So, here’s Democratic pollster Stanley Greenberg, who confirmed the partisan power of the Marriage Gap in January, reporting on his regression analysis of the individual level exit poll data.

“The marriage gap is one of the most important cleavages in electoral politics… The marriage gap is a defining dynamic in today’s politics, eclipsing the gender gap, with marital status a significant predictor of the vote, independent of the effects of age, race, income, education or gender.”

According to Greenberg, the exit poll showed Bush carried merely 44% of the single white females but 61% of the married white women—a 17 point difference. Among white men, Bush won 53% of the singles and 66% of the married—a 13 point difference.

By the way, Mickey Kaus blogged yesterday on Slate.com:

“Steve Sailer has boiled down the explanation for why some states become red and others become blue to three simple words. (“God” is not one of them.) … His equation sure works for San Francisco. … 6:01 P.M.”

My summary article is at: http://blog.vdare.com/archives/2005/05/08/affordable-family-formation-the-neglected-key-to-gops-future/

And it contains dozens of links to supporting documentation, like the above quote from Greenberg.

This is the Ecological Fallacy. You can’t infer that because Bush got a higher proportion of the vote in states with a high proportion of people of type X, that people with a high proportion of people of type X voted for Bush.

This may be true, but you need to demonstrate it with individual level data.

Question – “Is the Transportation Tax disproportionably spent in White Areas in order to drive development? Is the application of the tax racist? Most “great” cities in America are on the coast or are placed against a large natural barrier that restricts sprawl; thereby making people work out their problems, instead of building roads to allow people to flee the poor and brown people. Please look at data.”

Excellent question. I’ve been looking at the data since the last Presidential election and this geographic division between waterside liberal cities and conservative inland areas—what I call “dirt gap”—appears to underlie the famous division of the US into red (Republican) and blue (Democratic) states.

Let’s look at the 50 most populous metropolitan areas in the USA. Of the ones in blue states, 73 percent of their population lives in cities, such as New York, Los Angeles, and Chicago, where physical growth is restricted by unbridgeable water, compared to only 19 percent of the population of the biggest red state metropolises, such as Dallas, Atlanta, and Phoenix.

Why is this important? The Law of Supply and Demand controls housing prices. The greater supply of available land for suburban expansion in red metropolises keeps house prices down.

This leads to the Mortgage Gap. Bush carried the 20 states with the cheapest housing costs, while Kerry won the 9 states with the most expensive. The Mortgage Gap has been growing. Bush was victorious in the 26 states with the least home price inflation since 1980. Kerry triumphed in the 14 states with the most.

This, in turn, affects “affordable family formation.”

Bush did much better in states where non-Hispanic white people get married and have children.

The Marriage Gap: The single best correlation with Bush’s 2004 share of the vote by state that anybody has yet found is: the average years married by white women between age 18 and 44: an astonishing r-squared = 83 percent.(This has to be one of the highest r-squareds for a single unexpected factor ever seen in political science.) Bush carried the top 25 states ranked on “years married.”

The cost-of-housing index correlates with “years married” with an r-squared = 53 percent. Similarly, the housing inflation rate since 1980 and “years married” correlate at r-squared = 48 percent. Where people can afford to buy houses, they tend to get married more and at younger ages.

Finally, The Baby Gap: Bush carried 25 of the top 26 states in white total fertility (number of babies per white woman), while Kerry was victorious in the bottom 16. The correlation between white total fertility and Bush’s share produced an impressive r-squared = 74 percent.

The arrow of causality presumably points in both direction. Conservative, family oriented people tend to move to inland areas with cheaper costs of raising children well, so inland areas become more conservative. But, where somebody lives also affects their chance of getting married and the number of children they are likely to have. In regions where family formation is less affordable, voters are less likely to become family values voters.

You can read a fuller summary of this model of the US electoral map at http://blog.vdare.com/archives/2005/05/08/affordable-family-formation-the-neglected-key-to-gops-future/

It contains links to my other articles going into detail on each of the: Dirt Gap, Mortgage Gap, Marriage Gap, and Baby Gap.

Let me give this one more go.

My point isn’t about the specifics of Levitt’s argument, which I appreciate is about change, it’s about the methodology. That’s the reason I brought up the classic form of the Ecological Fallacy: that if areas with lots of people of type X have high Y, then it is wrong to infer that people of type X have high Y. Levitt doesn’t make this argument, his reasoning is at several removes from it – because he’s studying the absence of people and change in a rate. But I wonder if aggregation effects can have a similar impact on this argument.

If Levitt’s argument has the same structure as the classic Ecological Fallacy it should fail for the same reason. He says that areas with a high change in abortions (i.e. more “wanted” children) have a high fall in crime, therefore people who are aborted have high tendancy to commit crime. This is made in Figure 4.

I suspect that this may not be a valid inference for the same reason the classic form of the Ecological Fallacy fails.

