July 19, 2005

Levitt's state-by-state abortion-crime correlations

Something that hasn't been mentioned is what an uncertain reed the Freakonomics abortion-cut-crime theory is largely based upon: correlations between the abortion rates by state in the 1970s and the crime rates by state in the 1990s. Beyond all the other problems I've noted (such as the correlations only work for the decline in crime in the mid-1990s, and instead are reversed during the huge crime increase in the late 1980s and early 1990s), a massive weakness in his analysis is that people move. After two decades, a large fraction of the population is living in a different state. Even worse for Levitt's assumption, the odds that a grown child will be living in a different state than his mother was living in when she was pregnant with him are even higher.

Now, if all this movement was purely random, utterly unrelated to crime and abortion laws, then that would simply make any connection between abortion in the 1970s and crime in the 1990s harder to detect. But, we have good reason to assume that interstate migration is driven in part precisely by crime rates and by the general moral climate (of which the abortion rate is symbolic).

This is an enormously complicated subject, but let me give one example. Consider New York and California, which Levitt repeatedly points to as states that legalized abortion early (in 1970) and had crime fall-offs earlier in the 1990s. If you look at white people, you'll see that NY and California continued to attract affluent whites who could afford to insulate their kids from crime and moral decay, while they shed large numbers of less well-to-do whites who were worried that they couldn't afford to provide their kids with a good upbringing.

So, comparing the white populations of NY and California in the 1970s versus the 1990s is a classic apples and oranges comparison, and within each state, too. Among whites, the populations became increasingly affluent, older, and the family sizes shrunk. In contrast, the white populations of socially conservative (and thus low abortion) states became relatively younger and less affluent and thus more crime prone.

But that's just one dynamic. When you throw in the significant differences in the racial makeup of states over time, it all gets extremely complicated. But the key point is that there is no reason to assume that you can make safe apples to apples comparisons of the populations in states in the 1970s versus 1990s, as Levitt assumes you can.
My published articles are archived at iSteve.com -- Steve Sailer

No comments: