Did Legalization of abortions reduce crime?
In their highly influential (and controversial) study, economists John Donohue and Steven Levitt argue that the legalization of abortion in the United States in the early 1970s caused a big crime drop roughly two decades later.
According to them, children whose births were not wanted are more likely to grow up in adverse family environment and engage in criminal behavior, and the legalization of abortion reduced the number of such high-risk individuals among population born after the early 1970s. Consistent with these observations, they find that states with high abortion rates in the 1970s and 1980s experienced larger crime declines in the 1990s.
While simple and intuitive, their finding has been challenged by several other economists. Ted Joyce (2004) and Christopher Foote and Christopher Goetz (2008) argue that the estimated crime-reducing effect of abortion becomes much weaker under alternative regression specifications. Donohue and Levitt responded to these criticisms by arguing the result remains robust when different data and regression specification are used. The technical details surrounding the debate are out of the scope of this course, but I would like to introduce you to one part of the debate, which deals with an important limitation of the fixed-effect regression.
A simplified version of the regression specification used by Donohue and Levitt compares the rates of abortion and crime across different states in the following way:
Under this specification, the crime rate in state in year () is assumed to be a linear function of the abortion rate in state up to year (), observable socioeconomic characteristics of state in year (), and state fixed effect ().
What does the state fixed effect do again? Its inclusion in the regression specification addresses the concern that observable data () may not be able to fully describe important differences between states. For example, the difference in violent crimes between Texas and Utah may be partly driven by the difference in the tendency toward self-defense and gun use between people in Texas and Utah, but the data on this important difference would be very hard to find. To address this concern, we include the state fixed effect to control for all time-invariant differences across states, and instead focus on the within-state variation in abortion and crime.
But perhaps this is not enough. When looking at multiple states over a long period of time, there may be important unobservable, time-varying differences across states. The state fixed effect takes account of the unobservable, time-invariant difference between states, but it is likely that some of these unobservable differences would change over a long period of time, which cannot be taken account of by state fixed effects.
Part of the debate surrounding the abortion theory had to do with how best to account for this unobservable time-varying difference in criminal environment across states.
Since the empirical analysis on this abortion theory is usually based on a few decades’ worth of data, it is likely that the unobservable differences between states must have changed somewhat over time. In short, when analyzing data over a long period of time (say, a few decades), a simple fixed-effect regression using panel data is usually inadequate to remove time-varying, unobservable differences between states. Unfortunately, there is no simple way to take account of this and the debate remains unsettled to this day.
- Donohue, J. J., and Steven D. Levitt. “The Impact of Legalized Abortion on Crime.” Quarterly Journal of Economics 116.2 (2001): 379-420.
- Foote, Christopher L., and Christopher F. Goetz. “The Impact of Legalized Abortion on Crime: Comment.” Quarterly Journal of Economics (2008): 407-423.
- Joyce, Ted. “Did Legalized Abortion Lower Crime?” Journal of Human Resources 39.1 (2004): 1-28.
© Songman Kang, Hanyang University