April 15, 2005

Show me the evidence:

I spent some time on the phone and emailing with Steven Levitt back in 1999, so I think I can explain the origin of his abortion-cuts-crime theory pretty fairly. After he became a father for the first time, he started thinking about the huge number of legal abortions, tens of millions over the last several decades. (My impression is that he is more pro-life than pro-choice).

So, he thought to himself, That must have had some kind of effect on society. (Actually, plausible as that sounds, as his further research showed, most of those abortions were of fetuses that wouldn't have been conceived if abortion hadn't been legalized, so the actual effect on who is alive today isn't as large, or at least it's not as direct, as he'd originally assumed.)

So, he thought to himself: what's changed that's driving down the crime rate in the later 1990s. How about the legalization of abortion in the early 1970s? Criminals tend to be in their early 20s, so the timing seems right.

He looked at some crime data for 1985 and 1997, and noticed that crime had declined more on average in high abortion states like New York. So, he and John J. Donohue wrote up a draft paper suggesting abortion cut crime and started discussing it at academic conferences, where it got a respectful hearing. In August of 1999, the paper got leaked to the Chicago Tribune, which splashed it big.

It struck me when I read it as possible but not for certain, so I started looking into it. Greg Cochran pointed out that if you look at murder rates per year over the century, they go up and down a lot long before abortion was legalized. With that in mind, I started looking at the murder rates by age cohort and it quickly jumped out at me that they had gone through a vast upheaval between 1985 and 1997 that overwhelmed any effect related to abortion: namely, the crack wars.

Young men born in the years after legalization (1970 in NY and California, where the crack wars got started, 1973 in the rest of the country) became extraordinarily murderous in the late 1980s and early 1990s. You could hardly attribute the post-legalization cohort's better behavior in the late 1990s to abortion being legalized without also attributing to abortion their horrible behavior in the early 1990s. If there is an abortion effect, common sense says that it should impact people earlier in life, rather than later when all sorts of other factors have had more time to have an impact. But it was the post-legalization young who went on the worst youth murder spree in American history. In fact, you could make just about as strong a case that the legalization of abortion contributed to the murder spree by post-legalization youth.

Why did crime go down earlier in high abortion states? The cracks wars tended to burn out earliest in places where they got started earliest, which typically were high abortion cities that had had liberal politics, like NYC, LA, and Washington D.C. (where abortion was de facto legal from 1970 onward). Meanwhile, the crack wars spread in the 1990s to more conservative, low-abortion states in the hinterland, driving up the crime rate there.

This news came as a surprise to Levitt when we debated his theory in Slate in 1999, because he hadn't really thought about what happened in between his datapoints in 1985 and 1997, even though it was huge news at the time.

This is where the story gets mysterious. Rather than say, Oh, well, it was just an unpublished paper, Levitt kept on pushing his abortion-cut-crime idea, making that the most hyped element in his book Freakonomics, despite having lots of other material that he could have given the primary emphasis to instead.

He's never come up with simple answers to my challenges. He's instead upped the statistical complexity level of his explanations to the point where people generally feel they have to take his explanations on faith. He's nice guy, so lots of people decide to trust him rather than go through all the work of crunching the numbers for themselves. But, under the mild-mannered exterior, he does have a bit of a stubborn ego, which I guess is the solution to the mystery. The shame is that he's a bright guy and doesn't need this one theory to make his reputation.

So, here's some new data that I don't think has ever been published in a readable table before It's the FBI's homicide offending rates per 100,000, with columns being ages (even ages only) and the rows being the approximate birth years (for the full table, click here). Pick an age out and scan down and see if you can see if legalization (1970-1973) had an effect that you can notice. (For graphs of this data, click here.)

Homicide Offending Rates by Birth Year by Age 1980-2002


14 16 18 20 22 24 26 28 30
Approx. Birth Year 1958

29 24 18 19 18


24 20 18 19 18


31 25 22 21 18 20


27 23 21 17 19 20


28 27 22 19 18 19 17


27 25 21 22 18 21 16

16 24 25 24 22 20 18 15

14 24 25 26 22 21 20 15

1966 4 14 19 30 26 26 23 17 14

1967 4 14 19 31 28 25 21 18 13

1968 3 11 24 39 31 23 19 16 12

1969 4 13 26 39 37 26 17 13 12
CA, NY Legalized 1970 3 15 30 41 33 26 17 14 11

1971 3 16 37 53 37 26 16 13 11

1972 4 21 49 51 33 23 17 13 13
Roe v. Wade 1973 5 24 61 56 34 23 14 14

1974 6 32 57 52 33 21 18 15

1975 7 34 61 47 29 18 17

1976 7 36 62 45 28 20 19

1977 8 39 50 40 27 20

1978 8 38 48 40 26 22

1979 10 32 42 34 26

1980 9 25 35 35 27

1981 8 21 32 37

1982 6 15 29 35

1983 4 13 33

1984 3 11 25

I've put in bold the maximum murder rate for each age group. For 14 year olds, for example, the worst cohort was those born in about 1979 (so, their peak murder year was around 1993). For 16 year olds, 1977. For 18 year olds -- and the highest murder rate for any age -- was for those born about 1976 (peaking about 1994). For 20 year olds, 1973. For 22 year olds, 1969 and 1971. And so forth.

What you can see is that there were two murder peaks over the last 30 years. The first was the powder cocaine wars that peaked around 1980, when the killers tended to be in their 20s and older. The second was the vast crack cocaine wars that peaked around 1990-1994, and the killers became progressively younger as the wars went on.

The crack wars were fought much more by teenagers than the earlier crime waves, and most of those were born after legalization in their region, especially because of how much the early legalizing of abortion New York area dominated the crack wars in the early years. What made the crack years so murderous was the entry into the killing of so many teens -- exactly the generation that was theorized by Levitt to have been culled by abortion into law-abidingness.

Now, the abortion rate was higher in the later 1970s than right after legalization, so I suppose Levitt could argue that there just wasn't enough abortion after legalization for the Levitt Effect to work its magic. But, the abortion rate for blacks went up quite fast right after legalization (probably because of urbanization), and it was of course young black males whose homicide rate went through the roof during the crack wars. So, that's not a very persuasive argument. Black fetuses were getting aborted in very large numbers in the mid-1970s, but the survivors are exactly among whom the murder rate rose the most during the crack wars.

So, look at that table and if you can see the Levitt Effect, let me know because I sure can't.

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

No comments: