Data re-evaluation shows abortion
associated with increased murder rate
Abortion Link to Crime Decrease Debunked
By Randall K. O'Bannon, Ph.D. and Laura Antkowiak
The May 2001 issue of Harvard University's Quarterly Journal of Economics featured the final version of a controversial paper claiming that legal abortion accounted for as much as half of the reduction in crime during the 1990s.
The paper argued that when fewer kids are born into the sort of social and economic conditions that supposedly "breed" criminals, fewer crimes are committed when these kids come of age.
Now, however, a Yale law professor and an Australian economist have reexamined the data and found several major flaws in the original analysis. Among other problems, the study was thrown off because it did not account for all abortions performed and because it did not adequately consider criminals by age group.
Crime did fall in the 1990s, but the late '80s and early '90s saw a dramatic increase in homicide rates among the young. Ironically - - given the very favorable treatment the original study received in the media - - these researchers say, if anything, abortion is associated with more, not less crime, and with a higher murder rate in particular.
The Donohue-Levitt Thesis: Abortion Decreases Crime
In their paper, "The Impact of Legalized Abortion on Crime," Stanford University law professor John J. Donohue III and University of Chicago economics professor Steven J. Levitt give what appears to be solid statistical backing to what has long been implicit in arguments the pro-abortion movement and its sympathizers among the cultural elite: abortion helps rid society of "those persons" who might later cause it problems.
Donohue and Levitt claim they stumbled on this thesis by trying to look for factors that would explain the large drop in crime seen in the United States in the 1990s. They chose to look at abortion. "I was just stunned at the magnitude of the abortions relative to births," Donohue told a reporter from the Chicago Tribune (8/8/1999). "It's such a huge number that it has to have had some big impact somewhere."
Donohue and Levitt say that the increase of abortions in the 1970s corresponds to the decline in crime seen in the 1990s, just about the time those babies would have turned 17 and entered an age group characterized by especially high crime rates. Even more significantly, Donohue and Levitt argue, that drop in crime began to show up first in the five states that legalized abortion prior to Roe v. Wade. Further, they contend that crime dropped most dramatically in those states with the highest abortion rates.
Donahue and Levitt assert that legal abortion explains as much as half the reduction in crime between 1991 and 1997. How do they come to this conclusion?
They compare the change in crime rates between 1985 and 1997 among three groups of states they label high, medium, and low abortion states. They find that during this time period, in the low abortion states, violent crime rose by 29.2%, property crime by 9.3% percent, and murder by 4.1% percent. Meanwhile, in the highest abortion rate states, violent crime fell by 2.4%, property crime by 23.1%, and murder by 25.9%.
Based on these figures, Donohue and Levitt estimate that crime was nearly 15-25% lower in 1997 than it would have been without legal abortion. Overall, between 1991 and 1997, Donahue and Levitt report that murder rates declined by more than 40% and violent and property crime by more than 30%. The two then conclude that abortion could be responsible for up to 50% of the total decrease in crime between 1991 and 1997.
Levitt and Donohue say their study suggests that the abortion effect, if unhindered, could yield a 1% annual drop in the crime rate for the next two decades. They complain, however, that "To the extent that the Hyde Amendment effectively restricted access to abortion, however, this prediction might be overly optimistic."
Though they claim to have no agenda, the language Donohue and Levitt use is remarkably similar to that of the abortion lobby. Levitt explicitly tied all the pieces together for a New York Times reporter earlier this year. "The thesis is about as simple as you could have," Levitt told the Times (4/14/01). "A difficult home environment leads to an increased risk of criminal activity. Increased abortion reduced unwantedness and therefore lower[ed] criminal activity."
The Lott-Whitley Rebuttal: Abortion Increases Murder Rate
Though some academics and commentators expressed doubts and reservations about Donohue and Levitt's thesis, it fell to Yale University law professor John Lott and Australian economist John Whitley to actually reexamine the data and some of the assumptions Donohue and Levitt had packed into their statistical analysis.
Donohue and Levitt made a number of miscalculations right at the beginning, Lott and Whitley point out. Though abortion was indeed legalized in New York, California, Alaska, Hawaii, and Washington prior to the 1973 Roe decision, there were also a sizeable number of abortions being performed in other states where abortion was legal for the life or "health" of the mother before 1973. Indeed, several of these states had abortion rates as high or higher than those states where abortion was legalized.
For whatever reason, Donohue and Levitt treat states other than the five they focus on as if no abortions had been performed. In fact thousands of abortions took place. This leaves unanswered the obvious question: did their crime rates also drop?
Donohue and Levitt also failed to adequately disaggregate (separate out) those crime statistics by the age, race, or sex of the offender. In failing to do so, Donahue and Levitt are in danger of mistakenly attributing reductions in crime to the impact of abortion among generations that were too old to havew been aborted in a particular year.
Lott and Whitley attempt to replicate Donohue and Levitt's analysis over all states. They refine the data to account for all recorded abortions prior to 1973 and correct for how the age of criminals changed over the years and across states. The revised statistical analysis now shows not a decline, but a slight increase in overall crime due to abortion.