I really enjoyed the discussion, particularly the delightfully self-effacing pieces from Tim and Steve. It set me thinking about the recent decline of economics in Australian universities. With the introduction of commerce and business degrees (which focus on marketing, accounting etc), the fraction of students studying economics is falling sharply. I think we could benefit dramatically from selling ourselves not as a moneymaking degree, but as the premier social science degree.

Currently, it’s difficult to see today that an 17 year old Aussie kid who wants to understand poverty, corruption or drug markets will study economics. More likely, they’ll choose sociology, social work or political science. From my time in the US, I get the sense that this isn’t as much of a problem there (perhaps because choosing a major happens midway through college rather than at the end of high school). But economics as a social science isn’t doing well downunder.

http://slate.msn.com/id/33569/entry/33571/

(By the way, when his publicist recommended to him a couple of months ago that we resume our debate this year to publicize his book, he turned her down. Prudently, I would add.)

The format was that Dr. Levitt went first, I second, and we’d each get one response.

Dr. Levitt presented his argument, then I pointed out that he had forgotten to look at homicide rates by narrowly-defined age groups, which showed the exact opposite of what he was claiming. The generation born right after legalization went on the worst violent crime spree in history. I wrote:

“Why, then, is this generation born in 1975 to 1979 now committing relatively fewer crimes as it ages? It makes no sense to give the credit to abortion, which so catastrophically failed to keep them on the straight and narrow when they were juveniles. Instead, the most obvious explanation is the ups and downs of the crack business, which first drove violent crime up in the late ‘80s and early ‘90s, then drove it down in the mid and late ‘90s. That’s why the crime rate has fallen fastest exactly where it had previously grown fastest as a result of crack—in the biggest cities (e.g., New York) and among young black males. This generation born right after legalization is better behaved today in part because so many of its bad apples are now confined to prisons, wheelchairs, and coffins. For example, over the last two decades the U.S. has doubled the number of black males in prison, to nearly 1 million.

“More encouragingly, the biggest decline in murder from 1993 to 1997 was among the newest generation of black males aged 14 to 17. These kids born mostly in the early ‘80s survived abortion levels similar to those faced by the crime-ridden 1975-to-1979 generation. Yet, their murder rate in 1997 was less than half that of the 14- to 17-year-olds of 1993. Seeing their big brothers gunned down in drive-by shootings and their big sisters becoming crack whores may have scared them straight.

“Admittedly, it’s still theoretically possible that without abortion the black youth murder rate would have, say, sextupled instead of merely quintupling. Still, there’s a more interesting question: Why did the places with the highest abortion rates in the ‘70s (e.g., NYC and Washington D.C.) tend to suffer the worst crack-driven crime waves in the early ‘90s?”

Dr. Levitt’s reply, which he recalls as having “destroyed his core argument,” in fact completely dodged my final question.

I wrote a draft reply brusquely pointing out some of the logical and empirical shortcomings of his response and emailed it to him first, as a courtesy before sending it to Slate.com. He made clear to me that his feelings were hurt by my reply.

He seemed like a nice guy, and it seemed to me that only people who wanted to believe his abortion-cut-crime theory in the first place would find his convoluted response terribly persuasive.

So, I rewrote my draft to focus on what has been the stronger appeal to the public of his theory: his mechanism for why abortion would cut crime—by pre-emptively eliminating less “wanted” children. While this simplistic model makes sense when you first hear it, I pointed out the reasons why it was much less plausible when you take a more sophisticated view of human behavior.

The big problem with his wantedness theory is that, as he admits in “Freakonomics,” legalizing abortion drove up the number of unwanted pregnancies by tens of millions over the last 35 years. As he admits, the birthrate only went down by 6%, but the pregnancy rate went up by almost 30%. This makes his simplistic wantedness mechanism look dubious for reasons I explained to him back in 1999.

Dr. Levitt has failed to address these criticisms, and continues to hype his “wantedness” theory even though he has admitted, such as in an interview with Tim Harford, that legalizing abortion greatly increased unwanted pregnancies.

Nut graf:

Interestingly, at the time, Sailer refused to respond directly to my arguments. His response in Slate completely side-stepped the fact that I had destroyed his core argument. He wrote, for instance, “…rather than mud wrestle in numbers here, I’ll privately send you my technical suggestions. In this essay I’ll step back and explain why this straightforward insight [that abortion reduces crime] might not work in practice.” I should note that I am still waiting for those technical suggestions he promised to arrive!! And if you compare his Slate arguments to his “new” article in the American Conservative, you will see that his thinking has not progressed very far on the issue.

http://www.iSteve.com/abortion.htm

Considering how little critical thought has been applied to Dr. Levitt’s abortion-cut-crime theory, he has obviously tapped into a deep hunger to believe that legalizing abortion had a good social effect

What Dr. Levitt wants you to ignore is that the first generation born after the legalization of abortion went on to commit more homicides and other serious violent crimes as 14-17 year olds than any other generation for which we have data.

Moreover, the youth violent crime rates went up first, in the later 1980s, in those metropolises that legalized abortion, de jure or de facto, in 1970, three years ahead of the rest of the country: NYC, LA, DC. The huge wave of crack killings by 14-17 year olds then spread to the rest of the country in the early 1990s, peaking in 1993 and remaining very high in 1994, before dropping sharply.

And the violent crime rate went up most among the demographic group that had been the heaviest consumer of legalized abortion during the early years of legalization: urban blacks. In 1993, the homicide rate for the cohort of black males born in the second half of the 1970s (after legalization) was more than four times higher than for the cohort of black males born in the second half of the 1960s, before legalization.

Correlation does not prove causation, but it’s safe to say that there is about as much evidence for the theory that abortion helped drive up the juvenile serious violent crime as that it drove it down.

Large assertions require large proof, and Dr. Levitt hasn’t come close to meeting the burden of proof.

I think John Quiggen has it right: there is a unifying theme, and that theme is the need to base public policy on actual data. I read the book as a plea for greater empiricism in policy-making. The role of incentives in motivating behavior seems to me to be a secondary theme, central to Steve’s worldview maybe, but not developed as thoroughly in this book.

Where incentives are discussed, the message seems to be one of caution: warning us that it is far too easy for policy-makers to develop perverse incentives because the motivators of human behavior are more complex than we typically realize. Again, a good message and one that points to the importance of empirically testing one’s hypotheses before implementing them as policy.

I think the book offers a similar caution about conventional wisdom and other non-data-based arguments for policies. You may be able to make a solid, logical argument for why a particular policy will “cause” a particular desired “effect.” But no matter how sound or intuitive the logical argument is, the data needs to back it up.

I wish the book offered two additional cautions:

1) As important as data is, we need to be very careful to distinguish strong correlation from causation when we interpret data. In fairness, the book touches on this point, but I don’t think it goes far enough. I’d like to see more on this topic—maybe in a future book.

2)The fact that policies that are proved effective by empirical data are not necessarily the best possible policies. Reading the book, one could conclude that the best way to end crime is to reduce civil liberties, hire more police, build prisons, lock up more people for longer times, and encourage abortions. The data suggest that this would lower crime. But there must be a better approach, one that focuses on root causes and eliminates the need for these more draconian approaches. Let’s explore the data to help us find those more humane approaches, whatever they may be!

One other quick point: The book made me recall mathematician John Allen Paulos’s books on numerical illiteracy and its consequences—particularly the chapters that deal with Americans’ poor understanding of statistics and probability. I’d recommend these as supplemental reading to anyone who hasn’t read them yet.

The attribution that if countries with more Protestants tend to have higher suicide rates, then Protestants must be more likely to commit suicide, is false. The same logic applies to the attribution that if countries with increasing number of Protestants tend to have increasing suicide rates, then increasing Protestantism is causing more suicide. Ecological effects are a result of aggregation, and so have influence in cases more wide ranging than in the simple Durkheim example I quoted.

I’m wondering whether part of Levitt’s analysis is at heart about making inferences on individuals from aggregate data, and is affected by the ecological fallacy because of this. If you want make the link that people more prone to commit crime are selectively aborted, surely you have to use individual, not aggregate level, data in order to do this.

This is just a tentative suggestion, I’m not stupid enough to start pronoucing on things based on my shallow reading in this area. The reason I’m interested is that – in my training – the first thing you’d discuss is the Ecological Fallacy (and the Modifiable Areal Unit Problem, though this doesn’t apply in Levitt’s case). It feels strange to me that this isn’t mentioned at all and I’m still interesting as to whether it has been raised anywhere.

(1) I’m not sure how far Steven’s work is “data driven”. It obviously is more empirical than deriving models from theoretical premises, but from what I’ve seen of it his work seems to me to be hypothesis driven. He comes up with an idea, finds some data, and tests it. An example of “data driven” research is the sort of thing Danny Dorling does, getting social data and then using exploratory methods to tell us what it says about the world.

(2) I’ve read the abortion-crime paper, though I’ve yet to follow up the subsequent discussion. It strikes me that part of the data analysis is vulnerable to the Ecological Fallacy (states with low crime have high abortions, therefore this is because of an absence of the sort of people who commit crime, thanks to abortion). This sort-of-parallels Durkheim’s classic example that if countries with more Protestants tend to have higher suicide rates, then Protestants must be more likely to commit suicide. This issue isn’t addressed in the paper, is it followed up anywhere?

I devoured the book and imho the unifying theme of the book is telling the truth (or tricking the truth) out of the data. The other unifying or at least underlying theme appears to look broadly at the role of parenting and the consequences of selecting to avoid parenting. Loved it.

I am hoping that you will continue to ask questions and trick the data.

Question –
Is the Transportation Tax disproportionably spent in White Areas in order to drive development? Is the application of the tax racist?

Most “great” cities in America are on the coast or are placed against a large natural barrier that restricts sprawl; thereby making people work out their problems, instead of building roads to allow people to flee the poor and brown people.

Please look at data.

Detroit would be the only exception to the rule, but may be not. Maybe due to the Auto Industry, the road construction dollars far exceed the natural barrier that Lake Erie provides.