It’s time again to drag out America’s new favourite dataset couple into the limelight – Abortion and Murder! To be specific, I’m going to talk a bit about the age-based criticism of the Levitt and Donohue paper that Steve Sailer has been pushing for some time now.
First, a short recap: Sailer’s critique is based around the fact that the age group 14-17, that started enjoying the crime-stopping benefits of abortion rights in 1987 or so, responded to this improvement by going on a significant murder spree. (A quickie regression between cumulative abortion impact for the age group and the 14-17 murder rate yields a negative adjusted R2) Levitt explains this phenomenon by pointing to the Crack Wars of the early 1990-ies. Sailer, in turn, is fond of using this graph to support his point:
Now, while this certainly is a bump in the road for Levitt and Donohue, they can stage a strong comeback by pointing to the 18-24 age group instead, where the magic of OLS gives them plenty more to write home (or to a top journal) about.
Especially when controlling for a few variables like the incarceration rate, the cumulative abortion impact of the 18-24 group yields a big R2, as well as significant coefficients.
There are two major problems remaining, however, in addition to the murderous teen thing. Both problems are visible in the graph below:
Problem one: After the crack epidemic subsided, the 18-24 group settled down on pretty much the level of murderousness displayed during the (largely abortion-impact-free) period of 1970-86. This is the same pattern as among the 14-17 year-olds, where the “abortion effect” could not be detected. This could be a mere coincidence, of course. If abortion had not been present, the 18-24 year-olds might very well have continued to murder each other at unprecedented rates, due to some unknown force unrelated to the crack wars, while the 14-17 age group would have returned to normalcy. This, however, leads us to problem number two.
Problem two: The only real murder reduction in the US has taken place among the older generations – the 25 years and above category has seen its murder rate decline since the beginning of the 1980-ies. These generations are the ones that are largely untouched by abortion. (The first abortion-improved cohort entered the 25+ category in 1998)
Finally, my own take, really short version: The crack epidemic made youths more murderous than usual, and when it subsided they went back to murdering each other at normal rates. Meanwhile, ever-increasing incarceration rates made it a lot harder to stay on the street in the long run as a career criminal, and thus, the oldies got to murder each other less than in the good ol’ days. Levitt’s abortion-crime relationship for the 18-24 age group is pure coincidence.
PS.
So, what point am I trying to make, aside from the ones Sailer has already made?
What I found interesting is that the “Levitt effect” really appears to exist at first glance, but that it is confined to 18-24 year-olds. I don’t find that correlation very convincing in terms of being the cause behind the murder decline, however, because of the reasons stated above.
Second, I really don’t think Sailer pushes the most striking factoid hard enough: Murders have only declined among older people, while all youth groups – including the big-R-squared 18-24 gang that Levitt depends on – show constant murder rates, if you ignore the crack spike. Instead, Sailer focuses more on the 14-17 discrepancy, but I think the fact that there really is no murder reduction *at all* to show for all those abortions is a lot more striking, and thus likely to change minds. If I am incorrect in these views, link away , and I will update / correct / delete. (I am a regular Sailer reader and a fan, after all)
Posted by dobeln at 05:14 AM