Donohue and Levitt examine three crime categories: violent crime, property crime, and murder. Lott and Whitley direct their focus to murder rates, having found only a very small and unclear relationship between abortion and the other two types of crime. In contrast to Donohue and Levitt, they find no difference in the rise and fall of murder rates between the states that first legalized abortion and the rest of the country for the years 1976-1998. On average, however, the early- legalizing states experienced consistently higher homicide rates in nearly every year between 1976 and 1998.
If Donohue and Levitt had been correct - - that abortion was driving a decline in murder - - murder rates in the states legalizing abortion three years before Roe should have started climbing and then started falling three years before murder rates in the rest of the country started climbing and started falling. The data show this did not happen.
Recall Donohue and Levitt's basic argument that the children born after legal abortion are more "wanted" and have mothers who can provide better home environments. They theorize that these children will be less criminal. If Donohue and Levitt are correct, crime rates should drop first among the youngest age groups. Children killed by abortion in 1970 would have turned 10 in 1981. In 1981, we should see crime rates drop among the " wanted" 10-year-olds. As each year passed, we would see crime rates drop among 11-year-olds, 12-year-olds, and so on.
Instead, Lott and Whitley find the opposite story: across the country, the decline in homicide actually began among generations born before Roe.
The 10- to 15-year-old age group, the youngest examined by Lott and Whitley, was the last age group to see its homicide rates decline. Further, national murder rates among people in their late teens skyrocketed from the mid-1980s to the early '90s, just when the first victims of abortion would have been in that age group.
Employing what they believe to be a more accurate statistical model, Lott and Whitley next discover a strong association between abortion and an increase in murder. They measure the number of abortions by age, state, and year against the number of murders committed by individuals of a specific age and state in a given year, also considering in this equation variables measuring a state's economy, its law enforcement, and some population characteristics.
This type of analysis finds higher numbers of abortion to be correlated with a slight but significant increase in murders. Even with many other factors that could impact crime considered in this analysis, abortion is identifiably connected to increased murder.
In addition to those unborn lives lost due to abortion, Lott and Whitley say legalization could have meant as many as 1,850 more murders in 1998 than would have occurred otherwise.
Unlike Donohue and Levitt, Lott and Whitley offer no broad conclusions about the cause of this increase. They mention theories about the legalization of abortion increasing pre- marital sex and therefore the raising of children by single mothers, as well as the general "coarsening" effect of abortion on culture, but ultimately declare the validation of these theses to be beyond the scope of their paper.
Lott and Whitley simply assert that their analysis shows there are "many factors that reduce murder rates, but the legalization of abortion is not one of them. Of the over six thousand regressions that we estimated here, only one regression implied even a small reduction in the murder rate.
All the other estimates implied significant if very small to modest increases in the murder rates: legalizing abortion would increase murder rates by around 0.5 to 7 percent."Essentially, the message is, if you choose death, you get death. With interest.
[Editor's note. You can read the Lott/Whitley paper online at http://papers.ssrn.com/paper.taf?abstract_id=270126.]
Different Versions, Same Old Story
Donohue and Levitt went through at least two versions of their paper before putting it in the final form published in the May 2001 Quarterly Journal of Economics, so it is worth looking at what the changes they made tell us about their presuppositions and sensitivities.
Early on, Donohue and Levitt were much more direct in identifying the racial and economic factors involved in their analysis. "Teenagers, unmarried women and African Americans are all substantially more likely to seek abortions," Donohue and Levitt wrote in the 1999 version of their paper. "Children born to these mothers tend to be at higher risk for committing crime 17 years or so down the road, so abortion may reduce subsequent criminality through this selection effect."
The authors devoted considerable attention to efforts seeking to show why "selective abortion" and the "improved environment" women can provide for children if abortion is available are far more powerful explanations for abortion's effect on crime than the mere reduction in population. By 2001, the authors had removed virtually all these sorts of references to "selective abortion" from their paper. They also decreased their references to race, using instead the euphemism "economically disadvantaged."
Interestingly, Donohue and Levitt reach the same conclusion in 2001 as they did in 1999 - - that abortion could be responsible for as much as half the reduction in crime during the 1990s - - in spite of changes in the data and in the statistical models on which they base their conclusions. The 1999 paper was greeted with a lot of hype, but an independent look at their 1999 data tables showed that most of their results were statistically weak.
Donohue and Levitt nevertheless defended their claim because they said the majority of their tests continued to indicate a relationship between abortion and the drop in crime in the 1990s, and because they found no relationship between the two in the years prior to 1985.
In the final version, Donohue and Levitt calculate their abortion numbers in a different way. They substitute data for the actual abortion rate with a murky concept they call the "'effective' abortion rate."
Lott and Whitley's paper discusses some of the statistical problems with this measure, one of which is that Donohue and Levitt adjust abortion rates with a national 1985 arrest figure that they apply to all states and years. [Lott and Whitley say that using the national average for one year is problematic because there was actually a lot of variation in the arrest rates by age from state to state and year to year. Additionally, many crimes do not result in arrests, and many arrests do not occur in the same year a crime is committed.] Following this and other changes, Donohue and Levitt more confidently report stronger results in 2001.
Problems lie not just in what Donohue and Levitt did with their numbers. At least one scholar noted problems with Donohue and Levitt's initial data itself. Says economist Ted Joyce of Baruch College and the National Bureau of Economic Research, "Can you tease out that effect [of abortion] from crude aggregate data on crime 15 to 20 years later? I don't think so."