John Lott has this very interesting article about the thesis that Roe v. Wade reduced crime. The thesis has been most prominently advanced by economists John Donohue and Steven Levitt. They argue that Roe lead to declining crime rates by reducing the number of "unwanted" children. These unwanted children would have entered their crime-prone years shortly after 1990 — thereby explaining why crime rates began declining in the early 1990s.
Lott has some interesting responses. I found particularly powerful Lott's rejoinder that, when crime-rate declines are analyzed by age group, the Roe theory falters:
The "abortion decreases crime" theory runs into even more problems when the population is analyzed by age group. Suppose that liberalizing abortion in the early 1970s can indeed explain up to 80 percent of the drop in murder during the 1990s, as Donohue and Levitt claim. Deregulating abortion would then reduce criminality first among age groups born after the abortion laws changed, when the "unwanted," crime-prone elements began to be weeded out. Yet when we look at the declining murder rate during the 1990s, we find that this is not the case at all. Instead, murder rates began falling first among an older generation — those over 26 — born before Roe. It was only later that criminality among those born after Roe began to decline.
Lott goes on to contend that Roe actually increased crime, by increasing out-of-wedlock births. Social science has established a very clear link between single-parent families and crime. Lott concludes that any reduction in crime because of fewer crime-prone "unwanted" children was far outweighed by the increase in single-parent families. His conclusion seem reasonable to me.