. Indeed they finally peaked in 1996, and the black numbers peaked in 1993.

Then I looked up the abstract of the paper that Freakonomics cites on p. 223 as the source for the contention that "Infanticide fell dramatically." Here's what the authors of that study actually say:

We examined 1960-1998 U.S. mortality data for children under 5 years of age using an interrupted time series design. The legalization of abortion was not associated with a sudden change in child homicide trends. It was, however, associated with a steady decrease in the homicides of toddlers (i.e., 1- to 4-year-olds) in subsequent years. Although in the predicted direction, the decrease in homicides of children under 1 year of age was not statistically significant.

My published articles are archived at iSteve.com -- Steve Sailer