So, there is a chapter in Freakonomics on the line of research of Steve Levitt's that I was most familiar with already: the claim that an important cause of the large drop in crime in the 1990s was Roe vs. Wade and the births it prevented. To whatever extent I have heard this theory mentioned in sociological circles, which hasn't been often, it seems like the Good Liberal reaction is to think that it is absolutely ludicruous. To be honest, I'm not exactly sure why this is. Maybe there is the idea that In The Hands Of Conservatives, the argument could be used to support something heinous, like a forced sterilization program for mothers in unfortunate circumstances (which would, of course, be unequivocally wrong, but it would be wrong regardless of whether it would have any long-term consequences for crime rates).*
In any case, when I first read about the theory in a newspaper account shortly after the original research article was published, I mentally classified it as "intriguing but farfetched." It was intriguing to me since I am a fan of The Underappreciated Consequences of Demography as a causal trope. However, I'm generally skeptical of the trope of the Surprising Distal Cause for A Major Proximate Change. Besides, I'm generally a hard sell when it comes to any inferences about the causes of social change, and, just as I'm skeptical of bands that are too popular, I'm also skeptical of any theory that is catchy enough to garner prominent newspaper coverage.
Since the theory has nothing to do with any area of social research in which I'm personally engaged, I didn't give the theory much further consideration, but the "intrigued" part of my assessment did keep the theory in mind. Then, through a weird set of circumstances, last year I read an exchange on the theory that had appeared in the Journal of Human Resources (2004, issue 1). That exchange basically moved my assessment from "farfetched" to "plausible" to even acquiring something like a mental preponderance so long as one didn't push the theory too far (i.e., so long as one recognized other causes of the crime drop as well). It's not that I am ready to assert that the explanation is true, and I could certainly be convinced that it is false, but it would take something more than any evidence or counterarguments that I have seen so far.
Anyway, Freakonomics has a summary of the evidence for the causal inference that legalized abortion has resulted in a drop in crime rates. There are six main pieces of evidence mentioned. I'll list them along with my own rating on a scale of 0 -10 scale of how compelling I regard each to be (at least if true; in other words, if I learned that Levitt was overstating the evidence on a given point, my ultimate assessment would be revised downward accordingly). Here:
Exhibit A: "In the early 1990s, just as the first cohort of children born after Roe v. Wade was hitting its late teen years--the years during which young men enter their criminal prime--the rate of crime began to fall." (While this isn't discussed in Freakonomics, the JHR debate, incidentally, also includes evidence for the point that Roe-v-Wade did increase substantially increase the overall number of abortions, as opposed to just having legal abortions mostly replace illegal abortions, which is of course necessary for the Donahue/Levitt argument to go anywhere.)
Exhibit B: "In New York, California, Washington, Alaska, and Hawaii, a woman had been able to obtain a legal abortion for at least two years before Roe v. Wade. An indeed, those early-legalizing states saw crime begin to fall earlier than the other forty-five states and the District of Columbia..." [4; would be more if it was a better mix of states]
Exhibit C: "...the states with the highest abortion rates in the 1970s experienced the greatest crime drops in the 1990s, while states with low abortion rates experienced smaller crime drops. (This correlation exists even when controlling for a variety of factors that influence crime: a state's level of incarceration, number of police, and its economic situation.)" 
Exhibit D: "Moreover, there was no link between a given state's abortion rate and its crime rate _before_ the late 1980's--when the first cohort affected by legalized abortion was reaching its criminal prime..." [when added to Exhibit C, the 2 that I gave that becomes a 6-7]
Exhibit E: "In states with high abortion rates, the entire decline in crime was among the post-Roe cohort as opposed to older criminals." 
Exhibit F: "[S]tudies of Australia and Canada have since established a similar link between legalized abortion and crime." [6, provided that there is variation in the years these countries legalized abortion]
Again, especially consider how far the area is from my own areas of expertise, I can be swayed that I have been overly credulous here (or that am not giving the theory enough credibility). If you know something I don't here, please let me know.
* It's one thing that, in so much of sociology, the general expectation is that assessments of empirical claims are to be based first on their consistency with the prevailing political currents of the discipline before turning to consideration of the quality of the evidence itself. More irritating, however, is the extent to which the consideration of empirical claims turns first on how the claim could be interpreted In The Hands Of Conservatives. I would be able to afford at least a couple months of luxury apartment life in Boston if only I had a dollar for every time I heard a sociologist offer a counterargument along the lines of "But, if that were true, couldn't somebody say [politically reactionary thing]?" with the seeming belief that such a counterpoint has roughly equal epistemological weight to, say, actual empirical evidence. As if the overriding consideration for what is accepted as disciplinary wisdom should be its relative imperviousness to any unsavory FOX-news-spin.