broken windows theory

Fooled again! Pinker puts a nail in the coffin of the Freakonomics crime theory?

This keeps happening to me. Every time I pick up a social science book, my understanding of the dramatic crime drop in the ’90’s changes (all three books I reference are pre-2011, so apologies if I’m late to the party and you already know all this).

First there was Malcolm Gladwell’s Tipping Point, which postulated that the “Broken Windows Theory”  (the idea that policing minor crimes like vandalism has an effect on major crime as well) was responsible. That sounded mildly convincing, until I read Dubner and Levitt’s Freakonomics a few weeks later, which argues that legalized abortion actually correlates quite well with the drop in crime (simply, the mothers in environments most likely to have children prone to violence later in life started having abortions in the 70’s when it was legalized, thus there were fewer potential delinquents around in the 90’s). That seemed to blow Broken Windows out of the water (see earlier post here), and I remained convinced of the abortion correlation until coming to a section in Steven Pinker’s epic, The Better Angels of Our Nature, a few days ago. Now, I’ve got to say, I’ve switched again.

Pinker rebuts quite convincingly the theory popularized in Freakonomics, and puts forward two overarching explanations but admits that the decline “likely had multiple causes, and no one can be certain what they were, because too many things happened at once.” Since this is the third time my opinion has been swayed, I’m perfectly happy to remain agnostic on the issue. I am reproducing Pinker’s entire section below that refutes the abortion theory (without permission…hope he doesn’t mind) because I can’t find a thorough enough excerpt on the web. Do consider reading the entire book, however. It’s one of the best I’ve ever picked up.

From Steven Pinker’s The Better Angels of Our Nature, hardback, page 119-121 (without annotations):

The 1990s crime decline inspired one of the stranger hypotheses in the study of violence. When I told people I was writing a book on the historical decline of violence, I was repeatedly informed that the phenomenon had already been solved. rates of violence have come down, they explained to me, because after abortion was legalized by the 1973 Roe v. Wade U.S. Supreme Court decision, the unwanted children who would ordinarily have grown up to be criminals were not born in the first place, because their begrudging or unfit mothers had had abortions instead. I first heard of this theory in 2001 when it was proposed by the economists John Donohue and Steven Levitt, but it seemed too cute to be true. Any hypothesis that comes out of left field to explain a massive social trend with a single overlooked event will almost certainly turn out to be wrong, even if it has some data supporting it at the time. But Levitt, together with journalist Steven Dubner, popularized the theory in their bestseller Freakonomics, and now a large proportion of the public believes that crime went down in the 1990s because women aborted their crime-fated fetuses in the 1970s.

To be fair, Levitt went on to argue that Roe v. Wade was just one of four causes of the crime decline, and he has presented sophisticated correlational statistics in support of the connection. For example, he showed that the handful of states that legalized abortion before 1973 were the first to see their crime rates go down. But these statistics compare the two ends of a long, hypothetical, and tenuous causal chain – the availability of legal abortion as the first link and the decline in crime two decades later as the last – and ignore all the links in between. The links include the assumptions that legal abortion causes fewer unwanted children, that unwanted children are more likely to become criminals, and that the first abortion-culled generation was the one spearheading the 1990s crime decline. But there are other explanations for the overall correlation (for example, that the large liberal states that first legalized abortion were also the first states to see the rise and fall of the crack epidemic) and the intermediate links have turned out to be fragile or nonexistent.

To begin with, the freakonomics theory assumes that women were just as likely to have conceived unwanted children before and after 1973, and that the only difference was whether the children were born. But once abortion was legalized, couples may have treated it as a backup method of birth control and may have engaged in more unprotected sex. If the women conceived more unwanted children in the first place, the option of aborting more of them could leave the proportion of unwanted children the same. In fact, the proportion of unwanted children could even have increased if women were emboldened by the abortion option to have more unprotected sex in the heat of the moment, but then procrastinated or had second thoughts once they were pregnant. That may help explain why in the years since 1973 the proportion of children born to women in the most vulnerable categories – poor, single, teenage, and African American – did not decrease, as the freakonomics theory would predict. It increased, and by a lot.

What about differences among individual women within a crime-prone population? Here the freakonomics theory would seem to get things backwards. Among women who are accidentally pregnant and unprepared to raise a child, the ones who terminate their pregnancies are likely to be forward-thinking, realistic, and disciplined, whereas the ones who carry the child to term are more likely to be fatalistic, disorganized, or immaturely focused on the thought of a cute baby rather than an unruly adolescent. Several studies have borne this out. Young pregnant women who opt for abortions get better grades, are less likely to be on welfare, and are more likely to finish school than their counterparts who have miscarriages or carry their pregnancies to term. The availability of abortion thus may have led to a generation that is more prone to crime because it weeded out just the children who, whether through genes or environment, were most likely to exercise maturity and self-control.

Also, the freakonomists’ theory about the psychological causes of crime comes right out of “Gee, Officer Krupke” when a gang member says of his parents, “They didn’t wanna have me, but somehow I was had. Leapin’ lizards! That’s why I’m so bad!” And it is now as plausible. Though unwanted children may grow up to commit more crimes, it is more likely that women in crime-prone environments have more unwanted children than that unwantedness causes criminal behavior directly. In the studies that pit the effects of parenting against the effects of the children’s peer environment, holding genes constant, the peer environment almost always wins.

Finally, if easy abortion after 1973 sculpted a more crime-averse generation, the crime decline should have begun with the youngest group and then crept up the age brackets as they got older. The sixteen-year-olds of 1993, for example (who were born in 1977, when abortions were in full swing), should have committed fewer crimes than the sixteen-year-olds of 1983 (who were born in 1967, when abortion was illegal). By similar logic, the twenty-two-year-olds of 1993 should have remained violent, because they were born in pre-Roe 1971. Only in the late 1990s, when the first post-Roe generation reached their twenties, should the twenty-something age bracket have become less violent. In fact, the opposite happened. When the first post-Roe generation came of age in the late 1980s and early 1990s, they did not tug the homicide statistics downward; they indulged in an unprecedented spree of mayhem. The crime decline began when the older cohorts, born well before Roe, laid down their guns and knives, and from them the lower homicide rates trickled down the age scale.