In Freakonomics, John Donohue and Steven Levitt argued that legalized abortion led to a dramatic decline in crime during the 1990s.
A new NBER paper by Theodore J. Joyce of Baruch College finds little evidence to support the Donohue/Levitt hypothesis.
Donohue and Levitt?s presentation of the evidence relied too much on highly-aggregated regressions, run over limited time periods, with questionable specifications and no acknowledgement of the endogeneity of abortion. If they had started with simple time-series of age-specific arrest and homicide rates pre and post the points of abortion legalization, they would have been confronted with the lack of discontinuities consistent with large cohort effects.