I was thinking recently about getting Freakonomics, the super popular book by the economist Steven D. Levitt and Stephen J. Dubner, despite annoyance at the “rogue economist” pose adopted by a totally institutional figure (he teaches economics at U. Chicago, for God’s sake!) applying archiorthodoxe economic analytical premises and methodology to standard market questions. But I come to find out soon after that the most important contention in the book, that the legalization of abortion has caused crime to decrease, had already erupted in factual despute a full six years ago. Here is a debate between Levitt and his primary critic on this issue, Steve Sailer (you have to click on the different days to read all the responses and counter-responses). It’s a good example of how two presumably intelligent people with two different sets of statistics can talk entirely at cross-purposes.

Levitt shows a pretty strong correlation between falling crime and the advent of legalized abortion, being that crime rates began to fall 18 years after Roe v. Wade, that crime rates in the four states that legalized abortion three years earlier began to fall three years earlier as well, that states with higher abortion rates experienced a greater fall in the crime rate in the ’90’s than states with low abortion rates, and that the fall in crime for those years is restricted to the under-25 demographic. Quite suggestive, though hardly conclusive.

Then Sailer introduces a bit of a non sequiter. Rather than offering an alternative explanation for the decrease in rate of crimes as a whole, he chooses to concentrate only on the fluctuations in murder rate. Why, I’m not sure, though I seem to recall him in another article that he wrote claiming that he felt general crime statistics to be unreliable, since murder is virtually the only crime that almost by definition does not suffer much from under- or over-reporting. However, as I have indicated before, statistics can have comparative validity even if their absolute validity is suspect. In other words, a difference in numbers between different locations or groups or a change over time may indicate a definite trend or difference even if the figures themselves are dubious. If the same methodology is applied constantly, then changes or variations are probably caused by something real. The other problem is that focusing only on murder narrows both the number and type of cases, making the data more suspeptible to distortions that do not affect the whole spectrum of crimes. And in fact the principal factor that Sailer identifies, the crack epidemic of the late ’80’s and early ’90’s, is even by he himself acknowledged to be something of an aberration. Despite Levitt’s insinuations, he is not really suggesting that the crack problem had anything to do with abortion, but its statistical effect was probably disproportionately concentrated in a few types of crime: murder, drug trafficking (obviously), and maybe a few incidental categories like burglary and weapons charges. Probably few people, certainly not I nor Levitt, would dispute that Sailer’s analysis of what was happening with the murder rate during these years is probably correct, but it is precisely the factor he identifies as the primary culprit which makes it hard for me to believe that the murder rate is a better indication of the true state of crime during this period than general crime statistics (unless one could argue that the effect of abortion on crime would also be concentrated to the same degree in the same categories as crack, but there is no support offered for that possiblity).

However, rather than settling for a response of that nature, Levitt launches an even stranger non sequiter. He bizarrely insists that for Sailer’s hypothesis to be plausible the murder rate has to conform to the pattern of his general crime statistics over the time period in question and that for some reason there has to be a correlation between areas of high abortion rates and the epicenters of the crack explosion. These criteria are so obviously random and unrelated that I cannot believe them to be purely the result of confused thinking, particularly by an economist of good standing (who is reputed by his supporters to be a genius), and I am forced to conclude that it is probably a deliberate obfuscation of the issue on the part of Levitt. Levitt seems either blind or averse to the fact that Sailer never accepted the general crime statistics upon which he bases his analysis in the first place, and hence the two cannot even agree on the nature of the phenomenon that they are analyzing, let alone what caused it. Sailer is not so much disputing the reasons for the drop in crime as disputing that there even was a drop in crime. And yet one would expect the growth of an enormously profitable narcotics industry to temporarily overwhelm long-term structural changes that cause reduced crime, much as the advent of the heroin trade 30 years earlier caused a similar spike in mafia-type crime. And even Sailer acknowledges that the rise in the murder rate might have been even greater had abortion not been legal.

So Levitt seems to me to have a stronger case, but I have my doubts that abortion really decreases crime, not least because it is not certain that it is even an isolable phenomenon. Sailer very perceptively notes at the end that abortion has accompanied a broad number of social trends associated with the growth of the underclass which are not exactly conducive to decreased crime (see Dalrymple again). I think he is quite right that the number of unwanted babies aborted may be overwhelmed by the even greater number of babies conceived casually because it was assumed that thanks to abortion contraception was unnecessary and then not aborted for whatever reason. That, of course, we will probably never know. Secondly, I’m not sure that the abortion/crime statistics for other countries would back Levitt up. In much of Europe, for example, where abortion has been legal (except in Ireland) for at least as long as in the States, crime has been rising in recent years, esp. in the U.K., where crime rates have surpassed those in the U.S. in all major categories except murder. But then again, one can only debate that issue if one believes that crime in the U.S. has actually gone down in the first place…

3 Responses to “Orthodoxonomics”

  1. shonk Says:

    However, as I have indicated before, statistics can have comparative validity even if their absolute validity is suspect. In other words, a difference in numbers between different locations or groups or a change over time may indicate a definite trend or difference even if the figures themselves are dubious. If the same methodology is applied constantly, then changes or variations are probably caused by something real.

    This is, of course, absolutely correct, though the conditions you mention may not actually obtain in the case of crime reporting. The way in which crimes are reported and the data collated has changed in the last 30 years. The national databases and standardized reporting that is commonplace now didn’t exist 30 years ago. However, most people agree that this standardization should, ceteris paribus, lead to an increase in overall crime reporting. In other words, the fact that reported crimes are down probably means that actual crimes are down even more than the figures suggest.

    Of course, comparing crime reporting now to 30 years ago is different than comparing 1997 to 1985 (or 1991 or whatever) and it doesn’t really address the Levitt/Sailer debate, anyway. That having been said, Sailer’s obsession with murder rates to the exclusion of all else strikes me as somewhat bizarre, especially given that murder rates have historically always been more volatile than general crime rates. And it’s interesting to note that Levitt’s prediction of a continuing gradual decline in crime rates has, six years on, been borne out by the evidence (although there is some evidence that rates are beginning to stabilize).

  2. shonk Says:

    Oh, and I should mention that I’m also annoyed by Levitt’s “rogue economist” image when he’s a U. of Chicago prof applying standard econometrics and neoclassical economic analysis to various studies.

    Also, a while ago I linked to some more examples of “freaky economics.”

  3. mock Says:

    While I’m sure Levitt could’ve toned down some of the hyperpromotion, I get the feeling he didn’t really contribute much to the book, as it’s pretty clear that Dubner wrote everything and basically just dramatized Levitt’s previous research.

Leave a Reply

If your comment doesn't appear right away, it was probably eaten by our spam-killing bot. If your comment was not, in fact, spam (and if you're actually reading this, it probably wasn't), please send me an email and I'll try to extricate your comment from our electronic spam purgatory.