April 4, 2006

Why crime has fallen sharply in New York City

Somehow, I don't think that in their debates over why crime has fallen so sharply in New York City, Malcolm Gladwell and Steven D. Levitt are going to get around to mentioning this new article in the New York Times, "New York City Losing Blacks, Census Shows:"

"Reversing a tide from the South who altered the complexion of the city earlier in the 20th century, the number of American-born blacks leaving the city has exceeded the number arriving since at least the late 1970's."

Most of the statistics in the article understate the relevant size of the outflow of African Americans because they talk about total blacks, not African Americans. Until recently, the outflow of high crime rate African-Americans from New York City was balanced off by an influx of lower crime rate black immigrants from Africa and the Caribbean.

In 1997, according to a liberal advocacy group, the National Center on Institutions and Alternatives, non-Hispanic blacks in New York state were imprisoned 12.8 times as often per capita as non-Hispanic whites.

From this, you can see one reason why the media elite in New York City is so rabidly pro-immigration. When foreign blacks and Mexicans squeeze African-Americans out of New York City, in what I call "economic ethnic cleansing," the NYC crime rate falls. The chance that David Brooks is going to get mugged goes down. However, these African-Americans aren't, as New York City pundits often seem to assume, being deported, they are just moving to fly-over country and bringing their problems with them.

Newhouse News reporter Jonathan Tilove has pointed out another reason that probably contributed to the exceptional crime decline in NYC, the striking shortage of black men: "There are 36 percent more black women than men in New York City." Some of the reasons are obvious: imprisonment, the huge number of murders during the crack years, AIDS, and the military. But there has also been some sex-selective immigration, as parents move to keep their sons off the streets. The article quotes a black lady from NYC who moved to a small town in North Carolina:

"I was divorced and moved here with my 11-year-old — I was afraid of the crime, and black boys don't fare too well in New York."

All this migration is relevant to the Levitt-Donohue theory that legalizing abortion cut crime because ever since I pointed out in my 1999 debate in Slate.com with Steven D. Levitt that his theory was based on insufficiently precise age groups (under 25 and over 25), and that when you look at age groups (e.g., 14-17) sufficiently precise enough to register the impact of abortion, then what you instantly see is that at the national level, the opposite of what they had predicted actually had happened. All Levitt could do was claim that his state-level data vindicated his theory. I didn't have the econometric skills to evaluate Levitt's claim, but when Boston Fed economists Christopher Foote and Christopher Goetz worked through Levitt-Donohue's state level analysis in 2005, they found two fatal mistakes in it that wiped out the whole effect.

In response, Levitt and Donohue pulled out a whole new database of state-level abortion data and announced that this vindicated them. At the AEI conference on the topic last week, Levitt's co-author John J. Donohue indignantly kept saying that my national-level graphs showing the exact opposite happened didn't disprove his theory. Which is true, but I never claimed that it disproved it, just that journalists should have done some simple reality checks rather than accept Donohue and Levitt's claim on faith. Donohue made a hilarious effort to hand the burden of proof off to the skeptics, which is exactly where it does not belong. You can watch the video here.

This NYT article, however, illustrates many of the dangers of the Donohue-Levitt methodology of looking at state abortion rates and state crime rates with a lag of approximately two decades between them.

The obvious problem is that people move. If movement was just random, well, that wouldn't be a problem for the Donohue-Levitt theory because it would just reduce the effect size in their statistical analysis. But if interstate migration was not neutral to perceptions of local differences in crime rates and/or morality (which can correlate with the abortion rate), the whole point of their state-level analysis could be in big trouble.

And, of course, in the real world, crime and morality are highly important drivers of migration. People with kids move out of big cities, in part to escape crime and big city immorality, while well-educated young people flock to big cities with liberal morals to meet other young singles with liberal morals. Immigrants from cultures with close-knit extended family morals, making them partly immune to the temptations of the street, move to big cities and help drive socially laxer African-Americans out of town. Over the decades, these processes, which accelerated over the last few decades, have had an enormous cumulative effect on who lives where.

These biases thus render the Donohue-Levitt state-level analysis unreliable.

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

No comments: