Since Donohue and Levitt's 2001 QJE article The Impact of Legalized Abortion on Crime (PDF), and in particular since the publication of Freakonomics, there has been quite a controversy in the United States about a possible causal relationship between higher abortion rates and the fall in crime. But what about other countries? A new paper by Leo H. Kahane from California State University, David Paton of the Nottingham University and Rob Simmons from Lancaster University to last week's ESPE 2006 conference in Verona adds to the evidence.
The Abortion-Crime Link: Evidence from England and Wales (PDF) uses panel data from 1983 to 1997 for the 42 police force areas in England and Wales to test the hypothesis that legalizing abortion contributes to lower crime rates. The authors have two robust findings:
Of the many results presented in this paper, there are two things we can say with confidence. First, increased unemployment rates are associated with increased crime. Second, the greater the proportion of children in care (our proxy for ‘social deprivation’), the greater the crime rate. These results are remarkably consistent and robust to various estimation methods.
But they are much more equivocal on the abortion-crime link:
We are unable to say with confidence, however, that abortion legalization in the U.K. significantly reduced crime in England and Wales some twenty years hence. We come to this conclusion by first noting ..that total recorded crime in the U.K. began to decrease at about the same time as in the U.S., despite the fact that abortion legalization occurred about five years earlier. Thus we have a discrepancy in the timing of the potential effect of abortion on crime between the U.S. and the U.K. On the other hand, regression models linking effective abortion rates in the U.K. and subsequent recorded crime suggest the same negative and significant correlation between the two variables (at least for total crime and some sub-categories) as that reported for the U.S. by D & L.
When the authors used a similar model and estimation methodology to Donohue and Levitt, they "are able to replicate the negative association between abortion rates and reported crime found by for the U.S." But the authors recognise this negative correlation may be spurious, and so test for endogeneity:
We test this latter explanation by modeling abortion rates as endogenous to crime. In this approach, we identify abortion rates using on the proportion of abortions provided in the patient’s area of residence and the proportion provided free of charge. When abortion is treated as being endogenous, the negative correlation between abortion rates and crime disappears.
Though finding "no clear connection between the two", they add some important caveats:
Lastly, it must be noted that few of our results are robust to different specifications and samples. For example, the significant negative correlation between abortion and crime rates in the exogenous models does not hold for some sub-categories (e.g. violent and sex crimes) and is not robust to some other specifications (excluding London, area-specific trends and first-difference estimation).
Similarly, although the finding of no significant correlation between abortion and crime in the endogenous models is somewhat more robust to our various experiments, we do find evidence of a negative correlation between abortion and violent crime in some specifications.
An unsually frank admission. Finally, they highlight the problems facing researchers in this area:
The fragility of the results in this paper serve to emphasize the difficulty researchers have in identifying causal effects of social change such as abortion legalization on crime rates some years hence, particularly given the myriad of other social changes occurring over the same time and which may dilute any effect.